diego-ellis-soto commited on
Commit
1ca14ec
·
0 Parent(s):

Initial audited Biodiversity Access Score app

Browse files
.gitattributes ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ outputs/results/isochrone_basi_100m.gpkg filter=lfs diff=lfs merge=lfs -text
2
+ outputs/results/isochrone_basi_100m.rds filter=lfs diff=lfs merge=lfs -text
3
+ www/Combined_logos.png filter=lfs diff=lfs merge=lfs -text
4
+ www/Reimagining_San_Francisco.png filter=lfs diff=lfs merge=lfs -text
5
+ outputs/results/gbif_cell_species_100m.rds filter=lfs diff=lfs merge=lfs -text
6
+ outputs/results/grid_results_100m.gpkg filter=lfs diff=lfs merge=lfs -text
.github/workflows/deploy.yml ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ name: Sync to Hugging Face hub
2
+ on:
3
+ push:
4
+ branches: [main]
5
+
6
+ # to run this workflow manually from the Actions tab
7
+ workflow_dispatch:
8
+
9
+ jobs:
10
+ sync-to-hub:
11
+ runs-on: ubuntu-latest
12
+ steps:
13
+ - uses: actions/checkout@v3
14
+ with:
15
+ fetch-depth: 0
16
+ lfs: true
17
+ - name: Push to hub
18
+ env:
19
+ HF_TOKEN: ${{ secrets.HF_TOKEN }}
20
+ run: git push -f https://cboettig:$HF_TOKEN@huggingface.co/spaces/boettiger-lab/sf_biodiv_access_shiny/ main
.gitignore ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # macOS system files
2
+ .DS_Store
3
+ data
4
+ .Rproj.user
5
+ .positai
6
+ old
7
+ startup_benchmarks.csv
8
+ startup_benchmarks.png
9
+
10
+
11
+ # R session files
12
+ .Rhistory
13
+ .RData
14
+ .Ruserdata
15
+ .Rproj.user/
16
+ Renviron
17
+ .env
18
+ .env.*
19
+ .DS_Store
20
+
21
+
22
+ # Secrets / local environment
23
+ .Renviron
24
+ .env
25
+ .env.*
26
+ *.key
27
+ *.pem
28
+
29
+ # macOS
30
+ .DS_Store
31
+
32
+ # RStudio
33
+ *.Rproj.user
34
+
35
+ # Temporary / cache
36
+ tmp/
37
+ temp/
38
+ cache/
39
+ *.log
40
+
41
+ # Docker/local build clutter
42
+ docker-compose.override.yml
43
+
44
+ # Large local analysis outputs not needed by deployed app
45
+ outputs/figures/
46
+ outputs/tables/
47
+
48
+ # Optional: raw or intermediate data not needed by Shiny
49
+ data/raw/
50
+ data/intermediate/
51
+
52
+
Dockerfile ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FROM quay.io/jupyter/minimal-notebook:ubuntu-24.04
2
+
3
+ USER root
4
+
5
+ # Install R
6
+ RUN curl -s https://raw.githubusercontent.com/boettiger-lab/repo2docker-r/refs/heads/main/install_r.sh | bash
7
+
8
+ WORKDIR /code
9
+
10
+ # Install R packages first.
11
+ # This layer only rebuilds when install.r changes.
12
+ COPY install.r .
13
+ RUN Rscript install.r
14
+
15
+ # Copy application and data after packages are installed.
16
+ COPY . .
17
+
18
+ # Hugging Face Spaces
19
+ EXPOSE 7860
20
+
21
+ CMD ["R", "--quiet", "-e", "shiny::runApp('app.R', host='0.0.0.0', port=7860)"]
README.md ADDED
@@ -0,0 +1,209 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ title: Sf Biodiv Access Shiny
3
+ emoji: 📚
4
+ colorFrom: blue
5
+ colorTo: yellow
6
+ sdk: docker
7
+ pinned: false
8
+ ---
9
+
10
+ # SF Biodiversity Access Decision Support Tool
11
+
12
+ <img src="www/Combined_logos.png" width="60%">
13
+
14
+ An interactive Shiny app built for the **Reimagining San Francisco (RSF) Initiative** to explore how equitably San Francisco residents can access urban biodiversity across different transportation modes and socioeconomic contexts.
15
+
16
+ ---
17
+
18
+ ## What the App Does
19
+
20
+ Users select a location anywhere in San Francisco — either by clicking the map or geocoding an address — and choose one or more transportation modes and travel-time thresholds (5, 10, or 15 minutes). The app then:
21
+
22
+ 1. **Generates travel-time isochrones** — polygons representing the area reachable within the chosen time budget — using Mapbox (driving, walking, cycling, traffic-aware driving) and SF Muni GTFS data (transit, walk + transit).
23
+ 2. **Summarizes biodiversity** within each isochrone: GBIF occurrence records, unique species richness, and breakdowns by taxonomic group (birds, mammals, plants).
24
+ 3. **Summarizes greenspace access**: OSM greenspace coverage (%), distance to nearest greenspace via raster, and NDVI from Sentinel-2.
25
+ 4. **Summarizes socioeconomic and environmental justice context**: area-weighted median income and population from ACS Census block groups, CalEnviroScreen Cumulative Impact scores, and SF Environmental Justice community burden scores.
26
+ 5. **Computes the Biodiversity Access Index (BAI)** — a composite score benchmarked against citywide empirical distributions — and displays it as a spider/radar plot across seven dimensions: mobility access, route access, biodiversity potential, observation intensity, environmental quality, greenspace cover, and equity context.
27
+ 6. **Displays partner RSF Program Projects** as a toggleable map layer.
28
+
29
+ ---
30
+
31
+ ## App Tabs
32
+
33
+ | Tab | Description |
34
+ |-----|-------------|
35
+ | **Isochrone Explorer** | Interactive map, isochrone generation, BAI spider plot, summary table, and metric plots |
36
+ | **Isochrone Comparer** | Pick two locations (one mode + travel time each) and compare them side by side: two BAI spider plots plus a difference table (transit score, biodiversity score, BAI, closest greenspace). Loads only when opened. |
37
+ | **GBIF Summaries** | Filter GBIF records by taxonomic class and family within the isochrone; species richness vs. sampling effort plot |
38
+ | **Community Science** | Map and table of partner community organizations |
39
+ | **About** | Full methodology, data sources, transport mode descriptions, and BAI explanation |
40
+
41
+ ---
42
+
43
+ ## Transportation Modes
44
+
45
+ | Mode | Engine | Notes |
46
+ |------|--------|-------|
47
+ | Driving | Mapbox Navigation API | Free-flow road network |
48
+ | Walking | Mapbox Navigation API | Pedestrian paths and crossings |
49
+ | Cycling | Mapbox Navigation API | Dedicated cycle lanes where available |
50
+ | Driving with Traffic | Mapbox Navigation API | Real-time + historical congestion |
51
+ | Transit (GTFS) | gtfsrouter | SF Muni timetable-based stop-to-stop routing |
52
+ | Walk + Transit (Muni) | Mapbox + gtfsrouter | First-mile walk + Muni ride + last-mile walk, all within total time budget |
53
+
54
+ ---
55
+
56
+ ## Map Layers
57
+
58
+ All layers are toggleable in the map's layer control panel:
59
+
60
+ - **Income** — Median household income per census block group (ACS 5-yr)
61
+ - **Greenspace** — OSM parks and green areas
62
+ - **Greenspace Distance** — Raster showing distance (m) to nearest greenspace
63
+ - **RSF Program Distance** — Raster showing distance (m) to nearest RSF program polygon
64
+ - **RSF Program Projects** — Partner project areas from the RSF Initiative
65
+ - **Hotspots / Coldspots (KnowBR)** — Block groups with anomalously high/low species richness relative to sampling effort
66
+ - **Species Richness** — Unique GBIF species per census block group
67
+ - **Data Availability** — GBIF occurrence records per block group
68
+ - **CalEnviroScreen (CI Score)** — Cumulative environmental burden by census tract
69
+ - **SF EJ Communities** — SF Environmental Justice community burden scores
70
+ - **Transit Routes** — All SF Muni routes from GTFS shapes, colored by official SFMTA route color
71
+ - **Transit Stops** — All SF Muni stops with AM peak headway info
72
+ - **Isochrones** — Generated travel-time polygons
73
+ - **NDVI Raster** — Sentinel-2 NDVI cropped to the isochrone union
74
+
75
+ ---
76
+
77
+ ## Biodiversity Access Index (BAI)
78
+
79
+ The BAI is a composite indicator benchmarked against **citywide empirical distributions** (ECDFs across all SF census block groups), not just the current session. It comprises seven equally-weighted dimensions:
80
+
81
+ | Dimension | Variable | Direction |
82
+ |-----------|----------|-----------|
83
+ | Mobility Access | Transit stops / km² | Higher = better |
84
+ | Route Access | Unique Muni routes crossing isochrone | Higher = better |
85
+ | Biodiversity Potential | Unique GBIF species | Higher = better |
86
+ | Observation Intensity | GBIF records / km² | Higher = better |
87
+ | Environmental Quality | Mean NDVI | Higher = better |
88
+ | Greenspace Cover | % OSM greenspace area | Higher = better |
89
+ | Equity Context | SF EJ burden score (inverted) | Lower burden = better |
90
+
91
+ All axes scaled 0–1. BAI = unweighted mean of all seven standardized components.
92
+
93
+ ---
94
+
95
+ ## Repository Structure
96
+
97
+ ```
98
+ SF_biodiv_access_shiny/
99
+ ├── app.R # Main app (sources Rscripts/setup_unified.R at startup)
100
+ ├── Dockerfile # HF Spaces: install.r + shiny::runApp('app.R', …)
101
+ ├── install.r # R package list for Docker
102
+ ├── www/ # Static assets (e.g. app_pastel.css, logos)
103
+ ├── Rscripts/
104
+ │ ├── setup_unified.R # Loads all app data: local data/cached + HuggingFace fallback
105
+ │ ├── iso_metrics.R # Shared point→isochrone→metrics→BAI→radar functions (used by both isochrone tabs)
106
+ │ └── prep/ # One-off builds → data/output/ (see run_all_prep.R)
107
+ └── data/
108
+ ├── cached/ # Downloaded / runtime cache (often gitignored)
109
+ ├── output/ # Prep outputs (often gitignored)
110
+ └── source/ # Raw inputs for prep (e.g. RSF polygons, GTFS extract)
111
+ ```
112
+
113
+ > **Note:** The `data/` directory is gitignored. See setup instructions below.
114
+
115
+ ---
116
+
117
+ ## Cyberinfrastructure & Data Sources
118
+
119
+ | Dataset | Source | Format | How Loaded |
120
+ |---------|--------|--------|-----------|
121
+ | GBIF occurrences (SF) | Global Biodiversity Information Facility | Parquet | DuckDB spatial queries in server |
122
+ | Census block groups + ACS | US Census / tidycensus | `.Rdata` | Downloaded from HuggingFace at startup |
123
+ | NDVI raster | Sentinel-2 (pre-processed) | GeoTIFF | Downloaded from HuggingFace at startup |
124
+ | OSM greenspace polygons | OpenStreetMap | Shapefile bundle | Downloaded from HuggingFace at startup |
125
+ | Greenspace distance rasters | Derived from OSM (see `making-greenspace-raster.R`) | GeoTIFF | Downloaded from HuggingFace at startup |
126
+ | RSF Program distance rasters | Derived from RSF polygons (see `making-rsfprogram-raster.R`) | GeoTIFF | Downloaded from HuggingFace at startup |
127
+ | Biodiversity hotspots/coldspots | KnowBR analysis on GBIF | Shapefile | Downloaded from HuggingFace at startup |
128
+ | SF Muni GTFS | SFMTA official GTFS feed | zip + rds + csv | Downloaded from HuggingFace at startup |
129
+ | CalEnviroScreen 4.0 | OEHHA | GeoPackage | Downloaded from HuggingFace at startup |
130
+ | SF EJ Communities | SF Environment | GeoPackage | Downloaded from HuggingFace at startup |
131
+ | RSF Program Projects | RSF Initiative | GeoPackage | Downloaded from HuggingFace at startup |
132
+
133
+ **Remote data** is hosted on HuggingFace at
134
+ [`boettiger-lab/sf_biodiv_access`](https://huggingface.co/datasets/boettiger-lab/sf_biodiv_access) and cached locally in `data/cached/` by `setup_unified.R`. The two most expensive products — the CBG × greenspace intersection and the transit-routing timetable — are precomputed in `Rscripts/prep/` and read at startup rather than recomputed, so the app starts quickly.
135
+
136
+ **GBIF queries** use [DuckDB](https://duckdb.org/) with the spatial extension, querying a local `.parquet` file directly via SQL `ST_Intersects` — avoiding loading the full dataset into memory.
137
+
138
+ ---
139
+
140
+ ## Local Development Setup
141
+
142
+ ### Prerequisites
143
+
144
+ Install required R packages:
145
+
146
+ ```r
147
+ install.packages(c(
148
+ "shiny", "shinydashboard", "leaflet", "mapboxapi", "tidyverse",
149
+ "tidycensus", "sf", "DT", "RColorBrewer", "terra", "data.table",
150
+ "mapview", "sjPlot", "sjlabelled", "bslib", "shinycssloaders",
151
+ "DBI", "duckdb", "dbplyr", "gtfsrouter", "tidytransit", "fmsb", "scales"
152
+ ))
153
+ ```
154
+
155
+ ### First-Time Setup
156
+
157
+ **Step 1: Build GBIF parquets** (one-time; needs local `gbif.duckdb` path in script)
158
+ ```r
159
+ source("Rscripts/prep/create_annotated_gbif_parquet.R")
160
+ # Creates: data/output/sf-gbif.parquet, data/output/gbif_census_ndvi_anno.parquet
161
+ ```
162
+
163
+ **Step 2: Pre-compute GTFS timetable and greenspace / RSF rasters** (one-time, slow)
164
+ ```r
165
+ source("Rscripts/prep/run_all_prep.R")
166
+ ```
167
+
168
+ **Step 3: Run the app**
169
+ ```r
170
+ shiny::runApp("app.R")
171
+ ```
172
+
173
+ ### Startup Performance
174
+
175
+ Typical startup time: **~6–12 seconds** depending on whether caches exist.
176
+
177
+ ---
178
+
179
+ ## Cloud Deployment (HuggingFace Spaces)
180
+
181
+ The app is deployed via Docker on HuggingFace Spaces using `app.R` + `Rscripts/setup_unified.R`. Data is downloaded from the HuggingFace dataset repository into `data/cached/` at startup — so no large files need to be bundled in the image.
182
+
183
+ See `Dockerfile` for the deployment configuration.
184
+
185
+ ---
186
+
187
+ ## Authors
188
+
189
+ **Diego Ellis Soto**, Avery Hill, Christopher J. Schell, Carl Boettiger, Rebecca Johnson
190
+ University of California Berkeley (ESPM) · California Academy of Sciences
191
+ Contact: [diego.ellissoto@berkeley.edu](mailto:diego.ellissoto@berkeley.edu)
192
+
193
+ ---
194
+
195
+ ## Status & Roadmap
196
+
197
+ This tool is a **decision-support prototype** co-developed with the RSF Data Working Group. The BAI should be treated as an exploratory indicator; variable weights and reference distributions are subject to revision through stakeholder co-development.
198
+
199
+ **Planned additions:**
200
+ - Impervious surface coverage
201
+ - National Walkability Index
202
+ - CDC Social Vulnerability Index
203
+ - NatureServe biodiversity / rarity maps
204
+ - Frequency-weighted multimodal transit accessibility
205
+ - Pre-cached transit isochrones for faster queries
206
+
207
+ ---
208
+
209
+ <img src="www/hexbin_RSF_logo.png" width="80">
Rscripts/iso_metrics.R ADDED
@@ -0,0 +1,709 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ============================================================================
2
+ # Shared isochrone + metric functions (sourced once by app.R at startup)
3
+ # ============================================================================
4
+ # These four functions are the single source of truth for turning a point into
5
+ # an isochrone, scoring it, and drawing its BAI spider plot. Both the main
6
+ # "Isochrone Explorer" tab and the "Isochrone Comparer" tab call them, so the
7
+ # scoring logic lives in exactly one place.
8
+ #
9
+ # build_isochrones() point + modes/times -> isochrone sf (mode,time,geom)
10
+ # compute_iso_metrics() isochrone sf + point -> per-isochrone metric data.frame
11
+ # add_bai() metric df + benchmarks -> df + 7 BAI axes + composite BAI
12
+ # draw_radar() BAI df -> spider/radar plot (base graphics)
13
+ #
14
+ # These functions read the *static* objects loaded by Rscripts/setup_unified.R
15
+ # directly as globals (cbg_vect_sf, osm_greenspace, the distance/NDVI rasters,
16
+ # rsf_projects, cbg_greenspace_coverage, gtfs_stops_sf, gtfs_routes_sf,
17
+ # gtfs_router, transit_iso_cache, cenv_sf, sf_ej_sf) plus the helpers/config
18
+ # defined at the top of app.R (mapbox_token, mode_palette, pretty_mode, ecdf01,
19
+ # standardize_iso_sf, build_walk_transit_isochrone, safe_biodiv_hotspots/coldspots).
20
+ #
21
+ # Two objects are created per-session inside server() and so are passed in as
22
+ # explicit arguments rather than read as globals:
23
+ # gbif_tab -- the session's DuckDB handle on the GBIF parquet
24
+ # city_benchmarks -- the citywide ECDF reference distributions for the BAI
25
+ #
26
+ # compute_iso_metrics() reports progress via withProgress(), so it must be
27
+ # called from within a Shiny reactive context (which both tabs satisfy).
28
+ # ============================================================================
29
+
30
+
31
+ # ----------------------------------------------------------------------------
32
+ # build_isochrones(): point + chosen modes/times -> combined isochrone sf
33
+ # ----------------------------------------------------------------------------
34
+ # point: named numeric c(lon = , lat = ) (the shape chosen_point() stores), or NULL.
35
+ # modes: character vector from c("driving","walking","cycling","driving-traffic",
36
+ # "transit","walk_transit"). times: numeric minutes. The transit_* / walk_*
37
+ # arguments only matter when a transit mode is selected; they default to the
38
+ # main tab's defaults so the comparer can pass a single mode/time and ignore them.
39
+ # Returns an sf with columns mode, time, geometry (EPSG:4326), or NULL if nothing
40
+ # could be built.
41
+ build_isochrones <- function(point, modes, times,
42
+ transit_hour = 9,
43
+ walk_to_stop_min = 5,
44
+ walk_from_stop_min = 5,
45
+ transit_departure_window_min = 10) {
46
+ if (is.null(point) || length(modes) == 0 || length(times) == 0) return(NULL)
47
+
48
+ location_sf <- st_as_sf(
49
+ data.frame(lon = point["lon"], lat = point["lat"]),
50
+ coords = c("lon", "lat"),
51
+ crs = 4326
52
+ )
53
+
54
+ iso_list <- list()
55
+ times <- as.numeric(times)
56
+
57
+ # --- Mapbox modes (driving / walking / cycling / driving-traffic) ----------
58
+ mapbox_modes <- intersect(modes, c("driving", "walking", "cycling", "driving-traffic"))
59
+ for (mode in mapbox_modes) {
60
+ for (t in times) {
61
+ iso <- tryCatch(
62
+ mb_isochrone(location_sf, time = t, profile = mode, access_token = mapbox_token),
63
+ # Surface the Mapbox error to the log instead of silently dropping it --
64
+ # otherwise a failed call just looks like "no isochrone" with no clue why.
65
+ error = function(e) { warning("mb_isochrone failed (", mode, " ", t, " min): ", conditionMessage(e)); NULL }
66
+ )
67
+ if (!is.null(iso)) {
68
+ iso_std <- standardize_iso_sf(iso, mode_name = mode, time_min = t)
69
+ if (!is.null(iso_std)) iso_list <- append(iso_list, list(iso_std))
70
+ }
71
+ }
72
+ }
73
+
74
+ # --- Transit (GTFS) --------------------------------------------------------
75
+ if ("transit" %in% modes && !is.null(gtfs_router) && !is.null(gtfs_stops_sf)) {
76
+ stop_dists <- st_distance(location_sf, gtfs_stops_sf)
77
+ nearest_idx <- which.min(stop_dists)
78
+ nearest_id <- as.character(gtfs_stops_sf$stop_id[nearest_idx])
79
+ dep_secs <- as.numeric(transit_hour) * 3600
80
+
81
+ for (t in times) {
82
+ iso_poly <- NULL
83
+
84
+ if (!is.null(transit_iso_cache) &&
85
+ !is.null(transit_iso_cache[[nearest_id]]) &&
86
+ !is.null(transit_iso_cache[[nearest_id]][[as.character(t)]])) {
87
+ iso_poly <- transit_iso_cache[[nearest_id]][[as.character(t)]]
88
+ }
89
+
90
+ if (is.null(iso_poly)) {
91
+ iso_result <- tryCatch(
92
+ gtfsrouter::gtfs_isochrone(
93
+ gtfs = gtfs_router,
94
+ from = nearest_id,
95
+ start_time = dep_secs,
96
+ end_time = dep_secs + t * 60,
97
+ from_is_id = TRUE
98
+ ),
99
+ error = function(e) NULL
100
+ )
101
+
102
+ if (!is.null(iso_result) && nrow(iso_result) > 2) {
103
+ reachable_sf <- gtfs_stops_sf |>
104
+ filter(stop_id %in% as.character(iso_result$stop_id))
105
+
106
+ if (nrow(reachable_sf) > 2) {
107
+ iso_poly <- st_convex_hull(st_union(reachable_sf))
108
+ } else if (nrow(reachable_sf) > 0) {
109
+ iso_poly <- st_union(st_buffer(st_transform(reachable_sf, 3857), 100)) |>
110
+ st_transform(4326)
111
+ }
112
+ }
113
+ }
114
+
115
+ if (!is.null(iso_poly)) {
116
+ iso_sf <- st_sf(
117
+ mode = "transit",
118
+ time = as.numeric(t),
119
+ geometry = st_geometry(st_as_sf(iso_poly)),
120
+ crs = 4326
121
+ )
122
+ iso_sf <- standardize_iso_sf(iso_sf, mode_name = "transit", time_min = t)
123
+ iso_list <- append(iso_list, list(iso_sf))
124
+ }
125
+ }
126
+ }
127
+
128
+ # --- Walk + Transit (Muni) -------------------------------------------------
129
+ if ("walk_transit" %in% modes && !is.null(gtfs_router) && !is.null(gtfs_stops_sf)) {
130
+ dep_secs <- as.numeric(transit_hour) * 3600
131
+ valid_times <- times[times > walk_to_stop_min]
132
+
133
+ for (t in valid_times) {
134
+ wt_iso <- tryCatch(
135
+ build_walk_transit_isochrone(
136
+ location_sf = location_sf,
137
+ total_time_min = t,
138
+ dep_secs = dep_secs,
139
+ walk_to_stop_min = walk_to_stop_min,
140
+ walk_from_stop_min = walk_from_stop_min,
141
+ gtfs_stops_sf = gtfs_stops_sf,
142
+ gtfs_router = gtfs_router,
143
+ mapbox_token = mapbox_token,
144
+ departure_window_min = transit_departure_window_min,
145
+ departure_step_min = 5,
146
+ max_last_mile_stops = 12,
147
+ include_first_mile_polygon = TRUE
148
+ ),
149
+ error = function(e) NULL
150
+ )
151
+
152
+ if (!is.null(wt_iso) && nrow(wt_iso) > 0) {
153
+ iso_list <- append(iso_list, list(wt_iso))
154
+ }
155
+ }
156
+ }
157
+
158
+ if (length(iso_list) == 0) return(NULL)
159
+
160
+ dplyr::bind_rows(iso_list) |>
161
+ st_as_sf() |>
162
+ st_make_valid() |>
163
+ st_transform(4326)
164
+ }
165
+
166
+
167
+ # ----------------------------------------------------------------------------
168
+ # compute_iso_metrics(): isochrone sf -> per-isochrone metric data.frame
169
+ # ----------------------------------------------------------------------------
170
+ # iso_data: the sf returned by build_isochrones(). point: the c(lon, lat) the
171
+ # isochrones were built from (used for the nearest-greenspace / nearest-RSF
172
+ # lookups), or NULL. gbif_tab: the session's DuckDB handle on the GBIF parquet.
