Download app.py from idiouf/African-Conservation-Human-Development-Observatory: direct link, hf CLI and curl.
- Browser
- Download file 71.8 kB
-
https://huggingface.co/spaces/idiouf/African-Conservation-Human-Development-Observatory/resolve/main/app.py
- Command line
-
hf download hf://spaces/idiouf/African-Conservation-Human-Development-Observatory/app.py
-
curl -L -o app.py https://huggingface.co/spaces/idiouf/African-Conservation-Human-Development-Observatory/resolve/main/app.py
71.8 kB
"""Africa Conservation & Human Development — an exploratory dashboard.
Built around the question Liesbet Steer put after Li et al. (2024): do the human
conditions around protected areas explain whether conservation and local
development pull together or against each other?
"""
from __future__ import annotations
import inspect
import os
from string import Template
import numpy as np
import pandas as pd
import plotly.express as px
import plotly.graph_objects as go
import streamlit as st
import streamlit.components.v1 as components
import analysis as an
import methods
import sources as src
st.set_page_config(
page_title="Africa Conservation & Human Development",
page_icon="🌍",
layout="wide",
initial_sidebar_state="auto",
menu_items={
"About": (
"Exploratory dashboard on protected areas and the human conditions "
"around them in Africa. Framed on Li et al. (2024), *The synergy "
"between protected area effectiveness and economic growth*, "
"Current Biology 34."
)
},
)
AUTHOR = {
"name": "Dr. Ibrahima Diouf",
"affiliation": "UCAD–ESP/LPAO-SF · Université de Labé–FST",
"links": [
("🌐", "Website", "https://scientific-hub-lab.github.io/Dr-Ibrahima-Diouf/"),
("🤗", "Hugging Face", "https://huggingface.co/idiouf"),
("🆔", "ORCID", "https://orcid.org/0000-0002-3157-0925"),
("in", "LinkedIn", "https://www.linkedin.com/in/dr-ibrahima-diouf-610872108/"),
("🎓", "Google Scholar",
"https://scholar.google.com/citations?user=phsMm7sAAAAJ&hl=fr"),
("🔗", "This Space",
"https://huggingface.co/spaces/idiouf/"
"African-Conservation-Human-Development-Observatory"),
],
}
SEQ_BLUE = ["#EEF4FB", "#D2E3F4", "#AACBE9", "#7FB0DC", "#5493CB",
"#2F76B4", "#1B5B95", "#114374", "#0A2C51"]
SEQ_GREEN = ["#F0F7F2", "#D3E9DA", "#A9D5BA", "#7CBE99", "#51A57A",
"#33885E", "#216A48", "#154E34", "#0C3322"]
DIVERGING = ["#8C2D3F", "#C05B5F", "#E39A88", "#F5D0BC", "#F2F2EF",
"#C4DCE4", "#8CBCD1", "#4E90B5", "#1F5F8F"]
CATEGORICAL = ["#2F6FED", "#0E9F9B", "#E8873D", "#A64CB0", "#3FA34D",
"#D6455D", "#7B6FE8", "#B08A24", "#3FA3D9", "#7A8896"]
PALETTES = {
"Dark": {
"bg": "#070F1A", "panel": "#0D1E30", "panel2": "#122840", "border": "#1D3A55",
"border_soft": "#162E45", "text": "#F2F7FC", "muted": "#9AB0C6",
"accent": "#4A9BFF", "accent_soft": "rgba(74,155,255,.15)",
"positive": "#3FBF8F", "negative": "#F2778D",
"shadow": "0 10px 30px rgba(0,0,0,.35)",
"hero1": "#0B2C52", "hero2": "#14503C", "hero3": "#5C2350",
"map_style": "carto-darkmatter",
},
"Light": {
"bg": "#F2F6FA", "panel": "#FFFFFF", "panel2": "#F6F9FD", "border": "#D8E3EF",
"border_soft": "#E6EDF5", "text": "#0F2439", "muted": "#54687E",
"accent": "#1668CE", "accent_soft": "rgba(22,104,206,.10)",
"positive": "#12805C", "negative": "#B93049",
"shadow": "0 8px 24px rgba(15,36,57,.08)",
"hero1": "#0B3A69", "hero2": "#1B6B4F", "hero3": "#7A2359",
"map_style": "carto-positron",
},
}
# ---------------------------------------------------------------------------
# Translations
# ---------------------------------------------------------------------------
TEXT = {
"English": {
"title": "Conservation & Human Development in Africa",
"subtitle": ("Where protected areas sit, what human conditions surround them, "
"and whether the two move together."),
"framing": ("Framed on Li et al. (2024), Current Biology — protected areas and "
"neighbouring communities fall into four relationships: synergy, "
"economy impaired, conservation impaired, retreat."),
"nav": "Sections",
"overview": "Overview", "areas": "Protected areas", "conditions": "Human conditions",
"relationships": "Relationships", "trajectories": "Trajectories", "data": "Data & method",
"filters": "Filters", "language": "Language / Langue", "appearance": "Appearance",
"theme_hint": "dark by default · switch to light in ⋮ → Settings → Appearance.",
"country": "Country", "countries": "Countries", "all_countries": "All of Africa",
"indicator": "Human development indicator", "year": "Reference year",
"kpi_pas": "Protected areas mapped", "kpi_countries": "Countries covered",
"kpi_area": "Total reported area", "kpi_pressure": "Median pressure index",
"map_title": "Protected areas on human development",
"map_hint": "Colour is the country indicator; each dot is a protected area.",
"no_data": "No data available for this selection.",
"loading": "Loading data from the open sources…",
"loading_map": "Downloading country boundaries…",
"sources_ok": "sources loaded", "sources_failed": "source(s) unavailable",
"select_pa": "Select a protected area",
"search": "Search by name", "iucn": "IUCN category", "min_area": "Minimum area (km²)",
"pressure": "Human pressure index", "pressure_help": (
"Rank-based 0–100 screening index: it says which parks sit in busier "
"settings, not how many people live within 10 km of a boundary."),
"pressure_built": "Built from:", "neighbours": "protected areas within 50 km",