173
+ # Returns a data.frame (one row per isochrone) carrying aggregate summaries as
174
+ # attributes: bio_percentile, city/mean transit scores, closest_greenspace(+dist),
175
+ # closest_rsf_program(+dist).
176
+ compute_iso_metrics <- function(iso_data, point, gbif_tab) {
177
+ if (is.null(iso_data) || nrow(iso_data) == 0) return(data.frame())
178
+
179
+ hotspot_union <- safe_biodiv_hotspots()
180
+ coldspot_union <- safe_biodiv_coldspots()
181
+
182
+ if (!is.null(hotspot_union)) hotspot_union <- st_union(hotspot_union)
183
+ if (!is.null(coldspot_union)) coldspot_union <- st_union(coldspot_union)
184
+
185
+ acs_wide <- cbg_vect_sf |>
186
+ mutate(population = popE, med_income = medincE)
187
+
188
+ results <- data.frame()
189
+ n_isos <- nrow(iso_data)
190
+
191
+ user_point_sf <- NULL
192
+ if (!is.null(point)) {
193
+ user_point_sf <- st_as_sf(
194
+ data.frame(lon = point["lon"], lat = point["lat"]),
195
+ coords = c("lon", "lat"),
196
+ crs = 4326
197
+ )
198
+ }
199
+
200
+ min_dist_val_global <- NA_real_
201
+ osm_greenspace_name_global <- NA_character_
202
+ if (!is.null(user_point_sf) && exists("greenspace_dist_raster") && exists("greenspace_osmid_raster")) {
203
+ try({
204
+ # Distance from the selected point to its nearest greenspace, read straight
205
+ # from the distance raster. terra::extract() returns data.frame(ID, value),
206
+ # so the distance is column 2 -- pulling column 1 (the point ID) was the bug
207
+ # that made this come back as "1" instead of the real distance in metres.
208
+ min_dist_val_global <- (greenspace_dist_raster |> extract(vect(user_point_sf)) |> pull(2))[1]
209
+ user_point_osm_id <- (greenspace_osmid_raster |> extract(vect(user_point_sf)) |> pull(2))[1]
210
+ osm_greenspace_name_global <- osm_greenspace |>
211
+ mutate(osm_id = as.numeric(osm_id)) |>
212
+ filter(osm_id == user_point_osm_id) |>
213
+ pull(name)
214
+ if (length(osm_greenspace_name_global) == 0 || is.na(osm_greenspace_name_global[1])) {
215
+ osm_greenspace_name_global <- "Unnamed Greenspace"
216
+ } else {
217
+ osm_greenspace_name_global <- osm_greenspace_name_global[1]
218
+ }
219
+ }, silent = TRUE)
220
+ }
221
+
222
+ min_rsf_dist_global <- NA_real_
223
+ rsf_program_name_global <- NA_character_
224
+ if (!is.null(user_point_sf) && exists("rsfprogram_dist_raster") && exists("rsfprogram_id_raster") && exists("rsf_projects")) {
225
+ try({
226
+ # Distance is column 2 (column 1 is the point ID) -- same fix as greenspace above.
227
+ min_rsf_dist_global <- (rsfprogram_dist_raster |> extract(vect(user_point_sf)) |> pull(2))[1]
228
+ user_point_rsf_pid <- (rsfprogram_id_raster |> extract(vect(user_point_sf)) |> pull(2))[1]
229
+ rsf_program_name_global <- rsf_projects |>
230
+ dplyr::filter(as.numeric(.data$polygon_id) == as.numeric(user_point_rsf_pid)) |>
231
+ dplyr::pull(.data$prj_name)
232
+ if (length(rsf_program_name_global) == 0 || is.na(rsf_program_name_global[1])) {
233
+ rsf_program_name_global <- "Unknown RSF program"
234
+ } else {
235
+ rsf_program_name_global <- rsf_program_name_global[1]
236
+ }
237
+ }, silent = TRUE)
238
+ }
239
+
240
+ withProgress(message = "Analyzing isochrones...", value = 0, {
241
+
242
+ for (i in seq_len(n_isos)) {
243
+ poly_i <- iso_data[i, ]
244
+ vect_poly_i <- vect(poly_i)
245
+
246
+ incProgress(
247
+ 1 / n_isos,
248
+ detail = paste0(
249
+ pretty_mode(as.character(poly_i$mode[[1]])),
250
+ " – ", poly_i$time[[1]], " min"
251
+ )
252
+ )
253
+
254
+ dist_hot_km <- if (!is.null(hotspot_union)) {
255
+ round(as.numeric(min(st_distance(poly_i, hotspot_union))) / 1000, 3)
256
+ } else NA_real_
257
+
258
+ dist_cold_km <- if (!is.null(coldspot_union)) {
259
+ round(as.numeric(min(st_distance(poly_i, coldspot_union))) / 1000, 3)
260
+ } else NA_real_
261
+
262
+ inter_acs <- tryCatch(intersect(vect(acs_wide), vect_poly_i) |> st_as_sf(), error = function(e) NULL)
263
+
264
+ pop_total <- 0
265
+ w_income <- NA_real_
266
+ if (!is.null(inter_acs) && nrow(inter_acs) > 0) {
267
+ inter_acs <- inter_acs |>
268
+ mutate(
269
+ area_num = as.numeric(st_area(st_transform(geometry, 3857))),
270
+ weighted_pop = population * (area_num / sum(area_num, na.rm = TRUE))
271
+ )
272
+ pop_total <- round(sum(inter_acs$weighted_pop, na.rm = TRUE))
273
+ w_income <- sum(inter_acs$med_income * inter_acs$area_num, na.rm = TRUE) /
274
+ sum(inter_acs$area_num, na.rm = TRUE)
275
+ }
276
+
277
+ iso_area_m2 <- as.numeric(st_area(st_transform(poly_i, 3857)))
278
+ iso_area_km2 <- round(iso_area_m2 / 1e6, 3)
279
+
280
+ # Greenspace area within this isochrone: use pre-cached per-CBG coverage
281
+ # (replaces slow per-isochrone raster cellSize() which took ~40-55 s each).
282
+ # Method: intersect isochrone with CBGs, scale each CBG's greenspace_m2 by
283
+ # the fraction of that CBG falling inside the isochrone, then sum.
284
+ gs_area_m2 <- tryCatch({
285
+ if (exists("cbg_greenspace_coverage") && !is.null(cbg_greenspace_coverage)) {
286
+ poly_proj <- st_transform(poly_i, 3857)
287
+ cbg_proj_i <- st_transform(cbg_vect_sf[, "GEOID"], 3857) |>
288
+ mutate(cbg_area_m2 = as.numeric(st_area(geometry))) |>
289
+ st_make_valid()
290
+ inter_df <- st_intersection(cbg_proj_i, poly_proj) |>
291
+ mutate(inter_area_m2 = as.numeric(st_area(geometry))) |>
292
+ st_drop_geometry() |>
293
+ left_join(
294
+ cbg_greenspace_coverage[, c("GEOID", "greenspace_m2", "cbg_area_m2")],
295
+ by = "GEOID", suffix = c("_iso", "_cbg")
296
+ ) |>
297
+ mutate(
298
+ greenspace_m2 = tidyr::replace_na(greenspace_m2, 0),
299
+ cbg_area_m2 = dplyr::coalesce(cbg_area_m2_cbg, cbg_area_m2_iso),
300
+ contrib = ifelse(cbg_area_m2 > 0,
301
+ inter_area_m2 / cbg_area_m2 * greenspace_m2,
302
+ 0)
303
+ )
304
+ sum(inter_df$contrib, na.rm = TRUE)
305
+ } else if (exists("greenspace_dist_raster")) {
306
+ # Fallback to raster method if cache is unavailable
307
+ dist_crop <- terra::crop(greenspace_dist_raster, vect_poly_i)
308
+ dist_mask <- terra::mask(dist_crop, vect_poly_i)
309
+ is_greenspace <- dist_mask == 0
310
+ cell_areas <- terra::cellSize(is_greenspace, unit = "m")
311
+ as.numeric(terra::global(cell_areas * is_greenspace, "sum", na.rm = TRUE)[1, 1])
312
+ } else {
313
+ 0
314
+ }
315
+ }, error = function(e) 0)
316
+ gs_percent <- ifelse(iso_area_m2 > 0, 100 * gs_area_m2 / iso_area_m2, 0)
317
+
318
+ mean_ndvi <- NA_real_
319
+ if (exists("ndvi")) {
320
+ ndvi_vals <- tryCatch(values(terra::mask(terra::crop(ndvi, vect_poly_i), vect_poly_i)), error = function(e) NA)
321
+ ndvi_vals <- ndvi_vals[!is.na(ndvi_vals)]
322
+ mean_ndvi <- ifelse(length(ndvi_vals) > 0, round(mean(ndvi_vals, na.rm = TRUE), 3), NA_real_)
323
+ }
324
+
325
+ iso_wkt <- st_as_text(st_geometry(poly_i)[[1]])
326
+ gbif_summary <- tryCatch({
327
+ gbif_tab |>
328
+ filter(sql(glue("ST_Intersects(ST_GeomFromText(geom_wkt), ST_GeomFromText('{iso_wkt}'))"))) |>
329
+ summarise(
330
+ n_records = n(),
331
+ n_species = n_distinct(species),
332
+ n_birds = n_distinct(case_when(class == "Aves" ~ species, TRUE ~ NA_character_)),
333
+ n_mammals = n_distinct(case_when(class == "Mammalia" ~ species, TRUE ~ NA_character_)),
334
+ n_plants = n_distinct(case_when(
335
+ class %in% c("Magnoliopsida", "Liliopsida", "Pinopsida", "Polypodiopsida",
336
+ "Equisetopsida", "Bryopsida", "Marchantiopsida") ~ species,
337
+ TRUE ~ NA_character_
338
+ ))
339
+ ) |>
340
+ collect()
341
+ }, error = function(e) NULL)
342
+ n_records <- if (!is.null(gbif_summary) && nrow(gbif_summary) > 0) gbif_summary$n_records[[1]] else 0L
343
+ n_species <- if (!is.null(gbif_summary) && nrow(gbif_summary) > 0) gbif_summary$n_species[[1]] else 0L
344
+ n_birds <- if (!is.null(gbif_summary) && nrow(gbif_summary) > 0) gbif_summary$n_birds[[1]] else 0L
345
+ n_mammals <- if (!is.null(gbif_summary) && nrow(gbif_summary) > 0) gbif_summary$n_mammals[[1]] else 0L
346
+ n_plants <- if (!is.null(gbif_summary) && nrow(gbif_summary) > 0) gbif_summary$n_plants[[1]] else 0L
347
+
348
+ n_transit_stops <- NA_real_
349
+ transit_access_score <- NA_real_
350
+ freq_weighted_score <- NA_real_
351
+ mean_headway_iso <- NA_real_
352
+ nearest_stop_m <- NA_real_
353
+ nearest_stop_name <- NA_character_
354
+
355
+ if (!is.null(gtfs_stops_sf)) {
356
+ inter_transit <- tryCatch(st_intersection(gtfs_stops_sf, poly_i), error = function(e) NULL)
357
+ n_transit_stops <- if (!is.null(inter_transit)) nrow(inter_transit) else 0
358
+
359
+ dist_transit <- st_distance(poly_i, gtfs_stops_sf)
360
+ nearest_stop_m <- round(as.numeric(min(dist_transit)), 0)
361
+ nearest_stop_name <- gtfs_stops_sf$stop_name[which.min(dist_transit)]
362
+
363
+ transit_access_score <- ifelse(iso_area_km2 > 0, round(n_transit_stops / iso_area_km2, 2), NA_real_)
364
+
365
+ freq_weighted_score <- if (!is.null(inter_transit) &&
366
+ "mean_headway_min" %in% names(inter_transit) &&
367
+ nrow(inter_transit) > 0 &&
368
+ iso_area_km2 > 0) {
369
+ hw <- inter_transit$mean_headway_min
370
+ hw <- hw[!is.na(hw) & hw > 0]
371
+ if (length(hw) > 0) round(sum(60 / hw) / iso_area_km2, 2) else NA_real_
372
+ } else NA_real_
373
+
374
+ mean_headway_iso <- if (!is.null(inter_transit) &&
375
+ "mean_headway_min" %in% names(inter_transit) &&
376
+ nrow(inter_transit) > 0) {
377
+ round(mean(inter_transit$mean_headway_min, na.rm = TRUE), 1)
378
+ } else NA_real_
379
+ }
380
+
381
+ # Unique Muni route IDs whose shapes intersect this isochrone
382
+ n_unique_routes <- 0L
383
+ if (!is.null(gtfs_routes_sf)) {
384
+ routes_inter <- tryCatch(
385
+ st_intersection(gtfs_routes_sf[, "route_id"], poly_i),
386
+ error = function(e) NULL
387
+ )
388
+ n_unique_routes <- if (!is.null(routes_inter) && "route_id" %in% names(routes_inter)) {
389
+ length(unique(routes_inter$route_id))
390
+ } else 0L
391
+ }
392
+
393
+ sampling_density_km2 <- ifelse(iso_area_km2 > 0, round(n_records / iso_area_km2, 2), NA_real_)
394
+
395
+ mean_ciscore <- if (!is.null(cenv_sf)) {
396
+ tryCatch({
397
+ ce_inter <- st_intersection(cenv_sf, poly_i)
398
+ if (nrow(ce_inter) > 0) {
399
+ ce_inter$a <- as.numeric(st_area(st_transform(ce_inter, 3857)))
400
+ round(weighted.mean(ce_inter$CIscore, w = ce_inter$a, na.rm = TRUE), 1)
401
+ } else NA_real_
402
+ }, error = function(e) NA_real_)
403
+ } else NA_real_
404
+
405
+ mean_traffic_pctl <- if (!is.null(cenv_sf)) {
406
+ tryCatch({
407
+ ce_inter <- st_intersection(cenv_sf, poly_i)
408
+ if (nrow(ce_inter) > 0) {
409
+ ce_inter$a <- as.numeric(st_area(st_transform(ce_inter, 3857)))
410
+ round(weighted.mean(ce_inter$Traffic_Pctl, w = ce_inter$a, na.rm = TRUE), 1)
411
+ } else NA_real_
412
+ }, error = function(e) NA_real_)
413
+ } else NA_real_
414
+
415
+ mean_ej_score <- if (!is.null(sf_ej_sf)) {
416
+ tryCatch({
417
+ ej_inter <- st_intersection(sf_ej_sf, poly_i)
418
+ if (nrow(ej_inter) > 0) {
419
+ ej_inter$a <- as.numeric(st_area(st_transform(ej_inter, 3857)))
420
+ valid <- ej_inter[!is.na(ej_inter$score), ]
421
+ if (nrow(valid) > 0) {
422
+ round(weighted.mean(valid$score, w = valid$a, na.rm = TRUE), 1)
423
+ } else NA_real_
424
+ } else NA_real_
425
+ }, error = function(e) NA_real_)
426
+ } else NA_real_
427
+
428
+ row_i <- data.frame(
429
+ Mode = pretty_mode(as.character(poly_i$mode[[1]])),
430
+ Time = as.numeric(poly_i$time[[1]]),
431
+ IsochroneArea_km2 = iso_area_km2,
432
+ DistToHotspot_km = dist_hot_km,
433
+ DistToColdspot_km = dist_cold_km,
434
+ EstimatedPopulation = pop_total,
435
+ MedianIncome = round(w_income, 2),
436
+ MeanNDVI = mean_ndvi,
437
+ GBIF_Records = n_records,
438
+ GBIF_Species = n_species,
439
+ Bird_Species = n_birds,
440
+ Mammal_Species = n_mammals,
441
+ Plant_Species = n_plants,
442
+ SamplingDensity_km2 = sampling_density_km2,
443
+ Greenspace_percent = round(gs_percent, 2),
444
+ Transit_Stops = n_transit_stops,
445
+ Unique_Muni_Routes = n_unique_routes,
446
+ Transit_Access_Score = transit_access_score,
447
+ Freq_Weighted_Score = freq_weighted_score,
448
+ Mean_Headway_min = mean_headway_iso,
449
+ Nearest_Stop_m = nearest_stop_m,
450
+ Nearest_Stop_Name = nearest_stop_name,
451
+ CalEnviro_CIscore = mean_ciscore,
452
+ CalEnviro_Traffic_Pctl = mean_traffic_pctl,
453
+ SF_EJ_Score = mean_ej_score,
454
+ closest_greenspace = osm_greenspace_name_global,
455
+ closest_greenspace_dist_m = min_dist_val_global,
456
+ closest_rsf_program = rsf_program_name_global,
457
+ closest_rsf_program_dist_m = min_rsf_dist_global,
458
+ stringsAsFactors = FALSE
459
+ )
460
+
461
+ results <- rbind(results, row_i)
462
+ }
463
+
464
+ }) # end withProgress
465
+
466
+ union_wkt <- st_as_text(st_geometry(st_union(iso_data))[[1]])
467
+ union_n_species <- tryCatch({
468
+ gbif_tab |>
469
+ filter(sql(glue("ST_Intersects(ST_GeomFromText(geom_wkt), ST_GeomFromText('{union_wkt}'))"))) |>
470
+ summarise(n_species = n_distinct(species)) |>
471
+ collect() |>
472
+ pull(n_species)
473
+ }, error = function(e) 0L)
474
+
475
+ attr(results, "bio_percentile") <- round(100 * ecdf(cbg_vect_sf$unique_species)(union_n_species), 1)
476
+
477
+ if (!is.null(gtfs_stops_sf)) {
478
+ sf_city_area_km2 <- 121.4
479
+ attr(results, "city_transit_score") <- round(nrow(gtfs_stops_sf) / sf_city_area_km2, 2)
480
+ attr(results, "mean_transit_score") <- round(mean(results$Transit_Access_Score, na.rm = TRUE), 2)
481
+ attr(results, "mean_transit_stops") <- round(mean(results$Transit_Stops, na.rm = TRUE), 1)
482
+ attr(results, "mean_muni_routes") <- round(mean(results$Unique_Muni_Routes, na.rm = TRUE), 1)
483
+ } else {
484
+ attr(results, "city_transit_score") <- NA_real_
485
+ attr(results, "mean_transit_score") <- NA_real_
486
+ attr(results, "mean_transit_stops") <- NA_real_
487
+ attr(results, "mean_muni_routes") <- NA_real_
488
+ }
489
+
490
+ if (nrow(results) > 0) {
491
+ closest_gs <- results |>
492
+ filter(!is.na(closest_greenspace_dist_m)) |>
493
+ slice_min(closest_greenspace_dist_m, n = 1)
494
+ if (nrow(closest_gs) > 0) {
495
+ attr(results, "closest_greenspace") <- closest_gs$closest_greenspace[1]
496
+ attr(results, "closest_greenspace_dist_m") <- closest_gs$closest_greenspace_dist_m[1]
497
+ } else {
498
+ attr(results, "closest_greenspace") <- "None"
499
+ attr(results, "closest_greenspace_dist_m") <- NA_real_
500
+ }
501
+
502
+ closest_rsf <- results |>
503
+ filter(!is.na(closest_rsf_program_dist_m)) |>
504
+ slice_min(closest_rsf_program_dist_m, n = 1)
505
+ if (nrow(closest_rsf) > 0) {
506
+ attr(results, "closest_rsf_program") <- closest_rsf$closest_rsf_program[1]
507
+ attr(results, "closest_rsf_program_dist_m") <- closest_rsf$closest_rsf_program_dist_m[1]
508
+ } else {
509
+ attr(results, "closest_rsf_program") <- "None"
510
+ attr(results, "closest_rsf_program_dist_m") <- NA_real_
511
+ }
512
+ }
513
+
514
+ results
515
+ }
516
+
517
+
518
+ # ----------------------------------------------------------------------------
519
+ # add_bai(): metric df + citywide benchmarks -> df + 7 BAI axes + composite BAI
520
+ # ----------------------------------------------------------------------------
521
+ # df: a data.frame from compute_iso_metrics(). city_benchmarks: the list of
522
+ # citywide reference distributions (the ECDFs each axis is scored against).
523
+ # Returns df with the seven *_std columns (each 0-1) plus BAI (their row mean),
524
+ # or NULL if df is empty.
525
+ add_bai <- function(df, city_benchmarks) {
526
+ if (is.null(df) || nrow(df) == 0) return(NULL)
527
+
528
+ ref <- city_benchmarks
529
+
530
+ eq_ref_inverted <- if (length(ref$ej) > 0) max(ref$ej, na.rm = TRUE) - ref$ej else numeric(0)
531
+ eq_obs_inverted <- ifelse(is.na(df$SF_EJ_Score), NA_real_, max(ref$ej, na.rm = TRUE) - df$SF_EJ_Score)
532
+
533
+ tmp <- df |>
534
+ mutate(
535
+ Mobility_Access_std = ecdf01(Transit_Access_Score, ref$transit_density),
536
+ Biodiversity_Potential_std = ecdf01(GBIF_Species, ref$biodiversity),
537
+ Observation_Intensity_std = ecdf01(SamplingDensity_km2, ref$sampling),
538
+ Environmental_Quality_std = ecdf01(MeanNDVI, ref$ndvi),
539
+ Greenspace_Cover_std = ecdf01(Greenspace_percent, ref$greenspace_cover),
540
+ Equity_Context_std = ecdf01(eq_obs_inverted, eq_ref_inverted),
541
+ Route_Access_std = ecdf01(Unique_Muni_Routes, ref$route_access)
542
+ )
543
+
544
+ tmp$BAI <- rowMeans(
545
+ tmp[, c(
546
+ "Mobility_Access_std",
547
+ "Biodiversity_Potential_std",
548
+ "Observation_Intensity_std",
549
+ "Environmental_Quality_std",
550
+ "Greenspace_Cover_std",
551
+ "Equity_Context_std",
552
+ "Route_Access_std"
553
+ )],
554
+ na.rm = TRUE
555
+ )
556
+
557
+ tmp
558
+ }
559
+
560
+
561
+ # Axis order shared by both radar functions, and the column each axis reads from.
562
+ # Keep these two vectors aligned -- the group labels below assume this clockwise
563
+ # order (1-2 Urban Access, 3-4 Biodiversity, 5-6 Environment, 7 EJ).
564
+ RADAR_AXIS_COLS <- c("Mobility_Access_std", "Route_Access_std", "Biodiversity_Potential_std",
565
+ "Observation_Intensity_std", "Environmental_Quality_std",
566
+ "Greenspace_Cover_std", "Equity_Context_std")
567
+ RADAR_AXIS_LABELS <- c("Stop\nDensity", "Route\nDiversity", "Species\nRichness",
568
+ "Obs.\nIntensity", "Vegetation\n(NDVI)", "Greenspace\nCover", "EJ\nContext")
569
+
570
+ # Radial group subheaders placed just beyond the axis labels. Shared by both
571
+ # radar functions so the grouping (Urban Access / Biodiversity / Environment /
572
+ # EJ) is drawn identically. Call after fmsb::radarchart() has drawn the chart.
573
+ draw_radar_group_labels <- function() {
574
+ n_ax <- 7
575
+ angs <- pi/2 - (0:(n_ax - 1)) * (2 * pi / n_ax)
576
+ cat_r <- 1.48 # radius just outside axis labels (~1.2-1.3)
577
+
578
+ text(cat_r * cos(mean(angs[1:2])), cat_r * sin(mean(angs[1:2])),
579
+ "Urban Access", cex = 0.72, col = "#2166ac", font = 2, xpd = TRUE)
580
+ text(cat_r * cos(mean(angs[3:4])), cat_r * sin(mean(angs[3:4])),
581
+ "Biodiversity", cex = 0.72, col = "#1b7837", font = 2, xpd = TRUE)
582
+ text(cat_r * cos(mean(angs[5:6])), cat_r * sin(mean(angs[5:6])),
583
+ "Environment", cex = 0.72, col = "#762a83", font = 2, xpd = TRUE)
584
+ text(cat_r * cos(angs[7]), cat_r * sin(angs[7]),
585
+ "Environmental\nJustice", cex = 0.72, col = "#b2182b", font = 2, xpd = TRUE)
586
+ }
587
+
588
+ # ----------------------------------------------------------------------------
589
+ # draw_radar(): BAI df -> spider/radar plot (base graphics, via fmsb)
590
+ # ----------------------------------------------------------------------------
591
+ # bai_df: the data.frame returned by add_bai() (one row per isochrone). title:
592
+ # the chart title. Draws directly to the current graphics device, so call it
593
+ # from inside a renderPlot({ }). Lines are coloured by transport mode.
594
+ draw_radar <- function(bai_df, title = "Biodiversity Access Index Profile") {
595
+ if (is.null(bai_df) || nrow(bai_df) == 0) return(NULL)
596
+
597
+ radar_df <- bai_df |>
598
+ mutate(
599
+ ModeTime = paste0(Mode, "_", Time, "m"),
600
+ `Stop\nDensity` = Mobility_Access_std,
601
+ `Route\nDiversity` = Route_Access_std,
602
+ `Species\nRichness` = Biodiversity_Potential_std,
603
+ `Obs.\nIntensity` = Observation_Intensity_std,
604
+ `Vegetation\n(NDVI)`= Environmental_Quality_std,
605
+ `Greenspace\nCover` = Greenspace_Cover_std,
606
+ `EJ\nContext` = Equity_Context_std
607
+ ) |>
608
+ select(
609
+ ModeTime,
610
+ `Stop\nDensity`,
611
+ `Route\nDiversity`,
612
+ `Species\nRichness`,
613
+ `Obs.\nIntensity`,
614
+ `Vegetation\n(NDVI)`,
615
+ `Greenspace\nCover`,
616
+ `EJ\nContext`
617
+ )
618
+
619
+ radar_mat <- as.data.frame(radar_df[, -1])
620
+ rownames(radar_mat) <- radar_df$ModeTime
621
+
622
+ radar_mat <- rbind(
623
+ rep(1, ncol(radar_mat)),
624
+ rep(0, ncol(radar_mat)),
625
+ radar_mat
626
+ )
627
+
628
+ labels <- rownames(radar_mat)[-(1:2)]
629
+ line_cols <- mode_palette[gsub("_(.*)$", "", labels)]
630
+ line_cols[is.na(line_cols)] <- "#666666"
631
+
632
+ # Extra margin so two-line axis labels aren't clipped
633
+ par(mar = c(2, 2, 3, 2))
634
+
635
+ fmsb::radarchart(
636
+ radar_mat,
637
+ axistype = 1,
638
+ pcol = line_cols,
639
+ plwd = 2,
640
+ plty = 1,
641
+ cglcol = "grey80",
642
+ cglty = 1,
643
+ cglwd = 0.8,
644
+ axislabcol = "grey40",
645
+ vlcex = 0.88,
646
+ title = title
647
+ )
648
+
649
+ draw_radar_group_labels()
650
+
651
+ legend(
652
+ "topright",
653
+ legend = labels,
654
+ col = line_cols,
655
+ lty = 1,
656
+ lwd = 2,
657
+ cex = 0.75,
658
+ bty = "n"
659
+ )
660
+ }
661
+
662
+
663
+ # ----------------------------------------------------------------------------
664
+ # draw_compare_radar(): overlay several BAI rows on ONE radar (Comparer tab)
665
+ # ----------------------------------------------------------------------------
666
+ # bai_rows: a list of single-row add_bai() data.frames (one per point).
667
+ # labels: legend text, one per row (e.g. "Point A: Walking 5 min").
668
+ # colors: one line colour per row -- this is where the A-vs-B colour scheme
669
+ # lives, so keep it in sync with the point markers / isochrones.
670
+ # Unlike draw_radar() (coloured by mode), this colours by *point* so two
671
+ # isochrones of the same mode are still distinguishable. Draws to the current
672
+ # device; call from inside renderPlot({ }).
673
+ draw_compare_radar <- function(bai_rows, labels, colors,
674
+ title = "Biodiversity Access Index — Point A vs Point B") {
675
+ if (length(bai_rows) == 0) return(NULL)
676
+
677
+ mat <- as.data.frame(do.call(rbind, lapply(bai_rows, function(d) as.numeric(d[1, RADAR_AXIS_COLS]))))
678
+ names(mat) <- RADAR_AXIS_LABELS
679
+ # fmsb wants the axis max (1) and min (0) as the first two rows.
680
+ mat <- rbind(rep(1, ncol(mat)), rep(0, ncol(mat)), mat)
681
+
682
+ par(mar = c(2, 2, 3, 2))
683
+ fmsb::radarchart(
684
+ mat,
685
+ axistype = 1,
686
+ pcol = colors,
687
+ pfcol = grDevices::adjustcolor(colors, alpha.f = 0.2), # translucent fills so overlaps read
688
+ plwd = 3,
689
+ plty = 1,
690
+ cglcol = "grey80",
691
+ cglty = 1,
692
+ cglwd = 0.8,
693
+ axislabcol = "grey40",
694
+ vlcex = 0.9,
695
+ title = title
696
+ )
697
+
698
+ draw_radar_group_labels()
699
+
700
+ legend(
701
+ "topright",
702
+ legend = labels,
703
+ col = colors,
704
+ lty = 1,
705
+ lwd = 3,
706
+ cex = 0.85,
707
+ bty = "n"
708
+ )
709
+ }
Rscripts/prep/build_cbg_greenspace_coverage.R ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ============================================================================
2
+ # Prep: Per-CBG greenspace coverage -> data/output
3
+ # ============================================================================
4
+ # Inputs: cbg_vect_sf + osm_greenspace (loaded via Rscripts/setup_unified.R,
5
+ # which pulls them from data/cached/ or HuggingFace)
6
+ # Output: data/output/cbg_greenspace_coverage.csv (GEOID, greenspace_m2, cbg_area_m2)
7
+ #
8
+ # The CBG x greenspace st_union + st_intersection is the most expensive single
9
+ # step at app startup. The inputs are static, so we precompute it here once and
10
+ # let setup_unified.R just read this small CSV instead of recomputing every run.
11
+ # Re-run when cbg_vect_sf or the greenspace polygons change, then upload
12
+ # data/output/cbg_greenspace_coverage.csv to HuggingFace (see upload_to_huggingface.R).
13
+ #
14
+ # Working directory must be the project root (folder containing data/ and Rscripts/).
15
+ # ============================================================================
16
+
17
+ # Reuse the app's canonical loader so coverage is computed from the exact same
18
+ # cbg_vect_sf + osm_greenspace the app uses (also gives us sf/dplyr, etc.).
19
+ source("Rscripts/setup_unified.R")
20
+
21
+ out_dir <- "data/output"
22
+ dir.create(out_dir, recursive = TRUE, showWarnings = FALSE)
23
+
24
+ cbg_proj <- st_transform(cbg_vect_sf[, "GEOID"], 3857) |>
25
+ mutate(cbg_area_m2 = as.numeric(st_area(geometry)))
26
+ gs_proj <- st_transform(osm_greenspace, 3857) |> st_make_valid()
27
+ gs_union <- st_union(gs_proj)
28
+ cbg_gs_inter <- st_intersection(cbg_proj, gs_union)
29
+
30
+ cbg_greenspace_coverage <- cbg_gs_inter |>
31
+ mutate(greenspace_m2 = as.numeric(st_area(geometry))) |>
32
+ st_drop_geometry() |>
33
+ group_by(GEOID) |>
34
+ summarise(greenspace_m2 = sum(greenspace_m2), .groups = "drop") |>
35
+ right_join(cbg_proj |> st_drop_geometry() |> dplyr::select(GEOID, cbg_area_m2), by = "GEOID") |>
36
+ mutate(
37
+ greenspace_m2 = tidyr::replace_na(greenspace_m2, 0),
38
+ GEOID = as.character(GEOID)
39
+ )
40
+
41
+ readr::write_csv(cbg_greenspace_coverage, file.path(out_dir, "cbg_greenspace_coverage.csv"))
42
+ message(glue::glue(
43
+ "[prep] wrote {nrow(cbg_greenspace_coverage)} CBG rows -> data/output/cbg_greenspace_coverage.csv"
44
+ ))
Rscripts/prep/build_equity_layers.R ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ============================================================================
2
+ # Prep: CalEnviroScreen + SF EJ → GeoPackages in data/output
3
+ # ============================================================================
4
+ # Reads local inputs from data/source/, writes derived layers for HuggingFace upload.
5
+
6
+ library(sf)
7
+ library(dplyr)
8
+
9
+ out_dir <- "data/output"
10
+ dir.create(out_dir, recursive = TRUE, showWarnings = FALSE)
11
+
12
+ # --- CalEnviroScreen (SF tracts only) -----------------------------------------
13
+
14
+ sf::st_read("data/source/calenviroscreen40gdb_F_2021.gdb", quiet = TRUE) |>
15
+ filter(grepl("san francisco", County, ignore.case = TRUE), !is.na(CIscore)) |>
16
+ select(
17
+ Tract, CIscore, CIscoreP,
18
+ PM2_5, PM2_5_Pctl, Traffic, Traffic_Pctl,
19
+ Poverty, Poverty_Pctl, HousBurd, HousBurd_Pctl,
20
+ County
21
+ ) |>
22
+ sf::st_transform(4326) |>
23
+ sf::st_make_valid() |>
24
+ sf::st_write(file.path(out_dir, "calenviro_sf.gpkg"), delete_dsn = TRUE, quiet = TRUE)
25
+
26
+ # --- SF Environmental Justice Communities -------------------------------------
27
+
28
+ sf::st_read(
29
+ "data/source/San Francisco Environmental Justice Communities Map_20260401/geo_export_e303e420-19f7-4166-9736-c1dbeda5b82e.shp",
30
+ quiet = TRUE
31
+ ) |>
32
+ sf::st_transform(4326) |>
33
+ sf::st_make_valid() |>
34
+ sf::st_write(file.path(out_dir, "sf_ej_communities_map.gpkg"), delete_dsn = TRUE, quiet = TRUE)
Rscripts/prep/create_annotated_gbif_parquet.R ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ============================================================================
2
+ # Prep: GBIF (DuckDB) → sf-gbif.parquet → CBG + NDVI → gbif_census_ndvi_anno.parquet
3
+ # ============================================================================
4
+ # 1) COPY San Francisco iNaturalist occurrences from attached gbif.duckdb → parquet
5
+ # 2) Join census block groups, extract NDVI, write annotated parquet for the app
6
+ #
7
+ # Requires: local gbif.duckdb path below; data/source/cbg_vect_sf.Rdata,
8
+ # data/source/SF_EastBay_NDVI_Sentinel_10.tif
9
+
10
+ # library(tidyverse)
11
+ library(dplyr)
12
+ library(tidyr)
13
+ library(readr)
14
+ library(purrr)
15
+ library(tibble)
16
+ library(stringr)
17
+ library(forcats)
18
+ library(lubridate)
19
+ library(ggplot2)
20
+ library(duckdb)
21
+ library(sf)
22
+ library(terra)
23
+ library(arrow)
24
+
25
+ # ============================================================================
26
+ # 1) Export raw GBIF points
27
+ # ============================================================================
28
+
29
+ scon <- dbConnect(duckdb())
30
+ scon |> dbExecute("INSTALL spatial; LOAD spatial;")
31
+ scon |> dbExecute("ATTACH '~/Data/Occurrences/GBIF/gbif.duckdb' AS gbifmain;")
32
+
33
+ scon |> dbExecute(
34
+ "COPY
35
+ (SELECT DISTINCT year,kingdom,phylum,class,\"order\",family,genus,species,decimallatitude,decimallongitude,coordinateuncertaintyinmeters,institutioncode FROM gbifmain.gbif
36
+ NATURAL JOIN gbifmain.join_lookup
37
+ WHERE stateprovince='California'
38
+ AND county_name='San Francisco'
39
+ AND institutioncode='iNaturalist'
40
+ AND decimallatitude IS NOT NULL
41
+ AND decimallongitude IS NOT NULL)
42
+ TO
43
+ 'data/output/sf-gbif.parquet'
44
+ (FORMAT parquet);
45
+ "
46
+ )
47
+
48
+ scon |> dbDisconnect(shutdown = TRUE)
49
+
50
+ # ============================================================================
51
+ # 2) Annotate: CBG + NDVI → gbif_census_ndvi_anno.parquet
52
+ # ============================================================================
53
+
54
+ gbif_in <- "data/output/sf-gbif.parquet"
55
+ cbg_rdata <- "data/source/cbg_vect_sf.Rdata"
56
+ ndvi_tif <- "data/source/SF_EastBay_NDVI_Sentinel_10.tif"
57
+ out_dir <- "data/output"
58
+ out_file <- file.path(out_dir, "gbif_census_ndvi_anno.parquet")
59
+
60
+ gbif <- arrow::read_parquet(gbif_in, as_data_frame = TRUE)
61
+ load(cbg_rdata)
62
+ ndvi <- terra::rast(ndvi_tif)
63
+
64
+ cbg_acs <- cbg_vect_sf |>
65
+ select(GEOID, medincE, popE, housingE)
66
+
67
+ pts <- gbif |>
68
+ filter(!is.na(decimallongitude), !is.na(decimallatitude)) |>
69
+ st_as_sf(coords = c("decimallongitude", "decimallatitude"), crs = 4326, remove = FALSE)
70
+
71
+ cbg_join <- st_join(pts, cbg_acs, join = st_within, left = TRUE)
72
+
73
+ cbg_join$ndvi_sentinel <- terra::extract(ndvi, cbg_join, ID = FALSE)$NDVI
74
+
75
+ out <- cbg_join |>
76
+ mutate(
77
+ geom_wkt = st_as_text(geometry),
78
+ longitude = decimallongitude,
79
+ latitude = decimallatitude
80
+ ) |>
81
+ st_drop_geometry() |>
82
+ rename(
83
+ institutionCode = institutioncode,
84
+ coordinateUncertaintyInMeters = coordinateuncertaintyinmeters
85
+ )
86
+
87
+ dir.create(out_dir, recursive = TRUE, showWarnings = FALSE)
88
+ arrow::write_parquet(out, out_file)
Rscripts/prep/implement_optimizations.R ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ============================================================================
2
+ # Prep: GTFS timetable, stop headways, and feed zip → data/output
3
+ # ============================================================================
4
+ # Inputs: data/source/muni_gtfs-current/*.txt
5
+ # Outputs: data/output/gtfs_timetable_monday.rds
6
+ # data/output/gtfs_stop_headways.csv
7
+ # data/output/sf_muni_gtfs.zip
8
+ #
9
+ # Run after updating the GTFS extract. Then upload data/output/* to HuggingFace.
10
+
11
+ # library(tidyverse)
12
+ library(dplyr)
13
+ library(tidyr)
14
+ library(readr)
15
+ library(purrr)
16
+ library(tibble)
17
+ library(stringr)
18
+ library(forcats)
19
+ library(lubridate)
20
+ library(ggplot2)
21
+ library(gtfsrouter)
22
+ library(tidytransit)
23
+ library(glue)
24
+
25
+ out_dir <- "data/output"
26
+ dir.create(out_dir, recursive = TRUE, showWarnings = FALSE)
27
+ out_dir_abs <- normalizePath(out_dir, mustWork = TRUE)
28
+
29
+ gtfs_dir <- "data/source/muni_gtfs-current"
30
+ if (!dir.exists(gtfs_dir)) {
31
+ stop(glue("GTFS folder not found: {gtfs_dir}"))
32
+ }
33
+ gtfs_txt <- list.files(gtfs_dir, pattern = "\\.txt$", full.names = FALSE)
34
+ if (length(gtfs_txt) == 0L) {
35
+ stop(glue("No .txt files under {gtfs_dir}"))
36
+ }
37
+
38
+ # --- Zip feed (same bytes consumers will download from HuggingFace) ----------
39
+ # Absolute zip path before setwd(); normalizePath(zip) is NA if the file does not exist yet
40
+ zip_abs <- file.path(out_dir_abs, "sf_muni_gtfs.zip")
41
+ old_wd <- getwd()
42
+ setwd(gtfs_dir)
43
+ utils::zip(zip_abs, files = gtfs_txt)
44
+ setwd(old_wd)
45
+
46
+ # --- gtfsrouter Monday timetable ------------------------------------------------
47
+ gr <- gtfsrouter::extract_gtfs(zip_abs, quiet = TRUE)
48
+ tt <- gtfsrouter::gtfs_timetable(gr, day = "Monday")
49
+ saveRDS(tt, file.path(out_dir_abs, "gtfs_timetable_monday.rds"), compress = "gzip")
50
+
51
+ # --- AM peak headways (inspectable CSV) ---------------------------------------
52
+ gt <- tidytransit::read_gtfs(zip_abs)
53
+ hw <- tidytransit::get_stop_frequency(gt, start_time = 7 * 3600, end_time = 9 * 3600) |>
54
+ group_by(stop_id) |>
55
+ summarise(
56
+ mean_headway_min = mean(mean_headway, na.rm = TRUE) / 60,
57
+ n_departures_peak = sum(n_departures, na.rm = TRUE),
58
+ .groups = "drop"
59
+ ) |>
60
+ mutate(stop_id = as.character(stop_id))
61
+
62
+ readr::write_csv(hw, file.path(out_dir_abs, "gtfs_stop_headways.csv"))
Rscripts/prep/making-greenspace-raster.R ADDED
@@ -0,0 +1,186 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Here we make 2 raster layers related to the nearest greenspaces in San Francisco
2
+ # at 30m resolution
3
+ # - 1) The distance to the nearest greenspace
4
+ # - 2) The osm_id of the nearest greenspace (for looking up the polygon)
5
+
6
+ # Load libraries
7
+ # library(tidyverse)
8
+ library(terra)
9
+ library(tidyterra)
10
+ library(sf)
11
+ library(duckdb)
12
+ library(duckdbfs)
13
+ library(glue)
14
+ library(tictoc)
15
+
16
+ # Connect to DuckDB
17
+ tcon <- dbConnect(duckdb::duckdb())
18
+
19
+ # Load spatial extension
20
+ tcon |> dbExecute("
21
+ INSTALL spatial;
22
+ LOAD spatial;
23
+ SET memory_limit = '200GB';
24
+ SET preserve_insertion_order = false;
25
+ SET threads TO 8;
26
+ ")
27
+
28
+ greenspace_shp <- "data/source/Greenspaces_osm_nad93/greenspaces_osm_nad83.shp"
29
+
30
+ greenspaces <- st_read(greenspace_shp, quiet = TRUE) |>
31
+ st_transform(3310)
32
+
33
+ # --- --- --- --- --- ---
34
+ # Commented out #
35
+ # --- --- --- --- --- ---
36
+
37
+ # # San Francisco county minus farallon islands
38
+ # sf <- st_read("data/source/CA_Counties/CA_Counties_TIGER2016.shp") %>%
39
+ # filter(NAME == "San Francisco") |>
40
+ # st_transform(3310) |>
41
+ # st_cast("POLYGON") %>%
42
+ # mutate(area = st_area(.)) |>
43
+ # slice_max(area)
44
+ #
45
+ library(tidycensus)
46
+ library(sf)
47
+ library(dplyr)
48
+
49
+ # requires a Census API key (install once via census_api_key())
50
+ sf <- get_acs(
51
+ geography = "county",
52
+ state = "CA",
53
+ variables = "B01003_001", # total population (dummy variable just to trigger geometry)
54
+ year = 2016,
55
+ geometry = TRUE
56
+ ) %>%
57
+ filter(NAME == "San Francisco County, California") %>%
58
+ st_transform(3310) %>%
59
+ st_cast("POLYGON") %>%
60
+ mutate(area = st_area(geometry)) %>%
61
+ slice_max(area, n = 1) %>%
62
+ select(GEOID, NAME, geometry)
63
+
64
+
65
+ empty.sr <- rast("data/source/slope.tif")
66
+
67
+
68
+ template.sr <- empty.sr %>%
69
+ # Give every cell an ID
70
+ mutate(cell_id = 1:ncell(.)) |>
71
+ # But filter to everything that isn't NA for slope values
72
+ filter(!is.na(prcnt_slope30)) |>
73
+ crop(sf |> st_transform(4326), mask = T) |>
74
+ crop(ext(c(-123, -122, 37.65, 37.85))) |>
75
+ trim()
76
+
77
+ # Convert template raster to points (centroids) for spatial joins
78
+ template_pts <- template.sr |>
79
+ as.points(na.rm = TRUE) |>
80
+ st_as_sf() |>
81
+ st_transform(4326) # Transform back to WGS84 for consistency
82
+
83
+ # Create directories if they don't exist
84
+ dir.create("data/intermediate", showWarnings = FALSE)
85
+ dir.create("data/output", showWarnings = FALSE)
86
+
87
+ # Write template to file for DuckDB
88
+ template_pts %>%
89
+ write_sf("data/intermediate/template_pts.gpkg")
90
+
91
+ # Load data into DuckDB and set up geometries
92
+ tcon %>% dbExecute(glue("
93
+ CREATE OR REPLACE TABLE greenspace_geo
94
+ AS
95
+ SELECT * EXCLUDE geom,
96
+ ST_TRANSFORM(geom, 'EPSG:4269', 'EPSG:3310', always_xy := true) AS geom3310,
97
+ ST_Centroid(geom) AS centroid_geom3310,
98
+ ST_SimplifyPreserveTopology(ST_TRANSFORM(geom, 'EPSG:4269', 'EPSG:3310', always_xy := true), 10) AS simple_geom3310
99
+ FROM ST_READ('{greenspace_shp}');
100
+
101
+ CREATE INDEX idx_grn_simple_geom ON greenspace_geo USING RTREE (simple_geom3310);
102
+ "))
103
+
104
+ # Load template grid into DuckDB
105
+ tcon %>% dbExecute("
106
+ CREATE OR REPLACE TABLE template_grid_geo
107
+ AS
108
+ SELECT
109
+ cell_id,
110
+ geom,
111
+ ST_TRANSFORM(geom, 'EPSG:4326', 'EPSG:3310', always_xy := true) AS geom3310
112
+ FROM ST_Read('data/intermediate/template_pts.gpkg');
113
+
114
+ CREATE INDEX idx_template_geom ON template_grid_geo USING RTREE (geom3310);
115
+ CREATE INDEX idx_template_cell_id ON template_grid_geo (cell_id);
116
+ ")
117
+
118
+
119
+ # Calculate nearest greenspace distances
120
+ nn_query <- glue("
121
+ CREATE OR REPLACE TABLE grn_distance_complete AS
122
+ WITH distances AS (
123
+ SELECT
124
+ template.cell_id,
125
+ template.geom AS template_geom,
126
+ ST_AsText(template.geom) AS geom_wkt,
127
+ osm_id,
128
+ ST_Distance(template.geom3310, green.simple_geom3310) AS distance_meters
129
+ FROM template_grid_geo AS template, greenspace_geo AS green
130
+ )
131
+ SELECT
132
+ cell_id,