"profile": "Profile", "context": "Country context", "nearby": "Nearest protected areas",
"education": "Education", "health": "Health", "income": "Income & poverty",
"typology": "The four relationships", "quadrant": "Where each country falls",
"cons_axis": "Conservation outcome", "dev_axis": "Human development outcome",
"period": "Period", "vs_li": "Compared with Li et al. (2024), global",
"this_dashboard": "This dashboard, Africa",
"upload_title": "Add protected area effectiveness scores",
"upload_help": ("Upload a CSV with a 'name' column and a 'score' column — "
"for example Canopy PACE scores — to unlock the "
"protected-area level analysis."),
"upload_matched": "matched to mapped protected areas",
"counterfactual_note": ("Li et al. compare each protected area with matched "
"counterfactual sites. That is not possible from the open "
"country series used here, so the reference point is the "
"African median change over the same period. Read this as "
"description, not causal attribution."),
"status": "Connector status", "licence": "Licence", "rows": "Rows",
"trend_title": "Trajectories, 1990–2024",
"pick_countries": "Pick countries to compare",
"conservation_ind": "Conservation indicator",
"no_overlap": "These two indicators have no period in common",
"typology_hint": ("Each country is placed by how its conservation and "
"human development changes compare with the African median."),
"coverage": "Coverage", "coverage_hint": "What each series actually contains",
"first_year": "From", "last_year": "To", "observations": "Observations",
"detailed": "Data, methods and analyses in detail",
"detailed_hint": "Provenance, every computation, and what each analysis supports",
"assoc_title": "Do the two axes move together?",
"assoc_pos": "Conservation and human development gains went together:",
"assoc_neg": "Conservation and human development gains pulled against each other:",
"assoc_none": "The two moved close to independently:",
"odds": "odds ratio",
"assoc_note": ("The median split fixes each margin at 50/50, so the four shares "
"are near-arithmetic; this association is the part the split does "
"not force. See Data & method."),
"demography": "Demography", "dependency": "Dependency",
"projection": "Projection", "unit_col": "Unit", "source_col": "Source",
},
"Français": {
"title": "Conservation et développement humain en Afrique",
"subtitle": ("Où se trouvent les aires protégées, quelles conditions humaines les "
"entourent, et si les deux évoluent ensemble."),
"framing": ("D'après Li et al. (2024), Current Biology — aires protégées et "
"communautés voisines relèvent de quatre relations : synergie, "
"économie entravée, conservation entravée, repli."),
"nav": "Sections",
"overview": "Vue d'ensemble", "areas": "Aires protégées", "conditions": "Conditions humaines",
"relationships": "Relations", "trajectories": "Trajectoires", "data": "Données et méthode",
"filters": "Filtres", "language": "Language / Langue", "appearance": "Apparence",
"theme_hint": "sombre par défaut · passez en clair via ⋮ → Settings → Appearance.",
"country": "Pays", "countries": "Pays", "all_countries": "Toute l'Afrique",
"indicator": "Indicateur de développement humain", "year": "Année de référence",
"kpi_pas": "Aires protégées cartographiées", "kpi_countries": "Pays couverts",
"kpi_area": "Superficie totale déclarée", "kpi_pressure": "Indice de pression médian",
"map_title": "Aires protégées et développement humain",
"map_hint": "La couleur est l'indicateur national ; chaque point est une aire protégée.",
"no_data": "Aucune donnée disponible pour cette sélection.",
"loading": "Chargement des données depuis les sources ouvertes…",
"loading_map": "Téléchargement des frontières nationales…",
"sources_ok": "sources chargées", "sources_failed": "source(s) indisponible(s)",
"select_pa": "Sélectionner une aire protégée",
"search": "Rechercher par nom", "iucn": "Catégorie UICN", "min_area": "Superficie minimale (km²)",
"pressure": "Indice de pression humaine", "pressure_help": (
"Indice de criblage 0–100 fondé sur les rangs : il indique quelles aires se "
"trouvent dans les contextes les plus peuplés, non combien de personnes vivent "
"à moins de 10 km d'une limite."),
"pressure_built": "Construit à partir de :",
"neighbours": "aires protégées à moins de 50 km",
"profile": "Profil", "context": "Contexte national", "nearby": "Aires protégées les plus proches",
"education": "Éducation", "health": "Santé", "income": "Revenu et pauvreté",
"typology": "Les quatre relations", "quadrant": "Position de chaque pays",
"cons_axis": "Résultat de conservation", "dev_axis": "Résultat de développement humain",
"period": "Période", "vs_li": "Comparaison avec Li et al. (2024), monde",
"this_dashboard": "Ce tableau de bord, Afrique",
"upload_title": "Ajouter des scores d'efficacité des aires protégées",
"upload_help": ("Déposez un CSV avec une colonne 'name' et une colonne 'score' — "
"par exemple les scores PACE de Canopy — pour activer l'analyse "
"au niveau des aires protégées."),
"upload_matched": "appariées aux aires protégées cartographiées",
"counterfactual_note": ("Li et al. comparent chaque aire protégée à des sites "
"contrefactuels appariés. Ce n'est pas possible à partir des "
"séries nationales ouvertes utilisées ici : la référence est "
"donc la variation médiane africaine sur la même période. "
"À lire comme une description, non comme une attribution causale."),
"status": "État des connecteurs", "licence": "Licence", "rows": "Lignes",
"trend_title": "Trajectoires, 1990–2024",
"pick_countries": "Choisir des pays à comparer",
"conservation_ind": "Indicateur de conservation",
"no_overlap": "Ces deux indicateurs n'ont aucune période commune",
"typology_hint": ("Chaque pays est situé selon la comparaison de ses variations "
"de conservation et de développement humain à la médiane africaine."),
"coverage": "Couverture", "coverage_hint": "Ce que contient réellement chaque série",
"first_year": "De", "last_year": "À", "observations": "Observations",
"detailed": "Données, méthodes et analyses en détail",
"detailed_hint": "Provenance, chaque calcul, et ce que permet chaque analyse",
"assoc_title": "Les deux axes évoluent-ils ensemble ?",
"assoc_pos": "Conservation et développement humain ont progressé de concert :",
"assoc_neg": "Conservation et développement humain se sont opposés :",
"assoc_none": "Les deux ont évolué de façon quasi indépendante :",
"odds": "rapport de cotes",
"assoc_note": ("La coupe médiane fixe chaque marge à 50/50 : les quatre parts sont "
"donc surtout arithmétiques ; cette association est ce que la coupe "
"n'impose pas. Voir Données et méthode."),
"demography": "Démographie", "dependency": "Dépendance",
"projection": "Projection", "unit_col": "Unité", "source_col": "Source",
},
}
# ---------------------------------------------------------------------------
# Streamlit compatibility + styling
# ---------------------------------------------------------------------------
_WIDTH_CACHE: dict[str, dict] = {}
def stretch(fn) -> dict:
key = getattr(fn, "__name__", str(fn))
if key not in _WIDTH_CACHE:
try:
supported = "width" in inspect.signature(fn).parameters
except (TypeError, ValueError):
supported = False
_WIDTH_CACHE[key] = {"width": "stretch"} if supported else {"use_container_width": True}
return _WIDTH_CACHE[key]
# Both palettes ship (see the Dockerfile), so light and dark both render
# correctly — native widgets, data grids and pop-overs included — and the reader
# can switch at any time from ⋮ → Settings → Appearance.
#
# Streamlit has no API to choose which one a first-time reader gets: with two
# palettes it follows the browser. DEFAULT_THEME closes that gap by writing the
# reader's preference key once, on a first visit only, and never touching it
# again. Set CONSERVATION_HD_THEME=auto to follow the browser instead.
DEFAULT_THEME = os.environ.get("CONSERVATION_HD_THEME", "dark").strip().lower()
def apply_default_theme(preference: str) -> None:
"""Seed the reader's theme choice on their very first visit.
Streamlit writes its own preference key eagerly, with the value "System",
before any component runs — so the mere presence of a key means nothing.
Only "System" counts as 'no choice made'; an explicit "Light" or "Dark" is
the reader's and is never touched. A sessionStorage flag makes the single
reload idempotent, so this can never loop.
"""
if preference not in ("dark", "light"):
return
name = preference.capitalize()
components.html(
f"""
<script>
(function () {{
try {{
const store = window.parent.localStorage;
if (window.parent.sessionStorage.getItem('themeSeeded')) return;
const keys = Object.keys(store).filter(k => k.startsWith('stActiveTheme-'));
const chosen = keys.some(k => {{
const v = (store.getItem(k) || '').replace(/"/g, '');
return v === 'Light' || v === 'Dark';
}});
window.parent.sessionStorage.setItem('themeSeeded', '1');
if (chosen) return; // the reader decided; leave it
const path = window.parent.location.pathname || '/';
store.setItem('stActiveTheme-' + path + '-v2', JSON.stringify('{name}'));
store.setItem('stActiveTheme-/-v2', JSON.stringify('{name}'));
window.parent.location.reload();
}} catch (e) {{ /* a sandboxed frame just means the browser default wins */ }}
}})();
</script>
""",
height=0,
)
def active_theme() -> str:
try:
return "Light" if st.context.theme.type == "light" else "Dark"
except Exception:
return "Dark" if DEFAULT_THEME != "light" else "Light"
CSS = Template("""
<style>
@import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700;800&display=swap');
:root {
--bg:$bg; --panel:$panel; --panel2:$panel2; --border:$border; --border-soft:$border_soft;
--text:$text; --muted:$muted; --accent:$accent; --accent-soft:$accent_soft;
--positive:$positive; --negative:$negative; --shadow:$shadow;
--radius:16px; --radius-sm:10px;
}
html, body, [data-testid="stAppViewContainer"] {
background:var(--bg); color:var(--text);
font-family:'Inter',-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;
-webkit-text-size-adjust:100%;
}
[data-testid="stHeader"] { background:transparent; }
.block-container { max-width:1560px; padding:1.1rem 1.5rem 2.2rem; }
h1,h2,h3,h4,p,span,label,li,div { color:var(--text); }
[data-testid="stSidebar"] { background:var(--panel); border-right:1px solid var(--border-soft); }
.side-brand { display:flex; align-items:center; gap:11px; padding:4px 2px 12px; }
.side-brand .glyph { width:40px; height:40px; flex:0 0 40px; border-radius:12px;
display:grid; place-items:center; font-size:1.25rem;
background:linear-gradient(140deg,$hero1,var(--accent)); box-shadow:var(--shadow); }
.side-brand .txt strong { font-size:.94rem; font-weight:800; line-height:1.15; display:block; }
.side-brand .txt span { font-size:.72rem; color:var(--muted); }