133
+ template_geom,
134
+ geom_wkt,
135
+ MIN(distance_meters) AS distance_to_greenspace_meters,
136
+ arg_min(osm_id, distance_meters) AS nearest_greenspace_osmid
137
+ FROM distances
138
+ GROUP BY cell_id, template_geom, geom_wkt;
139
+ ")
140
+
141
+ cat("Calculating nearest greenspace distances...\n")
142
+ tic()
143
+ tcon %>% dbExecute(nn_query)
144
+ toc()
145
+ # Takes 138 seconds
146
+ # 287 seconds on mac
147
+
148
+ # Get results from database
149
+ sf_dist_complete <- tcon %>%
150
+ tbl("grn_distance_complete") %>%
151
+ select(cell_id, distance_to_greenspace_meters, nearest_greenspace_osmid) %>%
152
+ collect()
153
+
154
+ # Create empty raster based on template
155
+ empty_raster <- empty.sr
156
+ values(empty_raster) <- NA
157
+
158
+ # Join results back to preserve exact cell alignment
159
+ aligned_results <- tibble(cells = 1:ncell(empty.sr)) %>%
160
+ left_join(sf_dist_complete %>% rename(cells = cell_id), by = "cells")
161
+
162
+ # Create distance raster
163
+ greenspace_nearest_dist <- empty_raster
164
+ values(greenspace_nearest_dist) <- aligned_results$distance_to_greenspace_meters
165
+ names(greenspace_nearest_dist) <- "greenspace_nearest_dist"
166
+ cropped_greenspace_nearest_dist <- greenspace_nearest_dist |>
167
+ crop(sf |> st_transform(4326), mask = T) |>
168
+ crop(ext(c(-123, -122, 37.65, 37.85))) |>
169
+ trim()
170
+
171
+ # Create nearest greenspace ID raster
172
+ greenspace_nearest_osmid <- empty_raster
173
+ values(greenspace_nearest_osmid) <- as.numeric(aligned_results$nearest_greenspace_osmid)
174
+ names(greenspace_nearest_osmid) <- "greenspace_nearest_osmid"
175
+ cropped_greenspace_nearest_osmid <- greenspace_nearest_osmid |>
176
+ crop(sf |> st_transform(4326), mask = T) |>
177
+ crop(ext(c(-123, -122, 37.65, 37.85))) |>
178
+ trim()
179
+
180
+ plot(cropped_greenspace_nearest_dist)
181
+ plot(cropped_greenspace_nearest_osmid)
182
+
183
+ # Write rasters
184
+ writeRaster(cropped_greenspace_nearest_dist, "data/output/nearest_greenspace_dist.tif", overwrite = TRUE)
185
+ # Use INT8S (64-bit signed integer) to store OSM IDs without precision loss
186
+ writeRaster(cropped_greenspace_nearest_osmid, "data/output/nearest_greenspace_osmid.tif", overwrite = TRUE, datatype = "INT8S")
Rscripts/prep/making-rsfprogram-raster.R ADDED
@@ -0,0 +1,173 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ============================================================================
2
+ # Prep: Nearest RSF Program rasters (30 m, same grid as greenspace)
3
+ # ============================================================================
4
+ # Produces:
5
+ # - nearest_rsfprogram_dist.tif — distance (m) to nearest RSF program polygon
6
+ # - nearest_rsfprogram_id.tif — polygon_id for lookup in RSF gpkg (prj_name)
7
+ #
8
+ # Source polygons: data/source/RSF_Program_Projects_polygons.gpkg
9
+ # Template grid: same slope.tif + SF county mask as making-greenspace-raster.R
10
+ # ============================================================================
11
+
12
+ # library(tidyverse)
13
+ library(dplyr)
14
+ library(tidyr)
15
+ library(readr)
16
+ library(purrr)
17
+ library(tibble)
18
+ library(stringr)
19
+ library(forcats)
20
+ library(lubridate)
21
+ library(ggplot2)
22
+ library(terra)
23
+ library(tidyterra)
24
+ library(sf)
25
+ library(duckdb)
26
+ library(glue)
27
+ library(tictoc)
28
+ library(tidycensus)
29
+
30
+ # ============================================================================
31
+ # DuckDB + spatial
32
+ # ============================================================================
33
+ tcon <- dbConnect(duckdb::duckdb())
34
+
35
+ tcon |> dbExecute("
36
+ INSTALL spatial;
37
+ LOAD spatial;
38
+ SET memory_limit = '200GB';
39
+ SET preserve_insertion_order = false;
40
+ SET threads TO 8;
41
+ ")
42
+
43
+ rsf_gpkg <- "data/source/RSF_Program_Projects_polygons.gpkg"
44
+
45
+ # ============================================================================
46
+ # SF county boundary + template grid (match greenspace raster script)
47
+ # ============================================================================
48
+ sf <- get_acs(
49
+ geography = "county",
50
+ state = "CA",
51
+ variables = "B01003_001",
52
+ year = 2016,
53
+ geometry = TRUE
54
+ ) |>
55
+ filter(NAME == "San Francisco County, California") |>
56
+ st_transform(3310) |>
57
+ st_cast("POLYGON") |>
58
+ mutate(area = st_area(geometry)) |>
59
+ slice_max(area, n = 1) |>
60
+ select(GEOID, NAME, geometry)
61
+
62
+ empty.sr <- rast("data/source/slope.tif")
63
+
64
+ template.sr <- empty.sr %>%
65
+ mutate(cell_id = 1:ncell(.)) %>%
66
+ filter(!is.na(prcnt_slope30)) %>%
67
+ crop(sf %>% st_transform(4326), mask = TRUE) %>%
68
+ crop(ext(c(-123, -122, 37.65, 37.85))) %>%
69
+ trim()
70
+
71
+ template_pts <- template.sr %>%
72
+ as.points(na.rm = TRUE) %>%
73
+ st_as_sf() %>%
74
+ st_transform(4326)
75
+
76
+ dir.create("data/intermediate", showWarnings = FALSE)
77
+ dir.create("data/output", showWarnings = FALSE)
78
+
79
+ template_pts |>
80
+ write_sf("data/intermediate/template_pts.gpkg")
81
+
82
+ # Escape single quotes in path for SQL (Windows paths unlikely here)
83
+ rsf_gpkg_sql <- gsub("'", "''", normalizePath(rsf_gpkg, winslash = "/", mustWork = TRUE))
84
+
85
+ tcon |> dbExecute(glue("
86
+ CREATE OR REPLACE TABLE rsf_geo AS
87
+ SELECT
88
+ polygon_id,
89
+ prj_name,
90
+ ST_TRANSFORM(geom, 'EPSG:4326', 'EPSG:3310', always_xy := true) AS geom3310,
91
+ ST_SimplifyPreserveTopology(
92
+ ST_TRANSFORM(geom, 'EPSG:4326', 'EPSG:3310', always_xy := true),
93
+ 10
94
+ ) AS simple_geom3310
95
+ FROM ST_Read('{rsf_gpkg_sql}');
96
+
97
+ CREATE INDEX idx_rsf_simple_geom ON rsf_geo USING RTREE (simple_geom3310);
98
+ "))
99
+
100
+ tcon |> dbExecute("
101
+ CREATE OR REPLACE TABLE template_grid_geo AS
102
+ SELECT
103
+ cell_id,
104
+ geom,
105
+ ST_TRANSFORM(geom, 'EPSG:4326', 'EPSG:3310', always_xy := true) AS geom3310
106
+ FROM ST_Read('data/intermediate/template_pts.gpkg');
107
+
108
+ CREATE INDEX idx_template_geom ON template_grid_geo USING RTREE (geom3310);
109
+ CREATE INDEX idx_template_cell_id ON template_grid_geo (cell_id);
110
+ ")
111
+
112
+ nn_query <- glue("
113
+ CREATE OR REPLACE TABLE rsf_distance_complete AS
114
+ WITH distances AS (
115
+ SELECT
116
+ template.cell_id,
117
+ template.geom AS template_geom,
118
+ ST_AsText(template.geom) AS geom_wkt,
119
+ rsf.polygon_id,
120
+ ST_Distance(template.geom3310, rsf.simple_geom3310) AS distance_meters
121
+ FROM template_grid_geo AS template, rsf_geo AS rsf
122
+ )
123
+ SELECT
124
+ cell_id,
125
+ template_geom,
126
+ geom_wkt,
127
+ MIN(distance_meters) AS distance_to_rsf_meters,
128
+ arg_min(polygon_id, distance_meters) AS nearest_rsf_polygon_id
129
+ FROM distances
130
+ GROUP BY cell_id, template_geom, geom_wkt;
131
+ ")
132
+
133
+ cat("Calculating nearest RSF program distances...\n")
134
+ tic()
135
+ tcon |> dbExecute(nn_query)
136
+ toc()
137
+
138
+ rsf_dist_complete <- tcon |>
139
+ tbl("rsf_distance_complete") |>
140
+ select(cell_id, distance_to_rsf_meters, nearest_rsf_polygon_id) |>
141
+ collect()
142
+
143
+ empty_raster <- empty.sr
144
+ values(empty_raster) <- NA
145
+
146
+ aligned_results <- tibble(cells = 1:ncell(empty.sr)) |>
147
+ left_join(rsf_dist_complete |> rename(cells = cell_id), by = "cells")
148
+
149
+ rsf_nearest_dist <- empty_raster
150
+ values(rsf_nearest_dist) <- aligned_results$distance_to_rsf_meters
151
+ names(rsf_nearest_dist) <- "rsfprogram_nearest_dist"
152
+
153
+ cropped_rsf_dist <- rsf_nearest_dist %>%
154
+ crop(sf %>% st_transform(4326), mask = TRUE) %>%
155
+ crop(ext(c(-123, -122, 37.65, 37.85))) %>%
156
+ trim()
157
+
158
+ rsf_nearest_id <- empty_raster
159
+ values(rsf_nearest_id) <- as.integer(round(as.numeric(aligned_results$nearest_rsf_polygon_id)))
160
+ names(rsf_nearest_id) <- "rsfprogram_nearest_id"
161
+
162
+ cropped_rsf_id <- rsf_nearest_id %>%
163
+ crop(sf %>% st_transform(4326), mask = TRUE) %>%
164
+ crop(ext(c(-123, -122, 37.65, 37.85))) %>%
165
+ trim()
166
+
167
+ plot(cropped_rsf_dist)
168
+ plot(cropped_rsf_id)
169
+
170
+ writeRaster(cropped_rsf_dist, "data/output/nearest_rsfprogram_dist.tif", overwrite = TRUE)
171
+ writeRaster(cropped_rsf_id, "data/output/nearest_rsfprogram_id.tif", overwrite = TRUE, datatype = "INT4S")
172
+
173
+ dbDisconnect(tcon, shutdown = TRUE)
Rscripts/prep/profile_startup.R ADDED
@@ -0,0 +1,444 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ============================================================================
2
+ # Startup Performance Profiling
3
+ # ============================================================================
4
+ # This script benchmarks each data loading step in setup.R to identify
5
+ # bottlenecks and guide optimization decisions (precomputation, caching, etc.)
6
+
7
+ # library(tidyverse)
8
+ library(profvis)
9
+
10
+ # Create benchmark dataframe to track timings
11
+ benchmarks <- tibble(
12
+ step = character(),
13
+ description = character(),
14
+ time_sec = numeric(),
15
+ time_pct = numeric()
16
+ )
17
+
18
+ start_overall <- Sys.time()
19
+
20
+ # ============================================================================
21
+ # 1. Library Loading
22
+ # ============================================================================
23
+ t1 <- Sys.time()
24
+
25
+ library(shiny)
26
+ library(shinydashboard)
27
+ library(leaflet)
28
+ library(mapboxapi)
29
+ # library(tidyverse)
30
+ library(tidycensus)
31
+ library(sf)
32
+ library(DT)
33
+ library(RColorBrewer)
34
+ library(terra)
35
+ library(data.table)
36
+ library(mapview)
37
+ # library(sjPlot)
38
+ # library(sjlabelled)
39
+ library(bslib)
40
+ library(shinycssloaders)
41
+ library(DBI)
42
+ library(duckdb)
43
+ library(dbplyr)
44
+ library(gtfsrouter)
45
+ library(tidytransit)
46
+ library(fmsb)
47
+ library(scales)
48
+
49
+ time_libraries <- as.numeric(difftime(Sys.time(), t1, units = "secs"))
50
+ benchmarks <- bind_rows(benchmarks, tibble(
51
+ step = "00_libraries",
52
+ description = "Load all 23 packages",
53
+ time_sec = time_libraries,
54
+ time_pct = NA
55
+ ))
56
+
57
+ # ============================================================================
58
+ # 2. Greenspace (OSM)
59
+ # ============================================================================
60
+ t1 <- Sys.time()
61
+ osm_greenspace <- st_read("/vsicurl/https://huggingface.co/datasets/boettiger-lab/sf_biodiv_access/resolve/main/greenspaces_osm_nad83.shp", quiet = TRUE) |>
62
+ st_transform(4326)
63
+ if (!"name" %in% names(osm_greenspace)) {
64
+ osm_greenspace$name <- "Unnamed Greenspace"
65
+ }
66
+ time_greenspace <- as.numeric(difftime(Sys.time(), t1, units = "secs"))
67
+ benchmarks <- bind_rows(benchmarks, tibble(
68
+ step = "01_greenspace",
69
+ description = "Load OSM greenspace polygons",
70
+ time_sec = time_greenspace,
71
+ time_pct = NA
72
+ ))
73
+
74
+ # ============================================================================
75
+ # 3. Greenspace Distance Rasters
76
+ # ============================================================================
77
+ t1 <- Sys.time()
78
+ greenspace_dist_raster <- terra::rast("/vsicurl/https://huggingface.co/datasets/boettiger-lab/sf_biodiv_access/resolve/main/nearest_greenspace_dist.tif")
79
+ greenspace_osmid_raster <- terra::rast("/vsicurl/https://huggingface.co/datasets/boettiger-lab/sf_biodiv_access/resolve/main/nearest_greenspace_osmid.tif")
80
+ time_gs_rasters <- as.numeric(difftime(Sys.time(), t1, units = "secs"))
81
+ benchmarks <- bind_rows(benchmarks, tibble(
82
+ step = "02_gs_rasters",
83
+ description = "Load greenspace distance rasters",
84
+ time_sec = time_gs_rasters,
85
+ time_pct = NA
86
+ ))
87
+
88
+ # ============================================================================
89
+ # 4. NDVI Raster
90
+ # ============================================================================
91
+ t1 <- Sys.time()
92
+ ndvi <- terra::rast("/vsicurl/https://huggingface.co/datasets/boettiger-lab/sf_biodiv_access/resolve/main/SF_EastBay_NDVI_Sentinel_10.tif")
93
+ time_ndvi <- as.numeric(difftime(Sys.time(), t1, units = "secs"))
94
+ benchmarks <- bind_rows(benchmarks, tibble(
95
+ step = "03_ndvi",
96
+ description = "Load NDVI raster",
97
+ time_sec = time_ndvi,
98
+ time_pct = NA
99
+ ))
100
+
101
+ # ============================================================================
102
+ # 5. CBG Vector Data
103
+ # ============================================================================
104
+ t1 <- Sys.time()
105
+ download.file(
106
+ 'https://huggingface.co/datasets/boettiger-lab/sf_biodiv_access/resolve/main/cbg_vect_sf.Rdata',
107
+ '/tmp/cbg_vect_sf.Rdata',
108
+ mode = 'wb',
109
+ quiet = TRUE
110
+ )
111
+ load('/tmp/cbg_vect_sf.Rdata')
112
+
113
+ if (!"unique_species" %in% names(cbg_vect_sf)) {
114
+ cbg_vect_sf$unique_species <- cbg_vect_sf$n_species
115
+ }
116
+ if (!"n_observations" %in% names(cbg_vect_sf)) {
117
+ cbg_vect_sf$n_observations <- cbg_vect_sf$n
118
+ }
119
+ if (!"median_inc" %in% names(cbg_vect_sf)) {
120
+ cbg_vect_sf$median_inc <- cbg_vect_sf$medincE
121
+ }
122
+ if (!"ndvi_mean" %in% names(cbg_vect_sf)) {
123
+ cbg_vect_sf$ndvi_mean <- cbg_vect_sf$ndvi_sentinel
124
+ }
125
+
126
+ time_cbg <- as.numeric(difftime(Sys.time(), t1, units = "secs"))
127
+ benchmarks <- bind_rows(benchmarks, tibble(
128
+ step = "04_cbg",
129
+ description = "Download + load census block groups",
130
+ time_sec = time_cbg,
131
+ time_pct = NA
132
+ ))
133
+
134
+ # ============================================================================
135
+ # 6. Hotspots/Coldspots
136
+ # ============================================================================
137
+ t1 <- Sys.time()
138
+ biodiv_hotspots <- st_read("/vsicurl/https://huggingface.co/datasets/boettiger-lab/sf_biodiv_access/resolve/main/hotspots.shp", quiet = TRUE) |>
139
+ st_transform(4326)
140
+ biodiv_coldspots <- st_read("/vsicurl/https://huggingface.co/datasets/boettiger-lab/sf_biodiv_access/resolve/main/coldspots.shp", quiet = TRUE) |>
141
+ st_transform(4326)
142
+ time_hotcold <- as.numeric(difftime(Sys.time(), t1, units = "secs"))
143
+ benchmarks <- bind_rows(benchmarks, tibble(
144
+ step = "05_hotcold",
145
+ description = "Load hotspots and coldspots",
146
+ time_sec = time_hotcold,
147
+ time_pct = NA
148
+ ))
149
+
150
+ # ============================================================================
151
+ # 7. RSF Projects
152
+ # ============================================================================
153
+ t1 <- Sys.time()
154
+ rsf_projects <- st_read("data/source/RSF_Program_Projects_polygons.gpkg", quiet = TRUE) |>
155
+ st_transform(4326)
156
+ time_rsf <- as.numeric(difftime(Sys.time(), t1, units = "secs"))
157
+ benchmarks <- bind_rows(benchmarks, tibble(
158
+ step = "06_rsf",
159
+ description = "Load RSF program projects",
160
+ time_sec = time_rsf,
161
+ time_pct = NA
162
+ ))
163
+
164
+ # ============================================================================
165
+ # 8. GBIF Parquet Reference (just get path, don't load yet)
166
+ # ============================================================================
167
+ time_gbif_setup <- 0.001 # negligible
168
+ benchmarks <- bind_rows(benchmarks, tibble(
169
+ step = "07_gbif_setup",
170
+ description = "GBIF parquet path (lazy load in server)",
171
+ time_sec = time_gbif_setup,
172
+ time_pct = NA
173
+ ))
174
+
175
+ # ============================================================================
176
+ # 9. GTFS Data
177
+ # ============================================================================
178
+ t1 <- Sys.time()
179
+
180
+ gtfs_path <- '/Users/diegoellis/Desktop/RSF_next_steps/GPFS_OSM_Transit/sf_muni_gtfs-current/'
181
+
182
+ # Stops
183
+ gtfs_stops_sf <- tryCatch({
184
+ read.csv(file.path(gtfs_path, 'stops.txt')) |>
185
+ st_as_sf(coords = c("stop_lon", "stop_lat"), crs = 4326)
186
+ }, error = function(e) {
187
+ warning("GTFS stops failed: ", e$message)
188
+ NULL
189
+ })
190
+
191
+ # Route shapes
192
+ gtfs_shapes_raw <- read.csv(file.path(gtfs_path, 'shapes.txt'))
193
+ gtfs_trips_raw <- read.csv(file.path(gtfs_path, 'trips.txt'))
194
+ gtfs_routes_raw <- read.csv(file.path(gtfs_path, 'routes.txt'))
195
+
196
+ shape_route_map <- gtfs_trips_raw |>
197
+ distinct(shape_id, route_id)
198
+
199
+ route_meta <- gtfs_routes_raw |>
200
+ select(route_id, route_short_name, route_long_name, route_color) |>
201
+ mutate(route_color_hex = paste0("#", trimws(route_color)))
202
+
203
+ shapes_split <- gtfs_shapes_raw |>
204
+ arrange(shape_id, shape_pt_sequence) |>
205
+ group_by(shape_id) |>
206
+ group_split()
207
+
208
+ shape_geoms <- lapply(shapes_split, function(s) {
209
+ st_linestring(cbind(s$shape_pt_lon, s$shape_pt_lat))
210
+ })
211
+
212
+ gtfs_routes_sf <- st_sf(
213
+ shape_id = sapply(shapes_split, function(s) s$shape_id[1]),
214
+ geometry = st_sfc(shape_geoms, crs = 4326)
215
+ ) |>
216
+ left_join(shape_route_map, by = "shape_id") |>
217
+ left_join(route_meta, by = "route_id")
218
+
219
+ time_gtfs_basic <- as.numeric(difftime(Sys.time(), t1, units = "secs"))
220
+ benchmarks <- bind_rows(benchmarks, tibble(
221
+ step = "08_gtfs_basic",
222
+ description = "Load GTFS stops and route shapes",
223
+ time_sec = time_gtfs_basic,
224
+ time_pct = NA
225
+ ))
226
+
227
+ # ============================================================================
228
+ # 10. gtfsrouter Initialization
229
+ # ============================================================================
230
+ t1 <- Sys.time()
231
+
232
+ gtfs_router <- tryCatch({
233
+ gtfs_zip_path <- tempfile(fileext = ".zip")
234
+ old_wd <- getwd()
235
+ setwd(gtfs_path)
236
+ utils::zip(gtfs_zip_path, files = list.files('.', pattern = "\\.txt$"))
237
+ setwd(old_wd)
238
+
239
+ gr <- gtfsrouter::extract_gtfs(gtfs_zip_path)
240
+ gtfsrouter::gtfs_timetable(gr, day = "Monday")
241
+ }, error = function(e) {
242
+ warning("gtfsrouter failed: ", e$message)
243
+ NULL
244
+ })
245
+
246
+ time_gtfs_router <- as.numeric(difftime(Sys.time(), t1, units = "secs"))
247
+ benchmarks <- bind_rows(benchmarks, tibble(
248
+ step = "09_gtfs_router",
249
+ description = "Initialize gtfsrouter + timetable",
250
+ time_sec = time_gtfs_router,
251
+ time_pct = NA
252
+ ))
253
+
254
+ # ============================================================================
255
+ # 11. tidytransit Headways
256
+ # ============================================================================
257
+ t1 <- Sys.time()
258
+
259
+ gtfs_stop_headways <- tryCatch({
260
+ gt <- tidytransit::read_gtfs(gtfs_path)
261
+ tidytransit::get_stop_frequency(gt, start_time = 7 * 3600, end_time = 9 * 3600) |>
262
+ group_by(stop_id) |>
263
+ summarise(
264
+ mean_headway_min = mean(mean_headway, na.rm = TRUE) / 60,
265
+ n_departures_peak = sum(n_departures, na.rm = TRUE),
266
+ .groups = "drop"
267
+ ) |>
268
+ mutate(stop_id = as.character(stop_id))
269
+ }, error = function(e) {
270
+ warning("tidytransit failed: ", e$message)
271
+ NULL
272
+ })
273
+
274
+ if (!is.null(gtfs_stop_headways) && !is.null(gtfs_stops_sf)) {
275
+ gtfs_stops_sf <- gtfs_stops_sf |>
276
+ mutate(stop_id = as.character(stop_id)) |>
277
+ left_join(gtfs_stop_headways, by = "stop_id")
278
+ }
279
+
280
+ time_gtfs_headways <- as.numeric(difftime(Sys.time(), t1, units = "secs"))
281
+ benchmarks <- bind_rows(benchmarks, tibble(
282
+ step = "10_gtfs_headways",
283
+ description = "Compute transit stop headways (AM peak)",
284
+ time_sec = time_gtfs_headways,
285
+ time_pct = NA
286
+ ))
287
+
288
+ # ============================================================================
289
+ # 12. CalEnviroScreen
290
+ # ============================================================================
291
+ t1 <- Sys.time()
292
+
293
+ calenviro_path <- '/Users/diegoellis/Downloads/calenviroscreen40gdb_F_2021.gdb'
294
+ if (!file.exists(calenviro_path)) {
295
+ calenviro_path <- '/Users/diegoellis/Desktop/Projects/Presentations/Data_Schell_Lab_Tutorial/calenviroscreen40gdb_F_2021.gdb'
296
+ }
297
+
298
+ cenv_sf <- tryCatch({
299
+ sf::st_read(calenviro_path, quiet = TRUE) |>
300
+ dplyr::filter(grepl("san francisco", County, ignore.case = TRUE), !is.na(CIscore)) |>
301
+ dplyr::select(
302
+ Tract, CIscore, CIscoreP,
303
+ PM2_5, PM2_5_Pctl, Traffic, Traffic_Pctl,
304
+ Poverty, Poverty_Pctl, HousBurd, HousBurd_Pctl,
305
+ County
306
+ ) |>
307
+ sf::st_transform(4326) |>
308
+ sf::st_make_valid()
309
+ }, error = function(e) {
310
+ warning("CalEnviroScreen failed: ", e$message)
311
+ NULL
312
+ })
313
+
314
+ time_calenviro <- as.numeric(difftime(Sys.time(), t1, units = "secs"))
315
+ benchmarks <- bind_rows(benchmarks, tibble(
316
+ step = "11_calenviro",
317
+ description = "Load CalEnviroScreen layer",
318
+ time_sec = time_calenviro,
319
+ time_pct = NA
320
+ ))
321
+
322
+ # ============================================================================
323
+ # 13. SF EJ Communities
324
+ # ============================================================================
325
+ t1 <- Sys.time()
326
+
327
+ sf_ej_path <- '/Users/diegoellis/Downloads/San Francisco Environmental Justice Communities Map_20251217/geo_export_a21b0a0a-7306-46fd-8381-06581cdbe6e9.shp'
328
+
329
+ sf_ej_sf <- tryCatch({
330
+ sf::st_read(sf_ej_path, quiet = TRUE) |>
331
+ dplyr::mutate(
332
+ symbol_hex = stringr::str_split(symbol_rgb, ",\\s*") |>
333
+ lapply(function(x) sprintf("#%02X%02X%02X",
334
+ as.integer(x[1]), as.integer(x[2]), as.integer(x[3]))) |>
335
+ unlist(),
336
+ ej_label = dplyr::case_when(
337
+ is.na(score) ~ "Not EJ",
338
+ score >= 21 ~ "High EJ burden (21-30)",
339
+ score >= 11 ~ "Moderate EJ burden (11-20)",
340
+ score >= 1 ~ "Low EJ burden (1-10)",
341
+ score == 0 ~ "Score 0",
342
+ TRUE ~ "Unknown"
343
+ )
344
+ ) |>
345
+ sf::st_transform(4326) |>
346
+ sf::st_make_valid()
347
+ }, error = function(e) {
348
+ warning("SF EJ layer failed: ", e$message)
349
+ NULL
350
+ })
351
+
352
+ time_ej <- as.numeric(difftime(Sys.time(), t1, units = "secs"))
353
+ benchmarks <- bind_rows(benchmarks, tibble(
354
+ step = "12_ej",
355
+ description = "Load SF EJ communities layer",
356
+ time_sec = time_ej,
357
+ time_pct = NA
358
+ ))
359
+
360
+ # ============================================================================
361
+ # SUMMARY
362
+ # ============================================================================
363
+ time_overall <- as.numeric(difftime(Sys.time(), start_overall, units = "secs"))
364
+
365
+ benchmarks <- benchmarks |>
366
+ mutate(time_pct = round(100 * time_sec / time_overall, 1))
367
+
368
+ # Print results
369
+ cat("\n")
370
+ cat("================================================================================\n")
371
+ cat("STARTUP PERFORMANCE BENCHMARK\n")
372
+ cat("================================================================================\n\n")
373
+
374
+ print(benchmarks |> select(step, description, time_sec, time_pct))
375
+
376
+ cat("\n")
377
+ cat("TOTAL STARTUP TIME: ", round(time_overall, 2), " seconds\n")
378
+ cat("================================================================================\n\n")
379
+
380
+ # ============================================================================
381
+ # OPTIMIZATION RECOMMENDATIONS
382
+ # ============================================================================
383
+ cat("OPTIMIZATION RECOMMENDATIONS:\n")
384
+ cat("================================================================================\n\n")
385
+
386
+ # Flag slow steps (>5 seconds or >10% of total)
387
+ slow_steps <- benchmarks |>
388
+ filter(time_sec > 5 | time_pct > 10) |>
389
+ arrange(desc(time_sec))
390
+
391
+ if (nrow(slow_steps) > 0) {
392
+ cat("CRITICAL BOTTLENECKS (>5s or >10%):\n\n")
393
+ for (i in 1:nrow(slow_steps)) {
394
+ row <- slow_steps[i, ]
395
+ cat(" •", row$step, "(", row$time_sec, "s,", row$time_pct, "%)\n")
396
+ cat(" Description:", row$description, "\n")
397
+
398
+ if (grepl("libraries", row$step)) {
399
+ cat(" Recommendation: Load packages only in server() if possible. Use lazy loading.\n")
400
+ } else if (grepl("download", row$step)) {
401
+ cat(" Recommendation: Cache downloaded files locally or use precomputed versions.\n")
402
+ } else if (grepl("gtfsrouter|tidytransit", row$step)) {
403
+ cat(" Recommendation: Pre-compute and cache GTFS timetable. Consider lazy loading for session.\n")
404
+ } else if (grepl("calenviro|ej", row$step)) {
405
+ cat(" Recommendation: Pre-filter to SF boundary. Store as .gpkg or parquet locally.\n")
406
+ }
407
+ cat("\n")
408
+ }
409
+ } else {
410
+ cat("No critical bottlenecks detected (all steps < 5s).\n\n")
411
+ }
412
+
413
+ # Create visualization
414
+ p <- ggplot(benchmarks, aes(x = reorder(step, -time_sec), y = time_sec, fill = time_pct)) +
415
+ geom_col() +
416
+ geom_text(aes(label = paste0(round(time_sec, 2), "s")), vjust = -0.3, size = 3) +
417
+ scale_fill_gradient(low = "green", high = "red", name = "% of Total") +
418
+ labs(
419
+ title = "Shiny App Startup Performance Profile",
420
+ subtitle = paste0("Total time: ", round(time_overall, 2), " seconds"),
421
+ x = "Loading Step",
422
+ y = "Time (seconds)"
423
+ ) +
424
+ theme_minimal() +
425
+ theme(
426
+ axis.text.x = element_text(angle = 45, hjust = 1, size = 9),
427
+ plot.title = element_text(face = "bold", size = 14),
428
+ legend.position = "right"
429
+ )
430
+
431
+ print(p)
432
+
433
+ ggsave("/Users/diegoellis/Desktop/Projects/Postdoc/Biodiversity_Access_Indicator/SF_biodiv_access_shiny/startup_benchmark.png",
434
+ p, width = 12, height = 6, dpi = 150)
435
+
436
+ cat("\nBenchmark plot saved to: startup_benchmark.png\n")
437
+
438
+ # ============================================================================
439
+ # Export benchmark data
440
+ # ============================================================================
441
+ write_csv(benchmarks,
442
+ "/Users/diegoellis/Desktop/Projects/Postdoc/Biodiversity_Access_Indicator/SF_biodiv_access_shiny/startup_benchmarks.csv")
443
+
444
+ cat("Benchmark data saved to: startup_benchmarks.csv\n\n")
Rscripts/prep/run_all_prep.R ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ============================================================================
2
+ # Prep: Run full data build → data/output
3
+ # ============================================================================
4
+ # Working directory must be the project root (folder containing data/ and Rscripts/).
5
+ #
6
+ # Order (fast steps first; heavy raster build last):
7
+ # 1. GBIF DuckDB → sf-gbif.parquet → gbif_census_ndvi_anno.parquet
8
+ # 2. CalEnviroScreen + SF EJ → GeoPackages
9
+ # 3. GTFS zip + timetable + stop headways (~20–30 s; cached for app startup)
10
+ # 4. Greenspace distance rasters (DuckDB; slow)
11
+ # 5. RSF Program nearest rasters (DuckDB; same grid as 4)
12
+ # 6. Per-CBG greenspace coverage CSV (precomputed so the app's first isochrone
13
+ # doesn't recompute the CBG x greenspace intersection)
14
+ #
15
+ # Then upload files from data/output/ to the HuggingFace dataset (manually or
16
+ # see comments in upload_to_huggingface.R).
17
+
18
+ source("Rscripts/prep/create_annotated_gbif_parquet.R")
19
+ source("Rscripts/prep/build_equity_layers.R")
20
+ source("Rscripts/prep/implement_optimizations.R")
21
+ source("Rscripts/prep/making-greenspace-raster.R")
22
+ source("Rscripts/prep/making-rsfprogram-raster.R")
23
+ source("Rscripts/prep/build_cbg_greenspace_coverage.R")
Rscripts/prep/upload_to_huggingface.R ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ============================================================================
2
+ # Prep: Stage data/output → data/hf_upload (optional mirror before upload)
3
+ # ============================================================================
4
+ # Canonical build artifacts live in data/output/ (from Rscripts/prep/* and
5
+ # run_all_prep.R). Copy them here, then upload data/hf_upload/ to HuggingFace:
6
+ #
7
+ # huggingface-cli upload boettiger-lab/sf_biodiv_access data/hf_upload/ . \
8
+ # --repo-type dataset
9
+ #
10
+ # Or upload files from data/output/ directly via the web UI.
11
+ #
12
+ # Already on HuggingFace (not staged from data/output): greenspaces_osm_nad83.{shp,...}
13
+ # Other manual assets: SF_EastBay_NDVI_Sentinel_10.tif, cbg_vect_sf.Rdata, hotspots/coldspots
14
+ #
15
+ # GTFS: sf_muni_gtfs.zip contains the feed; timetable .rds + headways .csv are
16
+ # precomputed (~20–30 s) so the app does not rebuild them every session.
17
+ # ============================================================================
18
+
19
+ library(glue)
20
+
21
+ out_dir <- "data/output"
22
+ upload_dir <- "data/hf_upload"
23
+ dir.create(upload_dir, recursive = TRUE, showWarnings = FALSE)
24
+
25
+ artifacts <- c(
26
+ "nearest_greenspace_dist.tif",
27
+ "nearest_greenspace_osmid.tif",
28
+ "nearest_rsfprogram_dist.tif",
29
+ "nearest_rsfprogram_id.tif",
30
+ "gbif_census_ndvi_anno.parquet",
31
+ "gtfs_timetable_monday.rds",
32
+ "gtfs_stop_headways.csv",
33
+ "sf_muni_gtfs.zip",
34
+ "calenviro_sf.gpkg",
35
+ "sf_ej_communities_map.gpkg",
36
+ "RSF_Program_Projects_polygons.gpkg",
37
+ "cbg_greenspace_coverage.csv"
38
+ )
39
+
40
+ for (f in artifacts) {
41
+ src <- file.path(out_dir, f)
42
+ dst <- file.path(upload_dir, f)
43
+ if (file.exists(src)) {
44
+ file.copy(src, dst, overwrite = TRUE)
45
+ } else {
46
+ warning(glue("Missing in data/output/: {f}"))
47
+ }
48
+ }
Rscripts/setup_unified.R ADDED
@@ -0,0 +1,284 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ============================================================================
2
+ # Setup: load all data for the app (sourced once at startup by app.R)
3
+ # ============================================================================
4
+ # Checks local files in data/cached/ first, falling back to HuggingFace
5
+ # downloads (see hf_or_local()). Loads everything the app needs in one pass.
6
+ #
7
+ # The two most expensive products -- the CBG x greenspace intersection and the
8
+ # transit-routing timetable -- are precomputed in Rscripts/prep/ and just read
9
+ # here, so startup stays fast without splitting the load across files.
10
+ # ============================================================================
11
+
12
+ # ----------------------------------------------------------------------------
13
+ # Libraries
14
+ # ----------------------------------------------------------------------------
15
+ # require(shinyjs) # commented: useShinyjs() enabled features the app never calls
16
+ library(shiny)
17
+ library(shinydashboard)
18
+ library(leaflet)
19
+ library(mapboxapi)
20
+ # library(tidyverse)
21
+ library(dplyr)
22
+ library(tidyr)
23
+ library(readr)
24
+ library(purrr)
25
+ library(tibble)
26
+ library(stringr)
27
+ library(forcats)
28
+ library(lubridate)
29
+ library(ggplot2)
30
+ # library(tidycensus) # commented: unused in app (CBG data is precomputed) -- speeds startup
31
+ library(sf)
32
+ library(DT)
33
+ # library(RColorBrewer) # commented: brewer.pal only appears in dead if(FALSE) blocks
34
+ library(terra)
35
+ # Commented out: no detectable use in the app, and loading them added ~1-2 s to
36
+ # startup (mapview/sjPlot/sjlabelled each pull large dependency trees). Re-enable
37
+ # the relevant line if you start using one.
38
+ # library(data.table)
39
+ # library(mapview)
40
+ # library(sjPlot)
41
+ # library(sjlabelled)
42
+ library(bslib)
43
+ library(shinycssloaders)
44
+ library(glue)
45
+
46
+ # ----------------------------------------------------------------------------
47
+ # HuggingFace base URL + cache dir + download helper
48
+ # ----------------------------------------------------------------------------
49
+ HF_BASE <- "https://huggingface.co/datasets/boettiger-lab/sf_biodiv_access/resolve/main"
50
+
51
+ # Use data/cached/ when running locally (writable), otherwise fall back to
52
+ # /tmp/sf_biodiv_cache/ for read-only environments like HuggingFace Spaces.
53
+ cache_dir <- if (file.access(".", mode = 2) == 0) "data/cached" else "/tmp/sf_biodiv_cache"
54
+ dir.create(cache_dir, recursive = TRUE, showWarnings = FALSE)
55
+
56
+ # Helper: if the file already exists in cache_dir, return that path. Otherwise
57
+ # attempt to download from HuggingFace into cache_dir. Returns the destination
58
+ # path regardless -- caller must check file.exists() if the download may fail.
59
+ hf_or_local <- function(filename) {
60
+ dest <- file.path(cache_dir, filename)
61
+ if (!file.exists(dest)) {
62
+ tryCatch(
63
+ download.file(glue::glue("{HF_BASE}/{filename}"), dest, mode = "wb", quiet = TRUE),
64
+ error = function(e) warning(glue::glue("HuggingFace download failed for {filename}: {e$message}")),
65
+ warning = function(w) warning(glue::glue("HuggingFace download warning for {filename}: {w$message}"))
66
+ )
67
+ }
68
+ dest
69
+ }
70
+
71
+ message("[setup] loading greenspace, CBG, RSF, CalEnviroScreen, SF EJ…")
72
+
73
+ # ----------------------------------------------------------------------------
74
+ # Greenspace (OSM polygons) -- "Greenspace" map layer + coverage calc
75
+ # ----------------------------------------------------------------------------
76
+ greenspace_shp <- file.path(cache_dir, "greenspaces_osm_nad83.shp")
77
+ if (!file.exists(greenspace_shp)) {
78
+ for (ext in c("shp", "dbf", "prj", "shx")) {
79
+ hf_or_local(glue::glue("greenspaces_osm_nad83.{ext}"))
80
+ }
81
+ }
82
+ osm_greenspace <- st_read(greenspace_shp, quiet = TRUE) |> st_transform(4326)
83
+ if (!"name" %in% names(osm_greenspace)) osm_greenspace$name <- "Unnamed Greenspace"
84
+
85
+ # ----------------------------------------------------------------------------
86
+ # GBIF observations (parquet) -- path only; queried via DuckDB in app.R / server
87
+ # ----------------------------------------------------------------------------
88
+ gbif_parquet <- hf_or_local("gbif_census_ndvi_anno.parquet")
89
+
90
+ # ----------------------------------------------------------------------------
91
+ # Census block groups (CBG) -- Income / Richness / Data map layers
92
+ # ----------------------------------------------------------------------------
93
+ load(hf_or_local("cbg_vect_sf.Rdata"))
94
+
95
+ if (!"unique_species" %in% names(cbg_vect_sf)) cbg_vect_sf$unique_species <- cbg_vect_sf$n_species
96
+ if (!"n_observations" %in% names(cbg_vect_sf)) cbg_vect_sf$n_observations <- cbg_vect_sf$n
97
+ if (!"median_inc" %in% names(cbg_vect_sf)) cbg_vect_sf$median_inc <- cbg_vect_sf$medincE
98
+ if (!"ndvi_mean" %in% names(cbg_vect_sf)) cbg_vect_sf$ndvi_mean <- cbg_vect_sf$ndvi_sentinel
99
+
100
+ # ----------------------------------------------------------------------------
101
+ # RSF Program Projects -- "RSF Program Projects" map layer
102
+ # ----------------------------------------------------------------------------
103
+ rsf_projects <- st_read(hf_or_local("RSF_Program_Projects_polygons.gpkg"), quiet = TRUE) |>
104
+ st_transform(4326)
105
+
106
+ # ----------------------------------------------------------------------------
107
+ # CalEnviroScreen 4.0 (pre-filtered to SF) -- map overlay
108
+ # ----------------------------------------------------------------------------
109
+ cenv_sf <- tryCatch({
110
+ sf::st_read(hf_or_local("calenviro_sf.gpkg"), quiet = TRUE)
111
+ }, error = function(e) {
112
+ warning("CalEnviroScreen failed to load: ", e$message); NULL
113
+ })
114
+
115
+ # ----------------------------------------------------------------------------
116
+ # SF Environmental Justice Communities -- map overlay
117
+ # ----------------------------------------------------------------------------
118
+ sf_ej_sf <- tryCatch({
119
+ sf::st_read(hf_or_local("sf_ej_communities_map.gpkg"), quiet = TRUE) |>
120
+ dplyr::mutate(
121
+ symbol_hex = stringr::str_split(symbol_rgb, ",\\s*") |>
122
+ lapply(function(x) sprintf("#%02X%02X%02X",
123
+ as.integer(x[1]), as.integer(x[2]), as.integer(x[3]))) |>
124
+ unlist(),
125
+ ej_label = dplyr::case_when(
126
+ is.na(score) ~ "Not EJ",
127
+ score >= 21 ~ "High EJ burden (21-30)",
128
+ score >= 11 ~ "Moderate EJ burden (11-20)",
129
+ score >= 1 ~ "Low EJ burden (1-10)",
130
+ score == 0 ~ "Score 0",
131
+ TRUE ~ "Unknown"
132
+ )
133
+ )
134
+ }, error = function(e) {
135
+ warning("SF EJ layer failed to load: ", e$message); NULL
136
+ })
137
+
138
+ message("[setup] loading distance rasters + NDVI (terra lazy: headers only)…")
139
+
140
+ # ----------------------------------------------------------------------------
141
+ # Greenspace + RSF distance/id rasters + NDVI (terra is lazy: reads headers, not
142
+ # cells, so these open instantly; pixels are read during isochrone analysis).
143
+ # ----------------------------------------------------------------------------
144
+ greenspace_dist_raster <- terra::rast(hf_or_local("nearest_greenspace_dist.tif"))
145
+ greenspace_osmid_raster <- terra::rast(hf_or_local("nearest_greenspace_osmid.tif"))
146
+ rsfprogram_dist_raster <- terra::rast(hf_or_local("nearest_rsfprogram_dist.tif"))
147
+ rsfprogram_id_raster <- terra::rast(hf_or_local("nearest_rsfprogram_id.tif"))
148
+ ndvi <- terra::rast(hf_or_local("SF_EastBay_NDVI_Sentinel_10.tif"))
149
+
150
+ # ----------------------------------------------------------------------------
151
+ # Per-CBG greenspace overlap (used in the socio summary + BAI benchmark).
152
+ # Precomputed by Rscripts/prep/build_cbg_greenspace_coverage.R and cached as a
153
+ # small CSV (downloaded from HuggingFace like the other prep artifacts). If that
154
+ # file is unavailable, fall back to the (slow) st_union + st_intersection here.
155
+ # ----------------------------------------------------------------------------
156
+ cov_path <- hf_or_local("cbg_greenspace_coverage.csv")
157
+ cbg_greenspace_coverage <- if (file.exists(cov_path)) {
158
+ # GEOID forced to character so leading zeros survive (FIPS codes; join key).
159
+ readr::read_csv(cov_path, col_types = readr::cols(GEOID = readr::col_character()),
160
+ show_col_types = FALSE)
161
+ } else {
162
+ message("[setup] precomputed greenspace coverage not found; computing inline (slow)…")
163
+ cbg_proj <- st_transform(cbg_vect_sf[, "GEOID"], 3857) |>
164
+ mutate(cbg_area_m2 = as.numeric(st_area(geometry)))
165
+ gs_proj <- st_transform(osm_greenspace, 3857) |> st_make_valid()
166
+ gs_union <- st_union(gs_proj)
167
+ cbg_gs_inter <- st_intersection(cbg_proj, gs_union)
168
+ coverage <- cbg_gs_inter |>
169
+ mutate(greenspace_m2 = as.numeric(st_area(geometry))) |>
170
+ st_drop_geometry() |>
171
+ group_by(GEOID) |>
172
+ summarise(greenspace_m2 = sum(greenspace_m2), .groups = "drop") |>
173
+ right_join(cbg_proj |> st_drop_geometry() |> dplyr::select(GEOID, cbg_area_m2), by = "GEOID") |>
174
+ mutate(
175
+ greenspace_m2 = tidyr::replace_na(greenspace_m2, 0),
176
+ GEOID = as.character(GEOID)
177
+ )
178
+ readr::write_csv(coverage, file.path(cache_dir, "cbg_greenspace_coverage.csv"))
179
+ coverage
180
+ }
181
+
182
+ message("[setup] loading GTFS (stops, routes, headways, routing timetable)…")
183
+
184
+ # ----------------------------------------------------------------------------
185
+ # GTFS (SF Muni)
186
+ # ----------------------------------------------------------------------------
187
+ gtfs_zip_path <- hf_or_local("sf_muni_gtfs.zip")
188
+
189
+ # Unzip for read.csv(stops.txt, …); gtfsrouter/tidytransit read the .zip directly.