.side-label { font-size:.68rem; font-weight:800; letter-spacing:.09em; text-transform:uppercase;
color:var(--muted); margin:14px 2px 6px; }
.hero { position:relative; background:linear-gradient(118deg,$hero1 0%,$hero2 55%,$hero3 100%);
border:1px solid rgba(255,255,255,.14); border-radius:20px; padding:26px 28px;
margin-bottom:14px; box-shadow:var(--shadow); overflow:hidden; }
.hero::after { content:""; position:absolute; inset:0;
background:radial-gradient(circle at 85% -20%,rgba(255,255,255,.18),transparent 55%);
pointer-events:none; }
.hero h1, .hero h1 span, .hero h1 [data-heading-text] { color:#fff !important; }
.hero h1 { font-size:clamp(1.55rem,3.4vw,2.7rem); line-height:1.06; margin:0 0 10px;
letter-spacing:-.035em; font-weight:800 !important; }
.hero h1 [data-testid="stHeaderActionElements"] { display:none; }
.hero p, .hero p span { color:rgba(255,255,255,.92) !important; font-size:clamp(.88rem,1.4vw,1rem);
margin:0; max-width:66ch; line-height:1.5; }
.hero-tags { display:flex; flex-wrap:wrap; gap:8px; margin-top:14px; }
.hero-tag { display:inline-flex; align-items:center; gap:6px; background:rgba(255,255,255,.14);
border:1px solid rgba(255,255,255,.20); color:#fff !important; border-radius:999px;
padding:4px 11px; font-size:.73rem; font-weight:600; white-space:nowrap; }
.source-line { color:var(--muted); font-size:.8rem; margin:0 2px 16px; line-height:1.5; }
.section-head { display:flex; align-items:baseline; justify-content:space-between;
flex-wrap:wrap; gap:6px; margin:18px 2px 10px; }
.section-head h2 { margin:0; font-size:clamp(1.12rem,2.2vw,1.42rem); font-weight:700;
letter-spacing:-.02em; }
.section-head .hint { color:var(--muted); font-size:.8rem; }
.kpi-grid { display:grid; grid-template-columns:repeat(auto-fit,minmax(215px,1fr));
gap:12px; margin-bottom:6px; }
.kpi { border:1px solid var(--border); border-radius:var(--radius);
background:linear-gradient(150deg,var(--panel),var(--panel2)); box-shadow:var(--shadow);
padding:15px 16px 14px; display:flex; gap:12px; align-items:flex-start;
transition:transform .15s ease,border-color .15s ease; }
.kpi:hover { transform:translateY(-2px); border-color:var(--accent); }
.kpi .ic { width:36px; height:36px; flex:0 0 36px; border-radius:10px; display:grid;
place-items:center; font-size:1.05rem; background:var(--accent-soft);
border:1px solid var(--border-soft); }
.kpi .lab { color:var(--muted); font-size:.7rem; font-weight:700; text-transform:uppercase;
letter-spacing:.06em; }
.kpi .val { font-size:clamp(1.25rem,3vw,1.6rem); font-weight:800; letter-spacing:-.02em;
line-height:1.18; margin-top:4px; }
.kpi .note { color:var(--muted); font-size:.74rem; margin-top:3px; }
.card { border:1px solid var(--border); border-radius:var(--radius); background:var(--panel);
padding:18px 20px; margin-bottom:12px; box-shadow:var(--shadow); }
.card h3 { margin-top:0; font-size:1.05rem; }
.small-muted { color:var(--muted); font-size:.82rem; line-height:1.6; }
.rel-grid { display:grid; grid-template-columns:repeat(auto-fit,minmax(210px,1fr)); gap:10px; }
.rel { border-radius:var(--radius); padding:14px 15px; border:1px solid var(--border);
background:var(--panel); }
.rel .nm { font-weight:800; font-size:.9rem; }
.rel .pc { font-size:1.5rem; font-weight:800; margin:4px 0 2px; }
.rel .ds { color:var(--muted); font-size:.76rem; line-height:1.5; }
.status-row { display:flex; align-items:center; gap:10px; padding:9px 12px;
border:1px solid var(--border); border-radius:var(--radius-sm); margin-bottom:7px;
background:var(--panel); flex-wrap:wrap; }
.dot { width:9px; height:9px; border-radius:50%; flex:0 0 9px; }
.dot.ok { background:var(--positive); }
.dot.cache { background:var(--accent); }
.dot.bad { background:var(--negative); }
.status-row .nm { font-weight:700; font-size:.85rem; min-width:180px; }
.status-row .dt { color:var(--muted); font-size:.78rem; }
div[data-testid="stPlotlyChart"] { background:var(--panel); border:1px solid var(--border);
border-radius:var(--radius); overflow:hidden; box-shadow:var(--shadow); }
[data-testid="stDataFrame"] { border:1px solid var(--border); border-radius:var(--radius-sm);
overflow:hidden; }
.stButton > button, .stDownloadButton > button { border-radius:var(--radius-sm);
border:1px solid var(--border); font-weight:650; min-height:42px; }
.stButton > button:hover, .stDownloadButton > button:hover { border-color:var(--accent);
background:var(--accent); color:#fff; }
div[role="radiogroup"] { display:flex; flex-wrap:wrap; gap:6px; }
div[role="radiogroup"] > label { background:var(--panel2); border:1px solid var(--border);
padding:7px 13px; border-radius:999px; margin:0; }
div[role="radiogroup"] > label:has(input:checked) { border-color:var(--accent);
background:var(--accent-soft); box-shadow:inset 0 0 0 1px var(--accent); }
[data-testid="stSidebar"] div[role="radiogroup"] { flex-direction:column; gap:4px; }
[data-testid="stSidebar"] div[role="radiogroup"] > label { width:100%; border-radius:var(--radius-sm); }
div[data-testid="stSegmentedControl"] button, div[data-testid="stButtonGroup"] button {
border-radius:999px !important; min-height:40px; font-weight:650 !important; }
[data-testid="stSelectbox"] > div > div, [data-testid="stMultiSelect"] > div > div,
[data-testid="stTextInput"] > div > div { border-radius:var(--radius-sm) !important; min-height:42px; }
.stTabs [data-baseweb="tab-list"] { gap:4px; border-bottom:1px solid var(--border-soft);