190
+ gtfs_unzip_dir <- file.path(cache_dir, "muni_gtfs")
191
+ dir.create(gtfs_unzip_dir, recursive = TRUE, showWarnings = FALSE)
192
+ if (!dir.exists(gtfs_unzip_dir) || length(list.files(gtfs_unzip_dir, pattern = "\\.txt$")) == 0L) {
193
+ unzip(gtfs_zip_path, exdir = gtfs_unzip_dir, overwrite = TRUE)
194
+ }
195
+ gtfs_path <- gtfs_unzip_dir
196
+
197
+ # --- Transit stops ("Transit Stops" overlay) --------------------------------
198
+ gtfs_stops_sf <- tryCatch({
199
+ read.csv(file.path(gtfs_path, "stops.txt")) |>
200
+ st_as_sf(coords = c("stop_lon", "stop_lat"), crs = 4326)
201
+ }, error = function(e) { warning("GTFS stops failed to load: ", e$message); NULL })
202
+
203
+ # --- Route shapes ("Transit Routes" overlay) --------------------------------
204
+ gtfs_routes_sf <- tryCatch({
205
+ gtfs_shapes_raw <- read.csv(file.path(gtfs_path, "shapes.txt"))
206
+ gtfs_trips_raw <- read.csv(file.path(gtfs_path, "trips.txt"))
207
+ gtfs_routes_raw <- read.csv(file.path(gtfs_path, "routes.txt"))
208
+
209
+ shape_route_map <- gtfs_trips_raw |> distinct(shape_id, route_id)
210
+ route_meta <- gtfs_routes_raw |>
211
+ select(route_id, route_short_name, route_long_name, route_color) |>
212
+ mutate(route_color_hex = paste0("#", trimws(route_color)))
213
+
214
+ shapes_split <- gtfs_shapes_raw |>
215
+ arrange(shape_id, shape_pt_sequence) |>
216
+ group_by(shape_id) |>
217
+ group_split()
218
+
219
+ shape_geoms <- lapply(shapes_split, function(s) {
220
+ st_linestring(cbind(s$shape_pt_lon, s$shape_pt_lat))
221
+ })
222
+
223
+ st_sf(
224
+ shape_id = sapply(shapes_split, function(s) s$shape_id[1]),
225
+ geometry = st_sfc(shape_geoms, crs = 4326)
226
+ ) |>
227
+ left_join(shape_route_map, by = "shape_id") |>
228
+ left_join(route_meta, by = "route_id")
229
+ }, error = function(e) { warning("GTFS route shapes failed to load: ", e$message); NULL })
230
+
231
+ # --- Stop headways (AM peak 7-9am): cached as CSV (readable / diffable).
232
+ # Joined into gtfs_stops_sf below so they show in the stop popups on the map.
233
+ hw_csv <- file.path(cache_dir, "gtfs_stop_headways.csv")
234
+ hw_rds <- file.path(cache_dir, "gtfs_stop_headways.rds")
235
+ if (!file.exists(hw_csv) && file.exists(hw_rds)) {
236
+ readRDS(hw_rds) |> readr::write_csv(hw_csv)
237
+ }
238
+
239
+ gtfs_stop_headways <- tryCatch({
240
+ headways_path <- hf_or_local("gtfs_stop_headways.csv")
241
+ if (file.exists(headways_path)) {
242
+ readr::read_csv(headways_path, show_col_types = FALSE) |>
243
+ mutate(stop_id = as.character(stop_id))
244
+ } else {
245
+ gt <- tidytransit::read_gtfs(gtfs_zip_path)
246
+ hw <- tidytransit::get_stop_frequency(gt, start_time = 7 * 3600, end_time = 9 * 3600) |>
247
+ group_by(stop_id) |>
248
+ summarise(
249
+ mean_headway_min = mean(mean_headway, na.rm = TRUE) / 60,
250
+ n_departures_peak = sum(n_departures, na.rm = TRUE),
251
+ .groups = "drop"
252
+ ) |>
253
+ mutate(stop_id = as.character(stop_id))
254
+ readr::write_csv(hw, hw_csv)
255
+ hw
256
+ }
257
+ }, error = function(e) { warning("tidytransit headway computation failed: ", e$message); NULL })
258
+
259
+ if (!is.null(gtfs_stop_headways) && !is.null(gtfs_stops_sf)) {
260
+ gtfs_stops_sf <- gtfs_stops_sf |>
261
+ mutate(stop_id = as.character(stop_id)) |>
262
+ left_join(gtfs_stop_headways, by = "stop_id")
263
+ }
264
+
265
+ # --- gtfsrouter timetable: routes transit isochrones (precomputed .rds) ------
266
+ gtfs_router <- tryCatch({
267
+ timetable_path <- hf_or_local("gtfs_timetable_monday.rds")
268
+ if (file.exists(timetable_path)) {
269
+ readRDS(timetable_path)
270
+ } else {
271
+ gr <- gtfsrouter::extract_gtfs(gtfs_zip_path)
272
+ result <- gtfsrouter::gtfs_timetable(gr, day = "Monday")
273
+ saveRDS(result, file.path(cache_dir, "gtfs_timetable_monday.rds"))
274
+ result
275
+ }
276
+ }, error = function(e) { warning("gtfsrouter failed to initialise: ", e$message); NULL })
277
+
278
+ # --- Pre-computed transit isochrone cache (keyed by nearest stop + time) -----
279
+ transit_iso_cache <- tryCatch({
280
+ p <- file.path(cache_dir, "transit_iso_cache.rds")
281
+ if (file.exists(p)) readRDS(p) else NULL
282
+ }, error = function(e) { NULL })
283
+
284
+ message("[setup] data load complete.")
SF_biodiv_access_shiny.Rproj ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Version: 1.0
2
+
3
+ RestoreWorkspace: Default
4
+ SaveWorkspace: Default
5
+ AlwaysSaveHistory: Default
6
+
7
+ EnableCodeIndexing: Yes
8
+ UseSpacesForTab: Yes
9
+ NumSpacesForTab: 2
10
+ Encoding: UTF-8
11
+
12
+ RnwWeave: Sweave
13
+ LaTeX: pdfLaTeX
app.R ADDED
The diff for this file is too large to render. See raw diff
 
code/iso_metric_definitions.R ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # =============================================================================
2
+ # Shared raw-metric definitions for SF Biodiversity Access isochrone scoring
3
+ # =============================================================================
4
+ # This file defines the scientific support used by BOTH:
5
+ # * Rscripts/iso_metrics.R (interactive user-generated isochrones)
6
+ # * 03_compute_isochrone_basi.R (citywide precomputed reference isochrones)
7
+ #
8
+ # NON-NEGOTIABLE SUPPORT DEFINITION
9
+ # ---------------------------------
10
+ # For each isochrone independently, the metric support is the set of SF 100 m
11
+ # grid cells whose projected grid-cell centroids fall within that single
12
+ # isochrone polygon. Metrics are then calculated from those selected cells.
13
+ # This deliberately makes the interactive and precomputed estimands identical.
14
+ #
15
+ # It is a 100 m grid-supported access statistic, not a claim of pixel-perfect
16
+ # clipping at the isochrone boundary.
17
+ # =============================================================================
18
+
19
+ ISO_METRIC_DEFINITION_VERSION <- "sf_bai_grid100m_centroid_support_v1"
20
+ # UTM Zone 10N is appropriate for San Francisco metric area/distance calculations.
21
+ # Do not use Web Mercator (EPSG:3857) for area-based density denominators.
22
+ ISO_METRIC_CRS <- 26910
23
+
24
+ ISO_PLANT_CLASSES <- c(
25
+ "Magnoliopsida", "Liliopsida", "Pinopsida", "Polypodiopsida",
26
+ "Equisetopsida", "Bryopsida", "Marchantiopsida"
27
+ )
28
+
29
+ iso_weighted_mean <- function(x, w) {
30
+ x <- suppressWarnings(as.numeric(x))
31
+ w <- suppressWarnings(as.numeric(w))
32
+ ok <- is.finite(x) & is.finite(w) & w > 0
33
+ if (!any(ok)) return(NA_real_)
34
+ sum(x[ok] * w[ok]) / sum(w[ok])
35
+ }
36
+
37
+ iso_safe_sum <- function(x, na_as_zero = TRUE) {
38
+ x <- suppressWarnings(as.numeric(x))
39
+ if (na_as_zero) x[!is.finite(x)] <- 0
40
+ if (!any(is.finite(x))) return(NA_real_)
41
+ sum(x[is.finite(x)])
42
+ }
43
+
44
+ iso_safe_min <- function(x) {
45
+ x <- suppressWarnings(as.numeric(x))
46
+ x <- x[is.finite(x)]
47
+ if (length(x) == 0) return(NA_real_)
48
+ min(x)
49
+ }
50
+
51
+ # Clean, explicit reference schema. `direction` describes whether a larger raw
52
+ # value is better for the access score. Step 5 stores raw values plus pctile_*
53
+ # columns; the Shiny app uses the RAW columns for external-observation scoring.
54
+ ISO_REFERENCE_METRICS <- list(
55
+ n_species_accessed = list(direction = "higher", label = "Species richness"),
56
+ n_birds_accessed = list(direction = "higher", label = "Bird richness"),
57
+ n_mammals_accessed = list(direction = "higher", label = "Mammal richness"),
58
+ n_plants_accessed = list(direction = "higher", label = "Plant richness"),
59
+ n_records_accessed = list(direction = "higher", label = "GBIF records"),
60
+ sampling_density_km2 = list(direction = "higher", label = "GBIF records / km2"),
61
+ mean_ndvi = list(direction = "higher", label = "Mean NDVI"),
62
+ greenspace_pct = list(direction = "higher", label = "Greenspace cover"),
63
+ n_stops_accessed = list(direction = "higher", label = "Transit stops"),
64
+ transit_access_score = list(direction = "higher", label = "Transit stops / km2"),
65
+ unique_muni_routes = list(direction = "higher", label = "Unique Muni routes"),
66
+ nearest_stop_m = list(direction = "lower", label = "Nearest stop distance"),
67
+ calenviro_ci_score = list(direction = "lower", label = "CalEnviroScreen burden"),
68
+ calenviro_traffic_pct = list(direction = "lower", label = "Traffic burden percentile"),
69
+ sf_ej_score = list(direction = "lower", label = "SF EJ burden")
70
+ )
code/iso_metrics_AUDITED.R ADDED
@@ -0,0 +1,688 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ============================================================================
2
+ # Shared isochrone + metric functions (sourced once by app.R at startup)
3
+ # ============================================================================
4
+ # These four functions are the single source of truth for turning a point into
5
+ # an isochrone, scoring it, and drawing its BAI spider plot. Both the main
6
+ # "Isochrone Explorer" tab and the "Isochrone Comparer" tab call them, so the
7
+ # scoring logic lives in exactly one place.
8
+ #
9
+ # build_isochrones() point + modes/times -> isochrone sf (mode,time,geom)
10
+ # compute_iso_metrics() isochrone sf + point -> per-isochrone metric data.frame
11
+ # draw_radar() BAI df -> spider/radar plot (base graphics)
12
+ #
13
+ # These functions read the *static* objects loaded by Rscripts/setup_unified.R
14
+ # directly as globals (cbg_vect_sf, osm_greenspace, the distance/NDVI rasters,
15
+ # rsf_projects, cbg_greenspace_coverage, gtfs_stops_sf, gtfs_routes_sf,
16
+ # gtfs_router, transit_iso_cache, cenv_sf, sf_ej_sf) plus the helpers/config
17
+ # defined at the top of app.R (mapbox_token, mode_palette, pretty_mode, ecdf01,
18
+ # standardize_iso_sf, build_walk_transit_isochrone, safe_biodiv_hotspots/coldspots).
19
+ #
20
+ # Two objects are created per-session inside server() and so are passed in as
21
+ # explicit arguments rather than read as globals:
22
+ # gbif_tab -- the session's DuckDB handle on the GBIF parquet
23
+ #
24
+ # compute_iso_metrics() reports progress via withProgress(), so it must be
25
+ # called from within a Shiny reactive context (which both tabs satisfy).
26
+ # ============================================================================
27
+
28
+
29
+ # Shared scientific definition used by both interactive and Step-5 calculations.
30
+ if (!exists("ISO_METRIC_DEFINITION_VERSION", inherits = TRUE)) {
31
+ source("Rscripts/iso_metric_definitions.R", local = TRUE)
32
+ }
33
+
34
+ # ----------------------------------------------------------------------------
35
+ # 100 m grid-supported metric cache
36
+ # ----------------------------------------------------------------------------
37
+ # Every scored raw metric uses the same support as corrected Step 5: the set of
38
+ # SF 100 m grid cells whose projected centroids fall within ONE isochrone.
39
+ # This makes user and reference metrics mathematically identical at 100 m support.
40
+ .iso_metric_cache <- new.env(parent = emptyenv())
41
+
42
+ load_iso_metric_grid <- function() {
43
+ if (exists("grid", envir = .iso_metric_cache, inherits = FALSE)) {
44
+ return(get("grid", envir = .iso_metric_cache, inherits = FALSE))
45
+ }
46
+
47
+ path <- Sys.getenv(
48
+ "ISO_METRIC_GRID_GPKG",
49
+ unset = file.path("outputs", "results", "grid_results_100m.gpkg")
50
+ )
51
+ if (!file.exists(path)) {
52
+ stop("Harmonized isochrone metrics require: ", path)
53
+ }
54
+
55
+ g <- sf::st_read(path, quiet = TRUE) |> sf::st_make_valid()
56
+ required <- c(
57
+ "cell_id", "n_records", "mean_ndvi", "gs_area_m2", "n_stops",
58
+ "nearest_stop_m", "calenviro_ci_score", "calenviro_traffic_pct", "sf_ej_score"
59
+ )
60
+ missing <- setdiff(required, names(g))
61
+ if (length(missing) > 0) {
62
+ stop("grid_results_100m.gpkg lacks required metric columns: ", paste(missing, collapse = ", "))
63
+ }
64
+
65
+ gp <- sf::st_transform(g, ISO_METRIC_CRS)
66
+ gp$cell_area_m2 <- as.numeric(sf::st_area(gp))
67
+ cp <- sf::st_centroid(gp) |> dplyr::select(cell_id)
68
+
69
+ out <- list(
70
+ grid_metric = gp,
71
+ centroids_metric = cp,
72
+ attrs = gp |>
73
+ sf::st_drop_geometry() |>
74
+ dplyr::transmute(
75
+ cell_id, cell_area_m2,
76
+ n_records = tidyr::replace_na(as.numeric(n_records), 0),
77
+ mean_ndvi = as.numeric(mean_ndvi),
78
+ gs_area_m2 = tidyr::replace_na(as.numeric(gs_area_m2), 0),
79
+ n_stops = tidyr::replace_na(as.numeric(n_stops), 0),
80
+ nearest_stop_m = as.numeric(nearest_stop_m),
81
+ calenviro_ci_score = as.numeric(calenviro_ci_score),
82
+ calenviro_traffic_pct = as.numeric(calenviro_traffic_pct),
83
+ sf_ej_score = as.numeric(sf_ej_score)
84
+ )
85
+ )
86
+ assign("grid", out, envir = .iso_metric_cache)
87
+ out
88
+ }
89
+
90
+ load_iso_species_cell_lookup <- function() {
91
+ if (exists("species", envir = .iso_metric_cache, inherits = FALSE)) {
92
+ return(get("species", envir = .iso_metric_cache, inherits = FALSE))
93
+ }
94
+ path <- file.path("outputs", "results", "gbif_cell_species_100m.rds")
95
+ if (!file.exists(path)) {
96
+ warning(
97
+ "Missing ", path, ". Run corrected 03_compute_isochrone_basi.R before ",
98
+ "using definition-matched biodiversity percentiles."
99
+ )
100
+ out <- NULL
101
+ } else {
102
+ out <- readRDS(path)
103
+ ver <- attr(out, "iso_metric_definition_version")
104
+ if (is.null(ver) || !identical(ver, ISO_METRIC_DEFINITION_VERSION)) {
105
+ stop("GBIF cell lookup definition version does not match the Shiny metric definition.")
106
+ }
107
+ }
108
+ assign("species", out, envir = .iso_metric_cache)
109
+ out
110
+ }
111
+
112
+ load_iso_route_cell_lookup <- function() {
113
+ if (exists("routes", envir = .iso_metric_cache, inherits = FALSE)) {
114
+ return(get("routes", envir = .iso_metric_cache, inherits = FALSE))
115
+ }
116
+ path <- file.path("outputs", "results", "muni_route_cells_100m.rds")
117
+ if (!file.exists(path)) {
118
+ warning(
119
+ "Missing ", path, ". Run corrected 03_compute_isochrone_basi.R before ",
120
+ "using definition-matched route-access percentiles."
121
+ )
122
+ out <- NULL
123
+ } else {
124
+ out <- readRDS(path)
125
+ ver <- attr(out, "iso_metric_definition_version")
126
+ if (is.null(ver) || !identical(ver, ISO_METRIC_DEFINITION_VERSION)) {
127
+ stop("Muni route-cell lookup definition version does not match the Shiny metric definition.")
128
+ }
129
+ }
130
+ assign("routes", out, envir = .iso_metric_cache)
131
+ out
132
+ }
133
+
134
+ iso_metric_support <- function(poly_i, include_geometry = FALSE) {
135
+ cache <- load_iso_metric_grid()
136
+ poly_metric <- sf::st_transform(sf::st_make_valid(poly_i), ISO_METRIC_CRS)
137
+ hit <- lengths(sf::st_within(cache$centroids_metric, sf::st_geometry(poly_metric))) > 0
138
+ ids <- cache$centroids_metric$cell_id[hit]
139
+ attrs <- cache$attrs[cache$attrs$cell_id %in% ids, , drop = FALSE]
140
+ if (!include_geometry) return(attrs)
141
+ geom <- cache$grid_metric[cache$grid_metric$cell_id %in% ids, , drop = FALSE]
142
+ list(attrs = attrs, geometry = geom)
143
+ }
144
+
145
+ iso_metric_support_geometry <- function(poly_i) {
146
+ support <- iso_metric_support(poly_i, include_geometry = TRUE)
147
+ if (is.null(support$geometry) || nrow(support$geometry) == 0) {
148
+ return(sf::st_sfc(crs = ISO_METRIC_CRS))
149
+ }
150
+ sf::st_union(sf::st_geometry(support$geometry))
151
+ }
152
+
153
+ summarise_iso_metric_support <- function(poly_i) {
154
+ support <- iso_metric_support(poly_i, include_geometry = FALSE)
155
+ if (is.null(support) || nrow(support) == 0) {
156
+ return(list(
157
+ cell_ids = integer(0), n_cells = 0L, area_m2 = NA_real_, area_km2 = NA_real_,
158
+ n_records = NA_real_, n_species = NA_real_, n_birds = NA_real_,
159
+ n_mammals = NA_real_, n_plants = NA_real_, sampling_density = NA_real_,
160
+ mean_ndvi = NA_real_, greenspace_pct = NA_real_, n_stops = NA_real_,
161
+ transit_access = NA_real_, unique_routes = NA_real_, nearest_stop_m = NA_real_,
162
+ calenviro_ci = NA_real_, traffic_pct = NA_real_, sf_ej = NA_real_
163
+ ))
164
+ }
165
+
166
+ ids <- unique(support$cell_id)
167
+ area_m2 <- sum(support$cell_area_m2[is.finite(support$cell_area_m2)], na.rm = TRUE)
168
+ area_km2 <- area_m2 / 1e6
169
+ n_records <- sum(tidyr::replace_na(support$n_records, 0), na.rm = TRUE)
170
+ n_stops <- sum(tidyr::replace_na(support$n_stops, 0), na.rm = TRUE)
171
+
172
+ spp <- load_iso_species_cell_lookup()
173
+ if (is.null(spp)) {
174
+ n_species <- n_birds <- n_mammals <- n_plants <- NA_real_
175
+ } else {
176
+ s <- spp[spp$cell_id %in% ids, , drop = FALSE]
177
+ n_species <- dplyr::n_distinct(s$species, na.rm = TRUE)
178
+ n_birds <- dplyr::n_distinct(s$species[s$class == "Aves"], na.rm = TRUE)
179
+ n_mammals <- dplyr::n_distinct(s$species[s$class == "Mammalia"], na.rm = TRUE)
180
+ n_plants <- dplyr::n_distinct(s$species[s$class %in% ISO_PLANT_CLASSES], na.rm = TRUE)
181
+ }
182
+
183
+ route_lookup <- load_iso_route_cell_lookup()
184
+ unique_routes <- if (is.null(route_lookup)) {
185
+ NA_real_
186
+ } else {
187
+ dplyr::n_distinct(route_lookup$route_id[route_lookup$cell_id %in% ids], na.rm = TRUE)
188
+ }
189
+
190
+ list(
191
+ cell_ids = ids,
192
+ n_cells = length(ids),
193
+ area_m2 = area_m2,
194
+ area_km2 = area_km2,
195
+ n_records = n_records,
196
+ n_species = n_species,
197
+ n_birds = n_birds,
198
+ n_mammals = n_mammals,
199
+ n_plants = n_plants,
200
+ sampling_density = if (is.finite(area_km2) && area_km2 > 0) n_records / area_km2 else NA_real_,
201
+ mean_ndvi = iso_weighted_mean(support$mean_ndvi, support$cell_area_m2),
202
+ greenspace_pct = if (is.finite(area_m2) && area_m2 > 0) {
203
+ pmin(100, 100 * sum(tidyr::replace_na(support$gs_area_m2, 0), na.rm = TRUE) / area_m2)
204
+ } else NA_real_,
205
+ n_stops = n_stops,
206
+ transit_access = if (is.finite(area_km2) && area_km2 > 0) n_stops / area_km2 else NA_real_,
207
+ unique_routes = unique_routes,
208
+ nearest_stop_m = iso_safe_min(support$nearest_stop_m),
209
+ calenviro_ci = iso_weighted_mean(support$calenviro_ci_score, support$cell_area_m2),
210
+ traffic_pct = iso_weighted_mean(support$calenviro_traffic_pct, support$cell_area_m2),
211
+ sf_ej = iso_weighted_mean(support$sf_ej_score, support$cell_area_m2)
212
+ )
213
+ }
214
+
215
+ # ----------------------------------------------------------------------------
216
+ # build_isochrones(): point + chosen modes/times -> combined isochrone sf
217
+ # ----------------------------------------------------------------------------
218
+ # point: named numeric c(lon = , lat = ) (the shape chosen_point() stores), or NULL.
219
+ # modes: character vector from c("driving","walking","cycling","driving-traffic",
220
+ # "transit","walk_transit"). times: numeric minutes. The transit_* / walk_*
221
+ # arguments only matter when a transit mode is selected; they default to the
222
+ # main tab's defaults so the comparer can pass a single mode/time and ignore them.
223
+ # Returns an sf with columns mode, time, geometry (EPSG:4326), or NULL if nothing
224
+ # could be built.