overflow-x:auto; scrollbar-width:none; }
.stTabs [data-baseweb="tab"] { border-radius:var(--radius-sm) var(--radius-sm) 0 0;
padding:8px 14px; white-space:nowrap; font-weight:600; }
.stTabs [aria-selected="true"] { background:var(--accent-soft); }
div[data-testid="stExpander"] { border:1px solid var(--border) !important;
border-radius:var(--radius) !important; background:var(--panel); overflow:hidden; }
[data-testid="stHorizontalBlock"] { flex-wrap:wrap; gap:12px; }
[data-testid="stColumn"] { min-width:min(100%,320px); }
.signature { margin-top:34px; padding:22px 18px 18px; text-align:center;
border-top:1px solid var(--border); border-radius:var(--radius) var(--radius) 0 0;
background:linear-gradient(180deg,transparent,var(--panel2)); }
.sig-name { font-weight:800; font-size:1rem; }
.sig-affil { color:var(--muted); font-size:.82rem; margin-top:3px; }
.sig-links { display:flex; flex-wrap:wrap; justify-content:center; gap:8px; margin:14px 0 12px; }
.sig-links a { display:inline-flex; align-items:center; padding:7px 14px; border-radius:999px;
border:1px solid var(--border); background:var(--panel); color:var(--text) !important;
text-decoration:none; font-size:.8rem; font-weight:650; min-height:38px; }
.sig-links a { gap:8px; }
.sig-links a .ic, .side-links a .ic {
font-size:.9em; opacity:.85; font-style:normal; font-weight:700; line-height:1;
}
.sig-links a span:not(.ic) { text-decoration:underline; text-underline-offset:3px; }
.sig-links a:hover { border-color:var(--accent); background:var(--accent); color:#fff !important; }
.sig-links a:hover .ic { opacity:1; }
.side-links a { display:inline-flex; align-items:center; gap:5px; }
.sig-meta { color:var(--muted); font-size:.76rem; line-height:1.7; }
.side-links { display:flex; flex-wrap:wrap; gap:6px; margin-top:4px; }
.side-links a { font-size:.72rem; font-weight:650; padding:5px 10px; border-radius:999px;
border:1px solid var(--border); background:var(--panel2); color:var(--text) !important;
text-decoration:none; }
@media (max-width:900px) {
.block-container { padding:.7rem .8rem 1.6rem; }
.hero { padding:20px 18px; border-radius:16px; }
.kpi-grid { grid-template-columns:repeat(auto-fit,minmax(160px,1fr)); gap:9px; }
.kpi .lab { font-size:.63rem; }
.kpi .val { font-size:1.24rem; }
}
@media (max-width:700px) {
div[data-testid="stPlotlyChart"] .modebar { display:none !important; }
}
@media (max-width:560px) {
.kpi-grid { grid-template-columns:repeat(2,minmax(0,1fr)); }
.sig-links a { flex:1 1 calc(50% - 8px); justify-content:center; }
}
</style>
""")
def kpi(label: str, value: str, note: str = "", icon: str = "") -> str:
return ('<div class="kpi">'
f'<div class="ic">{icon}</div><div class="body">'
f'<div class="lab">{label}</div><div class="val">{value}</div>'
f'<div class="note">{note}</div></div></div>')
def kpi_grid(cards: list[str]) -> None:
st.markdown(f'<div class="kpi-grid">{"".join(cards)}</div>', unsafe_allow_html=True)
def section(title: str, hint: str = "") -> None:
st.markdown(f'<div class="section-head"><h2>{title}</h2><span class="hint">{hint}</span></div>',
unsafe_allow_html=True)
def fmt(value, decimals: int = 1) -> str:
if value is None or pd.isna(value):
return "—"
value = float(value)
if abs(value) >= 1_000_000_000:
return f"{value / 1_000_000_000:.2f} B"
if abs(value) >= 1_000_000:
return f"{value / 1_000_000:.1f} M"
if abs(value) >= 10_000:
return f"{value / 1_000:.0f} K"
return f"{value:,.{decimals}f}"
def style_figure(fig: go.Figure, height: int = 420, theme: str = "Dark",
legend_top: bool = True) -> go.Figure:
p = PALETTES[theme]
grid = "#1E3A54" if theme == "Dark" else "#E8EFF6"
fig.update_layout(
height=height, autosize=True,
paper_bgcolor=p["panel"], plot_bgcolor=p["panel"],
font=dict(family="Inter, -apple-system, Segoe UI, sans-serif", color=p["text"], size=13),
margin=dict(l=14, r=14, t=54, b=14),
title=dict(font=dict(size=16, color=p["text"]), x=0.01, xanchor="left", y=0.965),
hoverlabel=dict(bgcolor=p["panel2"], font_color=p["text"], bordercolor=p["border"]),
xaxis=dict(gridcolor=grid, zerolinecolor=grid, linecolor=grid,
title_font=dict(color=p["muted"], size=12),
tickfont=dict(color=p["muted"], size=11), automargin=True),
yaxis=dict(gridcolor=grid, zerolinecolor=grid, linecolor=grid,
title_font=dict(color=p["muted"], size=12),
tickfont=dict(color=p["muted"], size=11), automargin=True),
colorway=CATEGORICAL,
modebar=dict(bgcolor="rgba(0,0,0,0)", color=p["muted"], activecolor=p["accent"]),
)
if legend_top:
fig.update_layout(legend=dict(orientation="h", yanchor="bottom", y=1.005,
xanchor="right", x=1, bgcolor="rgba(0,0,0,0)",
font=dict(size=11, color=p["muted"]), title_text=""))
return fig
PLOT_CONFIG = {"displaylogo": False, "responsive": True,
"modeBarButtonsToRemove": ["select2d", "lasso2d", "autoScale2d",
"toggleSpikelines"]}
def show(fig: go.Figure, filename: str = "conservation_hd") -> None:
cfg = dict(PLOT_CONFIG)
cfg["toImageButtonOptions"] = {"format": "png", "filename": filename, "scale": 2}
st.plotly_chart(fig, config=cfg, theme=None, **stretch(st.plotly_chart))
# ---------------------------------------------------------------------------
# Continental map: country choropleth + protected area points
# ---------------------------------------------------------------------------
def _rings(geometry: dict) -> list[list]:
"""Exterior rings of a (Multi)Polygon geometry."""