225
+ build_isochrones <- function(point, modes, times,
226
+ transit_hour = 9,
227
+ walk_to_stop_min = 5,
228
+ walk_from_stop_min = 5,
229
+ transit_departure_window_min = 10) {
230
+ if (is.null(point) || length(modes) == 0 || length(times) == 0) return(NULL)
231
+
232
+ location_sf <- st_as_sf(
233
+ data.frame(lon = point["lon"], lat = point["lat"]),
234
+ coords = c("lon", "lat"),
235
+ crs = 4326
236
+ )
237
+
238
+ iso_list <- list()
239
+ times <- as.numeric(times)
240
+
241
+ # --- Mapbox modes (driving / walking / cycling / driving-traffic) ----------
242
+ mapbox_modes <- intersect(modes, c("driving", "walking", "cycling", "driving-traffic"))
243
+ for (mode in mapbox_modes) {
244
+ for (t in times) {
245
+ iso <- tryCatch(
246
+ mb_isochrone(location_sf, time = t, profile = mode, access_token = mapbox_token),
247
+ # Surface the Mapbox error to the log instead of silently dropping it --
248
+ # otherwise a failed call just looks like "no isochrone" with no clue why.
249
+ error = function(e) { warning("mb_isochrone failed (", mode, " ", t, " min): ", conditionMessage(e)); NULL }
250
+ )
251
+ if (!is.null(iso)) {
252
+ iso_std <- standardize_iso_sf(iso, mode_name = mode, time_min = t)
253
+ if (!is.null(iso_std)) iso_list <- append(iso_list, list(iso_std))
254
+ }
255
+ }
256
+ }
257
+
258
+ # --- Transit (GTFS) --------------------------------------------------------
259
+ if ("transit" %in% modes && !is.null(gtfs_router) && !is.null(gtfs_stops_sf)) {
260
+ stop_dists <- st_distance(location_sf, gtfs_stops_sf)
261
+ nearest_idx <- which.min(stop_dists)
262
+ nearest_id <- as.character(gtfs_stops_sf$stop_id[nearest_idx])
263
+ dep_secs <- as.numeric(transit_hour) * 3600
264
+
265
+ for (t in times) {
266
+ iso_poly <- NULL
267
+
268
+ if (!is.null(transit_iso_cache) &&
269
+ !is.null(transit_iso_cache[[nearest_id]]) &&
270
+ !is.null(transit_iso_cache[[nearest_id]][[as.character(t)]])) {
271
+ iso_poly <- transit_iso_cache[[nearest_id]][[as.character(t)]]
272
+ }
273
+
274
+ if (is.null(iso_poly)) {
275
+ iso_result <- tryCatch(
276
+ gtfsrouter::gtfs_isochrone(
277
+ gtfs = gtfs_router,
278
+ from = nearest_id,
279
+ start_time = dep_secs,
280
+ end_time = dep_secs + t * 60,
281
+ from_is_id = TRUE
282
+ ),
283
+ error = function(e) NULL
284
+ )
285
+
286
+ if (!is.null(iso_result) && nrow(iso_result) > 2) {
287
+ reachable_sf <- gtfs_stops_sf |>
288
+ filter(stop_id %in% as.character(iso_result$stop_id))
289
+
290
+ if (nrow(reachable_sf) > 2) {
291
+ iso_poly <- st_convex_hull(st_union(reachable_sf))
292
+ } else if (nrow(reachable_sf) > 0) {
293
+ iso_poly <- st_union(st_buffer(st_transform(reachable_sf, 3857), 100)) |>
294
+ st_transform(4326)
295
+ }
296
+ }
297
+ }
298
+
299
+ if (!is.null(iso_poly)) {
300
+ iso_sf <- st_sf(
301
+ mode = "transit",
302
+ time = as.numeric(t),
303
+ geometry = st_geometry(st_as_sf(iso_poly)),
304
+ crs = 4326
305
+ )
306
+ iso_sf <- standardize_iso_sf(iso_sf, mode_name = "transit", time_min = t)
307
+ iso_list <- append(iso_list, list(iso_sf))
308
+ }
309
+ }
310
+ }
311
+
312
+ # --- Walk + Transit (Muni) -------------------------------------------------
313
+ if ("walk_transit" %in% modes && !is.null(gtfs_router) && !is.null(gtfs_stops_sf)) {
314
+ dep_secs <- as.numeric(transit_hour) * 3600
315
+ valid_times <- times[times > walk_to_stop_min]
316
+
317
+ for (t in valid_times) {
318
+ wt_iso <- tryCatch(
319
+ build_walk_transit_isochrone(
320
+ location_sf = location_sf,
321
+ total_time_min = t,
322
+ dep_secs = dep_secs,
323
+ walk_to_stop_min = walk_to_stop_min,
324
+ walk_from_stop_min = walk_from_stop_min,
325
+ gtfs_stops_sf = gtfs_stops_sf,
326
+ gtfs_router = gtfs_router,
327
+ mapbox_token = mapbox_token,
328
+ departure_window_min = transit_departure_window_min,
329
+ departure_step_min = 5,
330
+ max_last_mile_stops = 12,
331
+ include_first_mile_polygon = TRUE
332
+ ),
333
+ error = function(e) NULL
334
+ )
335
+
336
+ if (!is.null(wt_iso) && nrow(wt_iso) > 0) {
337
+ iso_list <- append(iso_list, list(wt_iso))
338
+ }
339
+ }
340
+ }
341
+
342
+ if (length(iso_list) == 0) return(NULL)
343
+
344
+ dplyr::bind_rows(iso_list) |>
345
+ st_as_sf() |>
346
+ st_make_valid() |>
347
+ st_transform(4326)
348
+ }
349
+
350
+
351
+ # ----------------------------------------------------------------------------
352
+ # compute_iso_metrics(): one independently calculated row per isochrone
353
+ # ----------------------------------------------------------------------------
354
+ # Scored metrics use the shared 100 m centroid-support definition above. This is
355
+ # intentionally identical to corrected Step 5. `gbif_tab` is retained in the
356
+ # signature for compatibility but is no longer used for scored richness/effort.
357
+ compute_iso_metrics <- function(iso_data, point, gbif_tab) {
358
+ if (is.null(iso_data) || nrow(iso_data) == 0) return(data.frame())
359
+
360
+ hotspot_union <- safe_biodiv_hotspots()
361
+ coldspot_union <- safe_biodiv_coldspots()
362
+ if (!is.null(hotspot_union)) hotspot_union <- st_union(hotspot_union)
363
+ if (!is.null(coldspot_union)) coldspot_union <- st_union(coldspot_union)
364
+
365
+ acs_wide <- cbg_vect_sf |>
366
+ mutate(population = popE, med_income = medincE)
367
+
368
+ user_point_sf <- NULL
369
+ if (!is.null(point)) {
370
+ user_point_sf <- st_as_sf(
371
+ data.frame(lon = point["lon"], lat = point["lat"]),
372
+ coords = c("lon", "lat"), crs = 4326
373
+ )
374
+ }
375
+
376
+ min_dist_val_global <- NA_real_
377
+ osm_greenspace_name_global <- NA_character_
378
+ if (!is.null(user_point_sf) && exists("greenspace_dist_raster") && exists("greenspace_osmid_raster")) {
379
+ try({
380
+ min_dist_val_global <- (greenspace_dist_raster |> extract(vect(user_point_sf)) |> pull(2))[1]
381
+ user_point_osm_id <- (greenspace_osmid_raster |> extract(vect(user_point_sf)) |> pull(2))[1]
382
+ osm_greenspace_name_global <- osm_greenspace |>
383
+ mutate(osm_id = as.numeric(osm_id)) |>
384
+ filter(osm_id == user_point_osm_id) |>
385
+ pull(name)
386
+ if (length(osm_greenspace_name_global) == 0 || is.na(osm_greenspace_name_global[1])) {
387
+ osm_greenspace_name_global <- "Unnamed Greenspace"
388
+ } else {
389
+ osm_greenspace_name_global <- osm_greenspace_name_global[1]
390
+ }
391
+ }, silent = TRUE)
392
+ }
393
+
394
+ min_rsf_dist_global <- NA_real_
395
+ rsf_program_name_global <- NA_character_
396
+ if (!is.null(user_point_sf) && exists("rsfprogram_dist_raster") && exists("rsfprogram_id_raster") && exists("rsf_projects")) {
397
+ try({
398
+ min_rsf_dist_global <- (rsfprogram_dist_raster |> extract(vect(user_point_sf)) |> pull(2))[1]
399
+ user_point_rsf_pid <- (rsfprogram_id_raster |> extract(vect(user_point_sf)) |> pull(2))[1]
400
+ rsf_program_name_global <- rsf_projects |>
401
+ dplyr::filter(as.numeric(.data$polygon_id) == as.numeric(user_point_rsf_pid)) |>
402
+ dplyr::pull(.data$prj_name)
403
+ if (length(rsf_program_name_global) == 0 || is.na(rsf_program_name_global[1])) {
404
+ rsf_program_name_global <- "Unknown RSF program"
405
+ } else {
406
+ rsf_program_name_global <- rsf_program_name_global[1]
407
+ }
408
+ }, silent = TRUE)
409
+ }
410
+
411
+ results <- data.frame()
412
+ n_isos <- nrow(iso_data)
413
+
414
+ withProgress(message = "Analyzing isochrones...", value = 0, {
415
+ for (i in seq_len(n_isos)) {
416
+ poly_i <- iso_data[i, ]
417
+ incProgress(
418
+ 1 / n_isos,
419
+ detail = paste0(pretty_mode(as.character(poly_i$mode[[1]])), " – ", poly_i$time[[1]], " min")
420
+ )
421
+
422
+ support <- summarise_iso_metric_support(poly_i)
423
+ iso_area_m2 <- as.numeric(st_area(st_transform(poly_i, ISO_METRIC_CRS)))
424
+ iso_area_km2 <- iso_area_m2 / 1e6
425
+
426
+ dist_hot_km <- if (!is.null(hotspot_union)) {
427
+ round(as.numeric(min(st_distance(poly_i, hotspot_union))) / 1000, 3)
428
+ } else NA_real_
429
+ dist_cold_km <- if (!is.null(coldspot_union)) {
430
+ round(as.numeric(min(st_distance(poly_i, coldspot_union))) / 1000, 3)
431
+ } else NA_real_
432
+
433
+ # Population/income remain descriptive, non-BAI fields and therefore retain
434
+ # the existing polygon-intersection implementation.
435
+ inter_acs <- tryCatch(intersect(vect(acs_wide), vect(poly_i)) |> st_as_sf(), error = function(e) NULL)
436
+ pop_total <- 0
437
+ w_income <- NA_real_
438
+ if (!is.null(inter_acs) && nrow(inter_acs) > 0) {
439
+ inter_acs <- inter_acs |>
440
+ mutate(
441
+ area_num = as.numeric(st_area(st_transform(geometry, ISO_METRIC_CRS))),
442
+ weighted_pop = population * (area_num / sum(area_num, na.rm = TRUE))
443
+ )
444
+ pop_total <- round(sum(inter_acs$weighted_pop, na.rm = TRUE))
445
+ w_income <- sum(inter_acs$med_income * inter_acs$area_num, na.rm = TRUE) /
446
+ sum(inter_acs$area_num, na.rm = TRUE)
447
+ }
448
+
449
+ # Frequency/headway are descriptive only; they are not reference-scored.
450
+ freq_weighted_score <- NA_real_
451
+ mean_headway_iso <- NA_real_
452
+ nearest_stop_name <- NA_character_
453
+ if (!is.null(gtfs_stops_sf)) {
454
+ inter_transit <- tryCatch(st_intersection(gtfs_stops_sf, poly_i), error = function(e) NULL)
455
+ if (!is.null(inter_transit) && nrow(inter_transit) > 0 &&
456
+ "mean_headway_min" %in% names(inter_transit)) {
457
+ hw <- inter_transit$mean_headway_min
458
+ hw <- hw[is.finite(hw) & hw > 0]
459
+ if (length(hw) > 0 && is.finite(support$area_km2) && support$area_km2 > 0) {
460
+ freq_weighted_score <- round(sum(60 / hw) / support$area_km2, 2)
461
+ mean_headway_iso <- round(mean(hw), 1)
462
+ }
463
+ }
464
+ if (is.finite(support$nearest_stop_m)) {
465
+ support_geom <- iso_metric_support_geometry(poly_i)
466
+ if (length(support_geom) > 0) {
467
+ support_sf <- st_sf(geometry = support_geom)
468
+ d <- tryCatch(st_distance(support_sf, st_transform(gtfs_stops_sf, st_crs(support_sf))), error = function(e) NULL)
469
+ if (!is.null(d) && length(d) > 0) {
470
+ nearest_stop_name <- gtfs_stops_sf$stop_name[which.min(as.numeric(d))]
471
+ }
472
+ }
473
+ }
474
+ }
475
+
476
+ row_i <- data.frame(
477
+ Mode = pretty_mode(as.character(poly_i$mode[[1]])),
478
+ Time = as.numeric(poly_i$time[[1]]),
479
+ IsochroneArea_km2 = round(iso_area_km2, 3),
480
+ MetricSupportArea_km2 = round(support$area_km2, 3),
481
+ MetricSupportNCells = as.integer(support$n_cells),
482
+ MetricDefinitionVersion = ISO_METRIC_DEFINITION_VERSION,
483
+ DistToHotspot_km = dist_hot_km,
484
+ DistToColdspot_km = dist_cold_km,
485
+ EstimatedPopulation = pop_total,
486
+ MedianIncome = round(w_income, 2),
487
+ MeanNDVI = as.numeric(support$mean_ndvi),
488
+ GBIF_Records = as.numeric(support$n_records),
489
+ GBIF_Species = as.numeric(support$n_species),
490
+ Bird_Species = as.numeric(support$n_birds),
491
+ Mammal_Species = as.numeric(support$n_mammals),
492
+ Plant_Species = as.numeric(support$n_plants),
493
+ SamplingDensity_km2 = as.numeric(support$sampling_density),
494
+ Greenspace_percent = as.numeric(support$greenspace_pct),
495
+ Transit_Stops = as.numeric(support$n_stops),
496
+ Unique_Muni_Routes = as.numeric(support$unique_routes),
497
+ Transit_Access_Score = as.numeric(support$transit_access),
498
+ Freq_Weighted_Score = freq_weighted_score,
499
+ Mean_Headway_min = mean_headway_iso,
500
+ Nearest_Stop_m = as.numeric(support$nearest_stop_m),
501
+ Nearest_Stop_Name = nearest_stop_name,
502
+ CalEnviro_CIscore = as.numeric(support$calenviro_ci),
503
+ CalEnviro_Traffic_Pctl = as.numeric(support$traffic_pct),
504
+ SF_EJ_Score = as.numeric(support$sf_ej),
505
+ closest_greenspace = osm_greenspace_name_global,
506
+ closest_greenspace_dist_m = min_dist_val_global,
507
+ closest_rsf_program = rsf_program_name_global,
508
+ closest_rsf_program_dist_m = min_rsf_dist_global,
509
+ stringsAsFactors = FALSE
510
+ )
511
+ results <- rbind(results, row_i)
512
+ }
513
+ })
514
+
515
+ # No union-based or mean-across-isochrone scoring attributes are created here.
516
+ # Each row remains the sole unit of percentile and BAI scoring.
517
+ if (nrow(results) > 0) {
518
+ closest_gs <- results |>
519
+ filter(!is.na(closest_greenspace_dist_m)) |>
520
+ slice_min(closest_greenspace_dist_m, n = 1)
521
+ attr(results, "closest_greenspace") <- if (nrow(closest_gs) > 0) closest_gs$closest_greenspace[1] else "None"
522
+ attr(results, "closest_greenspace_dist_m") <- if (nrow(closest_gs) > 0) closest_gs$closest_greenspace_dist_m[1] else NA_real_
523
+
524
+ closest_rsf <- results |>
525
+ filter(!is.na(closest_rsf_program_dist_m)) |>
526
+ slice_min(closest_rsf_program_dist_m, n = 1)
527
+ attr(results, "closest_rsf_program") <- if (nrow(closest_rsf) > 0) closest_rsf$closest_rsf_program[1] else "None"
528
+ attr(results, "closest_rsf_program_dist_m") <- if (nrow(closest_rsf) > 0) closest_rsf$closest_rsf_program_dist_m[1] else NA_real_
529
+ }
530
+
531
+ results
532
+ }
533
+
534
+
535
+ # Legacy census-block-group add_bai() scoring was removed. The app now uses
536
+ # add_reference_bai() from app.R, which requires same-mode × same-time
537
+ # precomputed isochrone references and never falls back to CBG ECDFs.
538
+
539
+
540
+ # Axis order shared by both radar functions, and the column each axis reads from.
541
+ # Keep these two vectors aligned -- the group labels below assume this clockwise
542
+ # order (1-2 Urban Access, 3-4 Biodiversity, 5-6 Environment, 7 EJ).
543
+ RADAR_AXIS_COLS <- c("Mobility_Access_std", "Route_Access_std", "Biodiversity_Potential_std",
544
+ "Observation_Intensity_std", "Environmental_Quality_std",
545
+ "Greenspace_Cover_std", "Equity_Context_std")
546
+ RADAR_AXIS_LABELS <- c("Stop\nDensity", "Route\nDiversity", "Species\nRichness",
547
+ "Obs.\nIntensity", "Vegetation\n(NDVI)", "Greenspace\nCover", "EJ\nContext")
548
+
549
+ # Radial group subheaders placed just beyond the axis labels. Shared by both
550
+ # radar functions so the grouping (Urban Access / Biodiversity / Environment /
551
+ # EJ) is drawn identically. Call after fmsb::radarchart() has drawn the chart.
552
+ draw_radar_group_labels <- function() {
553
+ n_ax <- 7
554
+ angs <- pi/2 - (0:(n_ax - 1)) * (2 * pi / n_ax)
555
+ cat_r <- 1.48 # radius just outside axis labels (~1.2-1.3)
556
+
557
+ text(cat_r * cos(mean(angs[1:2])), cat_r * sin(mean(angs[1:2])),
558
+ "Urban Access", cex = 0.72, col = "#2166ac", font = 2, xpd = TRUE)
559
+ text(cat_r * cos(mean(angs[3:4])), cat_r * sin(mean(angs[3:4])),
560
+ "Biodiversity", cex = 0.72, col = "#1b7837", font = 2, xpd = TRUE)
561
+ text(cat_r * cos(mean(angs[5:6])), cat_r * sin(mean(angs[5:6])),
562
+ "Environment", cex = 0.72, col = "#762a83", font = 2, xpd = TRUE)
563
+ text(cat_r * cos(angs[7]), cat_r * sin(angs[7]),
564
+ "Environmental\nJustice", cex = 0.72, col = "#b2182b", font = 2, xpd = TRUE)
565
+ }
566
+
567
+ # ----------------------------------------------------------------------------
568
+ # draw_radar(): BAI df -> spider/radar plot (base graphics, via fmsb)
569
+ # ----------------------------------------------------------------------------
570
+ # bai_df: the data.frame returned by same-mode × same-time reference scoring (one row per isochrone). title:
571
+ # the chart title. Draws directly to the current graphics device, so call it
572
+ # from inside a renderPlot({ }). Lines are coloured by transport mode.
573
+ draw_radar <- function(bai_df, title = "Biodiversity Access Index Profile") {
574
+ if (is.null(bai_df) || nrow(bai_df) == 0) return(NULL)
575
+
576
+ radar_df <- bai_df |>
577
+ mutate(
578
+ ModeTime = paste0(Mode, "_", Time, "m"),
579
+ `Stop\nDensity` = Mobility_Access_std,
580
+ `Route\nDiversity` = Route_Access_std,
581
+ `Species\nRichness` = Biodiversity_Potential_std,
582
+ `Obs.\nIntensity` = Observation_Intensity_std,
583
+ `Vegetation\n(NDVI)`= Environmental_Quality_std,
584
+ `Greenspace\nCover` = Greenspace_Cover_std,
585
+ `EJ\nContext` = Equity_Context_std
586
+ ) |>
587
+ select(
588
+ ModeTime,
589
+ `Stop\nDensity`,
590
+ `Route\nDiversity`,
591
+ `Species\nRichness`,
592
+ `Obs.\nIntensity`,
593
+ `Vegetation\n(NDVI)`,
594
+ `Greenspace\nCover`,
595
+ `EJ\nContext`
596
+ )
597
+
598
+ radar_mat <- as.data.frame(radar_df[, -1])
599
+ rownames(radar_mat) <- radar_df$ModeTime
600
+
601
+ radar_mat <- rbind(
602
+ rep(1, ncol(radar_mat)),
603
+ rep(0, ncol(radar_mat)),
604
+ radar_mat
605
+ )
606
+
607
+ labels <- rownames(radar_mat)[-(1:2)]
608
+ line_cols <- mode_palette[gsub("_(.*)$", "", labels)]
609
+ line_cols[is.na(line_cols)] <- "#666666"
610
+
611
+ # Extra margin so two-line axis labels aren't clipped
612
+ par(mar = c(2, 2, 3, 2))
613
+
614
+ fmsb::radarchart(
615
+ radar_mat,
616
+ axistype = 1,
617
+ pcol = line_cols,
618
+ plwd = 2,
619
+ plty = 1,
620
+ cglcol = "grey80",
621
+ cglty = 1,
622
+ cglwd = 0.8,
623
+ axislabcol = "grey40",
624
+ vlcex = 0.88,
625
+ title = title
626
+ )
627
+
628
+ draw_radar_group_labels()
629
+
630
+ legend(
631
+ "topright",
632
+ legend = labels,
633
+ col = line_cols,
634
+ lty = 1,
635
+ lwd = 2,
636
+ cex = 0.75,
637
+ bty = "n"
638
+ )
639
+ }
640
+
641
+
642
+ # ----------------------------------------------------------------------------
643
+ # draw_compare_radar(): overlay several BAI rows on ONE radar (Comparer tab)
644
+ # ----------------------------------------------------------------------------
645
+ # bai_rows: a list of single-row reference-scored BAI data.frames (one per point).
646
+ # labels: legend text, one per row (e.g. "Point A: Walking 5 min").
647
+ # colors: one line colour per row -- this is where the A-vs-B colour scheme
648
+ # lives, so keep it in sync with the point markers / isochrones.
649
+ # Unlike draw_radar() (coloured by mode), this colours by *point* so two
650
+ # isochrones of the same mode are still distinguishable. Draws to the current
651
+ # device; call from inside renderPlot({ }).
652
+ draw_compare_radar <- function(bai_rows, labels, colors,
653
+ title = "Biodiversity Access Index — Point A vs Point B") {
654
+ if (length(bai_rows) == 0) return(NULL)
655
+
656
+ mat <- as.data.frame(do.call(rbind, lapply(bai_rows, function(d) as.numeric(d[1, RADAR_AXIS_COLS]))))
657
+ names(mat) <- RADAR_AXIS_LABELS
658
+ # fmsb wants the axis max (1) and min (0) as the first two rows.
659
+ mat <- rbind(rep(1, ncol(mat)), rep(0, ncol(mat)), mat)
660
+
661
+ par(mar = c(2, 2, 3, 2))
662
+ fmsb::radarchart(
663
+ mat,
664
+ axistype = 1,
665
+ pcol = colors,
666
+ pfcol = grDevices::adjustcolor(colors, alpha.f = 0.2), # translucent fills so overlaps read
667
+ plwd = 3,
668
+ plty = 1,
669
+ cglcol = "grey80",
670
+ cglty = 1,
671
+ cglwd = 0.8,
672
+ axislabcol = "grey40",
673
+ vlcex = 0.9,
674
+ title = title
675
+ )
676
+
677
+ draw_radar_group_labels()
678
+
679
+ legend(
680
+ "topright",
681
+ legend = labels,
682
+ col = colors,
683
+ lty = 1,
684
+ lwd = 3,
685
+ cex = 0.85,
686
+ bty = "n"
687
+ )
688
+ }
code/setup_unified.R ADDED
@@ -0,0 +1,287 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ============================================================================
2
+ # Setup: Libraries
3
+ # ============================================================================
4
+ # require(shinyjs)
5
+ library(shiny)
6
+ library(shinydashboard)
7
+ library(leaflet)
8
+ library(mapboxapi)
9
+ # library(tidyverse)
10
+ library(dplyr)
11
+ library(tidyr)
12
+ library(readr)
13
+ library(purrr)
14
+ library(tibble)
15
+ library(stringr)
16
+ library(forcats)
17
+ library(lubridate)
18
+ library(ggplot2)
19
+ # library(tidycensus)
20
+ library(sf)
21
+ library(DT)
22
+ library(RColorBrewer)
23
+ library(terra)
24
+ library(data.table)
25
+ # library(mapview)
26
+ # library(sjPlot)
27
+ # library(sjlabelled)
28
+ library(bslib)
29
+ library(shinycssloaders)
30
+ library(glue)
31
+
32
+ # ============================================================================
33
+ # Setup: HuggingFace base URL and cache directory
34
+ # ============================================================================
35
+ HF_BASE <- "https://huggingface.co/datasets/boettiger-lab/sf_biodiv_access/resolve/main"
36
+
37
+ # Use data/cached/ when running locally (writable), otherwise fall back to
38
+ # /tmp/sf_biodiv_cache/ for read-only environments like HuggingFace Spaces.
39
+ cache_dir <- if (file.access(".", mode = 2) == 0) "data/cached" else "/tmp/sf_biodiv_cache"
40
+ dir.create(cache_dir, recursive = TRUE, showWarnings = FALSE)
41
+
42
+ # Helper: if the file already exists in data/cached/, return that path.
43
+ # Otherwise attempt to download from HuggingFace into data/cached/.
44
+ # Returns the destination path regardless — caller must check file.exists() if
45
+ # the download may fail (e.g. file not yet uploaded to HF).
46
+ hf_or_local <- function(filename) {
47
+ dest <- file.path(cache_dir, filename)
48
+ if (!file.exists(dest)) {
49
+ tryCatch(
50
+ download.file(glue::glue("{HF_BASE}/{filename}"), dest, mode = "wb", quiet = TRUE),
51
+ error = function(e) warning(glue::glue("HuggingFace download failed for {filename}: {e$message}")),
52
+ warning = function(w) warning(glue::glue("HuggingFace download warning for {filename}: {w$message}"))
53
+ )
54
+ }
55
+ dest
56
+ }
57
+
58
+ message("[setup_unified] loading greenspace polygons (cache / HuggingFace)…")
59
+
60
+ # ============================================================================
61
+ # Load Data: Greenspace (OSM polygons)
62
+ # ============================================================================
63
+ # Shapefile bundle on HuggingFace — download sidecars into cache if needed.
64
+ greenspace_shp <- file.path(cache_dir, "greenspaces_osm_nad83.shp")
65
+ if (!file.exists(greenspace_shp)) {
66
+ for (ext in c("shp", "dbf", "prj", "shx")) {
67
+ hf_or_local(glue::glue("greenspaces_osm_nad83.{ext}"))
68
+ }
69
+ }
70
+ osm_greenspace <- st_read(greenspace_shp, quiet = TRUE) |> st_transform(4326)
71
+ if (!"name" %in% names(osm_greenspace)) osm_greenspace$name <- "Unnamed Greenspace"
72
+
73
+ message("[setup_unified] loading greenspace distance rasters + NDVI…")
74
+
75
+ # ============================================================================
76
+ # Load Data: Greenspace distance rasters
77
+ # ============================================================================
78
+ greenspace_dist_raster <- terra::rast(hf_or_local("nearest_greenspace_dist.tif"))
79
+ greenspace_osmid_raster <- terra::rast(hf_or_local("nearest_greenspace_osmid.tif"))
80
+
81
+ rsfprogram_dist_raster <- terra::rast(hf_or_local("nearest_rsfprogram_dist.tif"))
82
+ rsfprogram_id_raster <- terra::rast(hf_or_local("nearest_rsfprogram_id.tif"))
83
+
84
+ # ============================================================================
85
+ # Load Data: NDVI raster
86
+ # ============================================================================
87
+ ndvi <- terra::rast(hf_or_local("SF_EastBay_NDVI_Sentinel_10.tif"))
88
+
89
+ message("[setup_unified] loading GBIF parquet + CBG polygons…")
90
+
91
+ # ============================================================================
92
+ # Load Data: GBIF observations (parquet, queried via DuckDB in server)
93
+ # ============================================================================
94
+ gbif_parquet <- hf_or_local("gbif_census_ndvi_anno.parquet")
95
+
96
+ # ============================================================================
97
+ # Load Data: Census block groups (CBG)
98
+ # ============================================================================
99
+ load(hf_or_local("cbg_vect_sf.Rdata"))
100
+
101
+ if (!"unique_species" %in% names(cbg_vect_sf)) cbg_vect_sf$unique_species <- cbg_vect_sf$n_species
102
+ if (!"n_observations" %in% names(cbg_vect_sf)) cbg_vect_sf$n_observations <- cbg_vect_sf$n
103
+ if (!"median_inc" %in% names(cbg_vect_sf)) cbg_vect_sf$median_inc <- cbg_vect_sf$medincE
104
+ if (!"ndvi_mean" %in% names(cbg_vect_sf)) cbg_vect_sf$ndvi_mean <- cbg_vect_sf$ndvi_sentinel
105
+
106
+ message("[setup_unified] computing CBG × greenspace overlap (vector intersect)…")
107
+
108
+ # ============================================================================
109
+ # Per-CBG greenspace overlap (computed here; no separate CSV on HuggingFace)
110
+ # ============================================================================
111
+ cbg_proj <- st_transform(cbg_vect_sf[, "GEOID"], 3857) |>
112
+ mutate(cbg_area_m2 = as.numeric(st_area(geometry)))
113
+ gs_proj <- st_transform(osm_greenspace, 3857) |> st_make_valid()
114
+ gs_union <- st_union(gs_proj)
115
+ cbg_gs_inter <- st_intersection(cbg_proj, gs_union)
116
+ cbg_greenspace_coverage <- cbg_gs_inter |>
117
+ mutate(greenspace_m2 = as.numeric(st_area(geometry))) |>
118
+ st_drop_geometry() |>
119
+ group_by(GEOID) |>
120
+ summarise(greenspace_m2 = sum(greenspace_m2), .groups = "drop") |>
121
+ right_join(cbg_proj |> st_drop_geometry() |> dplyr::select(GEOID, cbg_area_m2), by = "GEOID") |>
122
+ mutate(
123
+ greenspace_m2 = tidyr::replace_na(greenspace_m2, 0),
124
+ GEOID = as.character(GEOID)
125
+ )
126
+
127
+ message("[setup_unified] loading biodiversity hotspots / coldspots…")
128
+
129
+ # ============================================================================
130
+ # Load Data: Biodiversity hotspots / coldspots
131
+ # ============================================================================
132
+ hotspots_shp <- file.path(cache_dir, "hotspots.shp")
133
+ if (!file.exists(hotspots_shp)) {
134
+ for (ext in c("shp", "dbf", "prj", "shx")) hf_or_local(glue::glue("hotspots.{ext}"))
135
+ }
136
+ biodiv_hotspots <- st_read(hotspots_shp, quiet = TRUE) |> st_transform(4326)
137
+
138
+ coldspots_shp <- file.path(cache_dir, "coldspots.shp")
139
+ if (!file.exists(coldspots_shp)) {
140
+ for (ext in c("shp", "dbf", "prj", "shx")) hf_or_local(glue::glue("coldspots.{ext}"))
141
+ }
142
+ biodiv_coldspots <- st_read(coldspots_shp, quiet = TRUE) |> st_transform(4326)
143
+
144
+ message("[setup_unified] loading RSF, CalEnviroScreen, SF EJ layers…")
145
+
146
+ # ============================================================================
147
+ # Load Data: RSF Program Projects
148
+ # ============================================================================
149
+ rsf_projects <- st_read(hf_or_local("RSF_Program_Projects_polygons.gpkg"), quiet = TRUE) |>
150
+ st_transform(4326)
151
+
152
+ # ============================================================================
153
+ # Load Data: CalEnviroScreen 4.0 (pre-filtered to SF)
154
+ # ============================================================================
155
+ cenv_sf <- tryCatch({
156
+ sf::st_read(hf_or_local("calenviro_sf.gpkg"), quiet = TRUE)
157
+ }, error = function(e) {
158
+ warning("CalEnviroScreen failed to load: ", e$message); NULL
159
+ })
160
+
161
+ # ============================================================================
162
+ # Load Data: SF Environmental Justice Communities
163
+ # ============================================================================
164
+ sf_ej_sf <- tryCatch({
165
+ sf::st_read(hf_or_local("sf_ej_communities_map.gpkg"), quiet = TRUE) |>
166
+ dplyr::mutate(
167
+ symbol_hex = stringr::str_split(symbol_rgb, ",\\s*") |>
168
+ lapply(function(x) sprintf("#%02X%02X%02X",
169
+ as.integer(x[1]), as.integer(x[2]), as.integer(x[3]))) |>
170
+ unlist(),
171
+ ej_label = dplyr::case_when(
172
+ is.na(score) ~ "Not EJ",
173
+ score >= 21 ~ "High EJ burden (21-30)",
174
+ score >= 11 ~ "Moderate EJ burden (11-20)",
175
+ score >= 1 ~ "Low EJ burden (1-10)",
176
+ score == 0 ~ "Score 0",
177
+ TRUE ~ "Unknown"
178
+ )
179
+ )
180
+ }, error = function(e) {
181
+ warning("SF EJ layer failed to load: ", e$message); NULL
182
+ })
183
+
184
+ message("[setup_unified] loading GTFS (zip, stops, shapes, timetable, headways)…")
185
+
186
+ # ============================================================================
187
+ # Load Data: GTFS (SF Muni)
188
+ # ============================================================================
189
+
190
+ gtfs_zip_path <- hf_or_local("sf_muni_gtfs.zip")
191
+
192
+ # Unzip for read.csv(stops.txt, …); tidytransit/gtfsrouter read the .zip (gtfsio needs a zip path)
193
+ gtfs_unzip_dir <- file.path(cache_dir, "muni_gtfs")
194
+ dir.create(gtfs_unzip_dir, recursive = TRUE, showWarnings = FALSE)
195
+ if (!dir.exists(gtfs_unzip_dir) || length(list.files(gtfs_unzip_dir, pattern = "\\.txt$")) == 0L) {
196
+ unzip(gtfs_zip_path, exdir = gtfs_unzip_dir, overwrite = TRUE)
197
+ }
198
+ gtfs_path <- gtfs_unzip_dir
199
+
200
+ # --- Transit stops -----------------------------------------------------------
201
+ gtfs_stops_sf <- tryCatch({
202
+ read.csv(file.path(gtfs_path, "stops.txt")) |>
203
+ st_as_sf(coords = c("stop_lon", "stop_lat"), crs = 4326)
204
+ }, error = function(e) { warning("GTFS stops failed to load: ", e$message); NULL })
205
+
206
+ # --- Route shapes ------------------------------------------------------------
207
+ gtfs_routes_sf <- tryCatch({
208
+ gtfs_shapes_raw <- read.csv(file.path(gtfs_path, "shapes.txt"))
209
+ gtfs_trips_raw <- read.csv(file.path(gtfs_path, "trips.txt"))
210
+ gtfs_routes_raw <- read.csv(file.path(gtfs_path, "routes.txt"))
211
+
212
+ shape_route_map <- gtfs_trips_raw |> distinct(shape_id, route_id)
213
+ route_meta <- gtfs_routes_raw |>
214
+ select(route_id, route_short_name, route_long_name, route_color) |>
215
+ mutate(route_color_hex = paste0("#", trimws(route_color)))
216
+
217
+ shapes_split <- gtfs_shapes_raw |>
218
+ arrange(shape_id, shape_pt_sequence) |>
219
+ group_by(shape_id) |>
220
+ group_split()
221
+
222
+ shape_geoms <- lapply(shapes_split, function(s) {
223
+ st_linestring(cbind(s$shape_pt_lon, s$shape_pt_lat))
224
+ })
225
+
226
+ st_sf(
227
+ shape_id = sapply(shapes_split, function(s) s$shape_id[1]),
228
+ geometry = st_sfc(shape_geoms, crs = 4326)
229
+ ) |>
230
+ left_join(shape_route_map, by = "shape_id") |>
231
+ left_join(route_meta, by = "route_id")
232
+ }, error = function(e) { warning("GTFS route shapes failed to load: ", e$message); NULL })
233
+
234
+ # --- gtfsrouter timetable ----------------------------------------------------
235
+ gtfs_router <- tryCatch({
236
+ timetable_path <- hf_or_local("gtfs_timetable_monday.rds")
237
+ if (file.exists(timetable_path)) {
238
+ readRDS(timetable_path)
239
+ } else {
240
+ gr <- gtfsrouter::extract_gtfs(gtfs_zip_path)
241
+ result <- gtfsrouter::gtfs_timetable(gr, day = "Monday")
242
+ saveRDS(result, file.path(cache_dir, "gtfs_timetable_monday.rds"))
243
+ result
244
+ }
245
+ }, error = function(e) { warning("gtfsrouter failed to initialise: ", e$message); NULL })
246
+
247
+ # --- Pre-computed transit isochrone cache ------------------------------------
248
+ transit_iso_cache <- tryCatch({
249
+ p <- file.path(cache_dir, "transit_iso_cache.rds")
250
+ if (file.exists(p)) readRDS(p) else NULL
251
+ }, error = function(e) { NULL })
252
+
253
+ # --- Stop headways (AM peak 7-9am): cached as CSV (readable / diffable) -------
254
+ # gtfsrouter timetable stays .rds (opaque R object); this table is just columns.
255
+ hw_csv <- file.path(cache_dir, "gtfs_stop_headways.csv")
256
+ hw_rds <- file.path(cache_dir, "gtfs_stop_headways.rds")
257
+ if (!file.exists(hw_csv) && file.exists(hw_rds)) {
258
+ readRDS(hw_rds) |> readr::write_csv(hw_csv)
259
+ }
260
+
261
+ gtfs_stop_headways <- tryCatch({
262
+ headways_path <- hf_or_local("gtfs_stop_headways.csv")
263
+ if (file.exists(headways_path)) {
264
+ readr::read_csv(headways_path, show_col_types = FALSE) |>
265
+ mutate(stop_id = as.character(stop_id))
266
+ } else {
267
+ gt <- tidytransit::read_gtfs(gtfs_zip_path)
268
+ hw <- tidytransit::get_stop_frequency(gt, start_time = 7 * 3600, end_time = 9 * 3600) |>
269
+ group_by(stop_id) |>
270
+ summarise(
271
+ mean_headway_min = mean(mean_headway, na.rm = TRUE) / 60,
272
+ n_departures_peak = sum(n_departures, na.rm = TRUE),
273
+ .groups = "drop"
274
+ ) |>
275
+ mutate(stop_id = as.character(stop_id))
276
+ readr::write_csv(hw, hw_csv)
277
+ hw
278
+ }
279
+ }, error = function(e) { warning("tidytransit headway computation failed: ", e$message); NULL })
280
+
281
+ if (!is.null(gtfs_stop_headways) && !is.null(gtfs_stops_sf)) {
282
+ gtfs_stops_sf <- gtfs_stops_sf |>
283
+ mutate(stop_id = as.character(stop_id)) |>
284
+ left_join(gtfs_stop_headways, by = "stop_id")
285
+ }
286
+
287
+ message("[setup_unified] data load complete.")
install.r ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ options(repos = c(CRAN = "https://cloud.r-project.org"))
2
+
3
+ install.packages(c(
4
+
5
+ # Shiny / UI
6
+ "shiny",
7
+ "shinydashboard",
8
+ "shinycssloaders",
9
+ "bslib",
10
+ "DT",
11
+ "leaflet",
12
+ "leaflet.extras",
13
+ "htmltools",
14
+ "base64enc",
15
+
16
+ # Core data wrangling / plotting
17
+ # Install components directly rather than tidyverse meta-package
18
+ "dplyr",
19
+ "tidyr",
20
+ "readr",
21
+ "purrr",
22
+ "tibble",
23
+ "stringr",
24
+ "forcats",
25
+ "lubridate",
26
+ "ggplot2",
27
+ "scales",
28
+
29
+ # Spatial
30
+ "sf",
31
+ "terra",
32
+ "mapboxapi",
33
+
34
+ # Database
35
+ "DBI",
36
+ "dbplyr",
37
+ "duckdb",
38
+
39
+ # Transit
40
+ "gtfsrouter",
41
+ "tidytransit",
42
+ "tidycensus",
43
+
44
+ # Radar plots
45
+ "fmsb",
46
+
47
+ # Helpers
48
+ "glue",
49
+ "jsonlite",
50
+ "data.table"
51
+ ))
outputs/results/gbif_cell_species_100m.rds ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:35649dd22b7093bda327fc71d17eb0edb163160ca8e55810fcfd2cc1956d6467
3
+ size 895847
outputs/results/grid_results_100m.gpkg ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77a11fe1a88ff9eb1fc862fb449c0ba740cd5e3b7bafdd0c5792eaa6ecc3cfa7
3
+ size 8544256
outputs/results/isochrone_basi_100m.gpkg ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd37dc968f140ab9f89fd50acb1a340b775bc468465376fb5a511514d838ed98
3
+ size 88784896
outputs/results/isochrone_basi_100m.rds ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:adb323cfca3133dd3c8bd2e518fd4fed03ac4123259457aacd9d6a7a350b033d
3
+ size 23175655
outputs/results/muni_route_cells_100m.rds ADDED
Binary file (61.9 kB). View file
 
rsconnect/shinyapps.io/diego-ellis-soto/RSF_Biodiversity_Access.dcf ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ name: RSF_Biodiversity_Access
2
+ title: SF_biodiv_access_shiny
3
+ username: diego-ellis-soto
4
+ account: diego-ellis-soto
5
+ server: shinyapps.io
6
+ hostUrl: https://api.shinyapps.io/v1
7
+ appId: 13693040
8
+ bundleId: 9634358
9
+ url: https://diego-ellis-soto.shinyapps.io/RSF_Biodiversity_Access/
10
+ version: 1
11
+ asMultiple: FALSE
12
+ asStatic: FALSE
rsconnect/shinyapps.io/diego-ellis-soto/SF_biodiv_access.dcf ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ name: SF_biodiv_access
2
+ title: SF_biodiv_access
3
+ username: diego-ellis-soto
4
+ account: diego-ellis-soto
5
+ server: shinyapps.io
6
+ hostUrl: https://api.shinyapps.io/v1
7
+ appId: 13693447
8
+ bundleId: 9631093
9
+ url: https://diego-ellis-soto.shinyapps.io/SF_biodiv_access/
10
+ version: 1
11
+ asMultiple: FALSE
12
+ asStatic: FALSE
13
+ ignoredFiles: app_old.R
www/California_academy_logo.png ADDED
www/Combined_logos.png ADDED

Git LFS Details

  • SHA256: e44786edd4cce3836207e6243306e34c8fadc1a39a2484d2cbda35dfcf2b0dfe
  • Pointer size: 131 Bytes
  • Size of remote file: 201 kB
www/Reimagining_San_Francisco.png ADDED

Git LFS Details

  • SHA256: 71b5ea34c87176c52e7a9d8341d7482a9801dea7492bd638b17c5b1c0a79913b
  • Pointer size: 131 Bytes
  • Size of remote file: 760 kB
www/UC_Berkeley_logo.png ADDED
www/app_pastel.css ADDED
@@ -0,0 +1,89 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ /* SF Biodiversity Access — pastel chrome (shinydashboard overrides) */
2
+
3
+ .content-wrapper,
4
+ .right-side {
5
+ background-color: #f3f8f5 !important;
6
+ }
7
+
8
+ .box {
9
+ border-radius: 12px;
10
+ overflow: hidden;
11
+ border: 1px solid rgba(61, 92, 74, 0.1);
12
+ box-shadow: 0 2px 14px rgba(45, 74, 58, 0.07);
13
+ margin-bottom: 20px;
14
+ }
15
+
16
+ .box .box-header {
17
+ font-weight: 600;
18
+ letter-spacing: 0.01em;
19
+ }
20
+
21
+ .box.box-success > .box-header {
22
+ background: #c8e9d6 !important;
23
+ color: #1a4534 !important;
24
+ border-bottom: 1px solid rgba(26, 69, 52, 0.12);
25
+ }
26
+
27
+ .box.box-primary > .box-header {
28
+ background: #c9dcf5 !important;
29
+ color: #1f3d66 !important;
30
+ border-bottom: 1px solid rgba(31, 61, 102, 0.12);
31
+ }
32
+
33
+ .box.box-warning > .box-header {
34
+ background: #fce9c9 !important;
35
+ color: #5a4518 !important;
36
+ border-bottom: 1px solid rgba(90, 69, 24, 0.12);
37
+ }
38
+
39
+ .box.box-info > .box-header {
40
+ background: #c9ecf3 !important;
41
+ color: #1a4a55 !important;
42
+ border-bottom: 1px solid rgba(26, 74, 85, 0.12);
43
+ }
44
+
45
+ .box.box-danger > .box-header {
46
+ background: #f5d4d4 !important;
47
+ color: #5c2222 !important;
48
+ border-bottom: 1px solid rgba(92, 34, 34, 0.12);
49
+ }
50
+
51
+ .box-body {
52
+ background-color: #fcfdfc;
53
+ }
54
+
55
+ .well {
56
+ background-color: #f7faf8 !important;
57
+ border: 1px solid #e3ece6 !important;
58
+ border-radius: 10px !important;
59
+ box-shadow: none !important;
60
+ }
61
+
62
+ /* Main title bar: strong contrast; title uses Roboto (matches body), not Roboto Slab */
63
+ .skin-green .main-header {
64
+ border-bottom: 2px solid rgba(25, 55, 42, 0.35);
65
+ box-shadow: 0 3px 14px rgba(25, 50, 38, 0.18);
66
+ }
67
+
68
+ .skin-green .main-header .logo {
69
+ background: linear-gradient(135deg, #5a8568 0%, #3d6b52 100%) !important;
70
+ color: #f7fffa !important;
71
+ border-right: 1px solid rgba(0, 0, 0, 0.12);
72
+ font-family: "Roboto", -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif !important;
73
+ font-weight: 700 !important;
74
+ font-size: 1.5rem !important;
75
+ letter-spacing: 0.02em !important;
76
+ text-shadow: 0 1px 3px rgba(0, 0, 0, 0.22);
77
+ }
78
+
79
+ .skin-green .main-header .navbar {
80
+ background: linear-gradient(135deg, #5a8568 0%, #3d6b52 100%) !important;
81
+ }
82
+
83
+ .skin-green .main-header .navbar .sidebar-toggle {
84
+ color: #f7fffa !important;
85
+ }
86
+
87
+ .skin-green .main-header .navbar .sidebar-toggle:hover {
88
+ background-color: rgba(255, 255, 255, 0.14) !important;
89
+ }
www/hexbin_RSF_logo.png ADDED