kind, coords = geometry.get("type"), geometry.get("coordinates")
if kind == "Polygon":
return [coords[0]] if coords else []
if kind == "MultiPolygon":
return [poly[0] for poly in coords if poly]
return []
def _sample(scale: list[str], frac: float) -> str:
frac = min(max(float(frac), 0.0), 1.0)
try:
return px.colors.sample_colorscale(px.colors.make_colorscale(scale), [frac])[0]
except Exception:
return scale[int(frac * (len(scale) - 1))]
def continent_map(boundaries: dict, values: pd.DataFrame, points: pd.DataFrame,
value_label: str, unit: str, theme: str, height: int = 620) -> go.Figure:
"""Self-contained map: no tile server, no remote topojson.
Country polygons come from geoBoundaries and are drawn as filled paths on a
plain cartesian canvas, so the view renders identically online and offline.
"""
p = PALETTES[theme]
lookup = dict(zip(values["iso3"], values["value"])) if not values.empty else {}
finite = [v for v in lookup.values() if pd.notna(v)]
vmin, vmax = (min(finite), max(finite)) if finite else (0.0, 1.0)
span = (vmax - vmin) or 1.0
missing = "#22374D" if theme == "Dark" else "#E4EBF2"
fig = go.Figure()
for feature in boundaries.get("features", []):
rings = _rings(feature.get("geometry", {}))
if not rings:
continue
props = feature.get("properties", {})
iso3 = props.get("iso3", "")
value = lookup.get(iso3)
known = value is not None and pd.notna(value)
xs: list = []
ys: list = []
for ring in rings:
xs.extend([c[0] for c in ring] + [None])
ys.extend([c[1] for c in ring] + [None])
label = (f"<b>{props.get('name', iso3)}</b><br>{value:,.1f} {unit}"
if known else f"<b>{props.get('name', iso3)}</b><br>no data")
fig.add_trace(go.Scatter(
x=xs, y=ys, mode="lines", fill="toself",
fillcolor=_sample(SEQ_GREEN, (value - vmin) / span) if known else missing,
line=dict(color=p["border"], width=0.6),
hoveron="fills", hoverinfo="text", text=label,
showlegend=False, name=props.get("name", iso3),
))
fig.add_trace(go.Scatter(
x=[None], y=[None], mode="markers", hoverinfo="skip", showlegend=False,
marker=dict(colorscale=px.colors.make_colorscale(SEQ_GREEN), cmin=vmin, cmax=vmax,
color=[vmin], size=0.1, showscale=True,
colorbar=dict(title=dict(text=f"{value_label} ({unit})",
font=dict(color=p["muted"], size=11), side="top"),
orientation="h", thickness=12, len=0.62,
x=0.5, xanchor="center", y=-0.02, yanchor="top",
outlinewidth=0, bgcolor="rgba(0,0,0,0)",
tickfont=dict(color=p["muted"], size=10))),
))
if not points.empty:
sizes = points["area_km2"].fillna(points["area_km2"].median())
sizes = (sizes.clip(lower=1) ** 0.22)
sizes = 5 + 13 * (sizes - sizes.min()) / ((sizes.max() - sizes.min()) or 1)
custom = points[["country", "area_km2", "pressure"]].to_numpy()
fig.add_trace(go.Scatter(
x=points["lon"], y=points["lat"], mode="markers", name="Protected areas",
marker=dict(size=sizes, color=points["pressure"], colorscale=SEQ_BLUE,
line=dict(width=0.7, color=p["panel"]), opacity=0.92,
showscale=False),
text=points["name"], customdata=custom,
hovertemplate=("<b>%{text}</b><br>%{customdata[0]}"
"<br>%{customdata[1]:,.0f} km²"
"<br>pressure %{customdata[2]:.0f}<extra></extra>"),
showlegend=False,
))
fig = style_figure(fig, height, theme, legend_top=False)
fig.update_layout(
margin=dict(l=6, r=6, t=16, b=58),
xaxis=dict(visible=False, showgrid=False, zeroline=False),
yaxis=dict(visible=False, showgrid=False, zeroline=False,
scaleanchor="x", scaleratio=1),
plot_bgcolor=p["panel"], dragmode="pan",
)
return fig
# ---------------------------------------------------------------------------
# Cached data loading
# ---------------------------------------------------------------------------
@st.cache_data(show_spinner=False, ttl=60 * 60 * 12)
def load_protected_areas():
ds = src.fetch_protected_areas()
return ds.frame, ds.status
@st.cache_data(show_spinner=False, ttl=60 * 60 * 12)
def load_indicators():
ds = src.fetch_indicators()
return ds.frame, ds.status
@st.cache_data(show_spinner=False, ttl=60 * 60 * 12)
def load_un_population():
ds = src.fetch_un_population()
return ds.frame, ds.status
@st.cache_data(show_spinner=False, ttl=60 * 60 * 12)
def load_subnational():
ds = src.fetch_subnational_hdi()
return ds.frame, ds.status
@st.cache_data(show_spinner=False, ttl=60 * 60 * 24)
def load_boundaries(level: int = 0):
return src.fetch_boundaries(level)
@st.cache_data(show_spinner=False)
def enrich(pas: pd.DataFrame, indicators: pd.DataFrame) -> pd.DataFrame:
"""Attach country context, neighbour counts and the pressure index."""
if pas.empty:
return pas
codes = ["EN.POP.DNST", "SP.RUR.TOTL.ZS", "SP.POP.TOTL", "SP.DYN.LE00.IN",
"SE.SEC.ENRR", "NY.GNP.PCAP.PP.CD", "SI.POV.DDAY", "SH.DYN.MORT",
"ER.LND.PTLD.ZS", "AG.LND.FRST.ZS", "EG.ELC.ACCS.ZS"] + list(src.UN_INDICATORS)
out = an.attach_country_context(pas, indicators, codes)
out["neighbours_50km"] = an.neighbour_counts(out).values
out["pressure"] = an.pressure_index(out).values
return out
# ---------------------------------------------------------------------------
# Sidebar
# ---------------------------------------------------------------------------
apply_default_theme(DEFAULT_THEME)
theme = active_theme()
palette = PALETTES[theme]
with st.sidebar:
st.markdown(
'<div class="side-brand"><div class="glyph">🌍</div><div class="txt">'
"<strong>Conservation & Human Development</strong>"
"<span>Africa · protected areas & people</span></div></div>",
unsafe_allow_html=True,
)
language = st.radio(TEXT["English"]["language"], ["English", "Français"],
horizontal=True, key="lang")
t = TEXT[language]
st.caption(f"{t['appearance']} · {t['theme_hint']}")
st.markdown(CSS.substitute(**{k: v for k, v in palette.items() if k != "map_style"}),
unsafe_allow_html=True)
with st.spinner(t["loading"]):
pas_raw, pa_status = load_protected_areas()
wb_indicators, ind_status = load_indicators()
un_indicators, un_status = load_un_population()
shdi, shdi_status = load_subnational()
statuses = [pa_status, ind_status, un_status, shdi_status]
# `indicators_all` keeps the UN median-variant projections for the trajectory
# charts; `indicators` is observations only and is what every analytic view uses.
indicators_all = pd.concat([f for f in (wb_indicators, un_indicators) if not f.empty],
ignore_index=True) if (not wb_indicators.empty
or not un_indicators.empty) else pd.DataFrame()
indicators = an.observations_only(indicators_all)
pas = enrich(pas_raw, indicators) if not pas_raw.empty else pas_raw
COUNTRIES = sorted(pas["country"].dropna().unique().tolist()) if not pas.empty else []
if not COUNTRIES and not indicators.empty:
COUNTRIES = sorted(indicators["country"].dropna().unique().tolist())
# Welfare outcomes usable as the development axis of the typology. Demographic
# structure is context, not a welfare outcome, so only the two standard markers
# of the demographic transition join them.
HD_CODES = [c for c, m in src.INDICATORS.items()
if m["group"] in ("Education", "Health", "Income")]
HD_CODES += [c for c in ("UN:67", "UN:86") if c in src.UN_INDICATORS]
CONS_CODES = [c for c, m in src.INDICATORS.items() if m["group"] == "Conservation"]
CONTEXT_CODES = src.DEMOGRAPHY_CODES + src.DEPENDENCY_CODES
MAP_CODES = HD_CODES + [c for c in CONTEXT_CODES if c not in HD_CODES]
def label_of(code: str) -> str:
meta = src.ALL_INDICATORS.get(code, {})
return meta.get("label", code)
SOURCE_TAG = {"World Bank": "WB", "UN WPP 2024": "UN WPP"}
def label_with_source(code: str) -> str:
"""Selector label. Provenance is always visible, never inferred."""
meta = src.ALL_INDICATORS.get(code, {})
tag = SOURCE_TAG.get(meta.get("source", ""), meta.get("source", "?"))
return f"{meta.get('label', code)} · {tag}"
def unit_of(code: str) -> str:
return src.ALL_INDICATORS.get(code, {}).get("unit", "")
with st.sidebar:
st.markdown(f'<div class="side-label">{t["nav"]}</div>', unsafe_allow_html=True)
NAV = {
t["overview"]: f"🗺️ {t['overview']}",
t["areas"]: f"🌳 {t['areas']}",
t["conditions"]: f"👥 {t['conditions']}",
t["relationships"]: f"⚖️ {t['relationships']}",
t["trajectories"]: f"📈 {t['trajectories']}",
t["data"]: f"📚 {t['data']}",
}
nav_label = st.radio("Navigation", list(NAV.values()), index=0,
label_visibility="collapsed", key="nav")
page = next(k for k, v in NAV.items() if v == nav_label)
st.markdown(f'<div class="side-label">{t["filters"]}</div>', unsafe_allow_html=True)
hd_code = st.selectbox(t["indicator"], MAP_CODES, index=0,
format_func=label_with_source, key="hd_code")
country_filter = st.multiselect(t["countries"], COUNTRIES, default=[], key="country_filter",
placeholder=t["all_countries"])
st.markdown(f'<div class="side-label">{t["upload_title"]}</div>', unsafe_allow_html=True)
upload = st.file_uploader("PACE / effectiveness CSV", type=["csv"],
label_visibility="collapsed", key="pace_upload")
st.caption(t["upload_help"])
ok_count = sum(1 for s in statuses if s.ok)
bad = [s for s in statuses if not s.ok]
st.markdown(f'<div class="side-label">{t["status"]}</div>', unsafe_allow_html=True)
st.caption(f"{ok_count}/{len(statuses)} {t['sources_ok']}"
+ (f" · {len(bad)} {t['sources_failed']}" if bad else ""))
st.markdown(f'<div class="side-label">{AUTHOR["name"]}</div>', unsafe_allow_html=True)
st.markdown('<div class="side-links">' + "".join(
f'<a href="{u}" target="_blank" rel="noopener">'
f'<span class="ic">{i}</span>{n}</a>' for i, n, u in AUTHOR["links"]
) + "</div>", unsafe_allow_html=True)
effectiveness = pd.DataFrame()
match_report: dict = {}
if upload is not None:
ds = src.read_effectiveness_upload(upload)
statuses.append(ds.status)
if ds.ok:
pas, match_report = an.match_effectiveness(pas, ds.frame)
effectiveness = ds.frame
view = pas if not country_filter else pas[pas["country"].isin(country_filter)]
ind_view = (indicators if not country_filter
else indicators[indicators["country"].isin(country_filter)])
# ---------------------------------------------------------------------------
# Header
# ---------------------------------------------------------------------------
st.markdown(
f"""
<div class="hero">
<h1>{t['title']}</h1>
<p>{t['subtitle']}</p>
<div class="hero-tags">
<span class="hero-tag">🌳 {len(pas):,} {t['areas'].lower()}</span>
<span class="hero-tag">🌍 {len(COUNTRIES)} {t['countries'].lower()}</span>
<span class="hero-tag">📅 1990–2024</span>
<span class="hero-tag">📄 Li et al. 2024</span>
</div>
</div>
<div class="source-line">{t['framing']}</div>
""",
unsafe_allow_html=True,
)
if not pa_status.ok or not ind_status.ok:
failed = ", ".join(s.name for s in statuses if not s.ok)
st.error(
f"**{t['sources_failed'].capitalize()}: {failed}.** "
+ (" · ".join(f"{s.name}: {s.detail}" for s in statuses if not s.ok))
+ f" \n→ {t['data']}"
)
def map_style() -> str:
return palette["map_style"]
# ---------------------------------------------------------------------------
# Page: Overview
# ---------------------------------------------------------------------------
if page == t["overview"]:
total_area = view["area_km2"].sum() if "area_km2" in view else np.nan
kpi_grid([
kpi(t["kpi_pas"], f"{len(view):,}", "Wikidata / WDPA", "🌳"),
kpi(t["kpi_countries"], str(view["country"].nunique() if not view.empty else 0),
"Africa", "🌍"),
kpi(t["kpi_area"], f"{fmt(total_area)} km²", "reported", "📐"),
kpi(t["kpi_pressure"],
fmt(view["pressure"].median()) if "pressure" in view and not view.empty else "—",
"0–100", "👥"),
])
if "UN:83" in view.columns and view["UN:83"].notna().any():
kpi_grid([
kpi(label_of("UN:54"), fmt(view["UN:54"].median()), unit_of("UN:54"), "🏘️"),
kpi(label_of("UN:83"), fmt(view["UN:83"].median()), unit_of("UN:83"), "🧒"),
kpi(label_of("UN:67"), fmt(view["UN:67"].median()), unit_of("UN:67"), "🎂"),
kpi(label_of("UN:51"), fmt(view["UN:51"].median()), unit_of("UN:51"), "📈"),
])
st.caption("UN World Population Prospects 2024, median variant · "
"median across the countries in view, latest estimate year."
if language == "English" else
"Nations Unies, World Population Prospects 2024, variante médiane · "
"médiane des pays affichés, dernière année d'estimation.")
section(t["map_title"], t["map_hint"])
if view.empty:
st.info(t["no_data"])
else:
with st.spinner(t["loading_map"]):
boundaries, bnd_status = load_boundaries(0)
if not bnd_status.ok:
st.warning(f"{bnd_status.name}: {bnd_status.detail}")
latest = an.latest_values(ind_view, hd_code)
points = view.dropna(subset=["lat", "lon"]).copy()
show(
continent_map(boundaries, latest, points, label_of(hd_code),
src.ALL_INDICATORS[hd_code]["unit"], theme),
"africa_conservation_hd_map",
)
used = an.pressure_components(view)
st.caption(t["pressure_help"] + " " + t["pressure_built"] + " "
+ ", ".join(label_with_source(c) if c in src.ALL_INDICATORS
else t["neighbours"] for c in used) + ".")
left, right = st.columns([1, 1], gap="medium")
with left:
if latest.empty:
st.info(t["no_data"])
else:
top = latest.nlargest(15, "value").sort_values("value")
bar = px.bar(top, x="value", y="country", orientation="h",
color="value", color_continuous_scale=SEQ_GREEN,
title=label_of(hd_code),
labels={"value": src.ALL_INDICATORS[hd_code]["unit"], "country": ""})
bar.update_coloraxes(showscale=False)
bar.update_traces(hovertemplate="<b>%{y}</b><br>%{x:,.1f}<extra></extra>",
marker_line_width=0)
show(style_figure(bar, 470, theme, legend_top=False), "top_indicator")
with right:
counts = an.summarise_by_country(view).head(15).sort_values("n_pas")
if counts.empty:
st.info(t["no_data"])
else:
bar = px.bar(counts, x="n_pas", y="country", orientation="h",
color="n_pas", color_continuous_scale=SEQ_BLUE,
title=t["kpi_pas"], labels={"n_pas": "", "country": ""})
bar.update_coloraxes(showscale=False)
bar.update_traces(hovertemplate="<b>%{y}</b><br>%{x:,.0f}<extra></extra>",
marker_line_width=0)
show(style_figure(bar, 470, theme, legend_top=False), "pa_counts")
# ---------------------------------------------------------------------------
# Page: Protected areas
# ---------------------------------------------------------------------------
elif page == t["areas"]:
section(t["areas"], f"{len(view):,}")
if view.empty:
st.info(t["no_data"])
else:
c1, c2, c3 = st.columns([1.4, 1, 1], gap="medium")
with c1:
query = st.text_input(t["search"], "", key="pa_search", placeholder=t["search"])
with c2:
cats = sorted([c for c in view["iucn_category"].dropna().unique() if c])
cat = st.multiselect(t["iucn"], cats, default=[], key="pa_iucn")
with c3:
max_area = float(view["area_km2"].max()) if view["area_km2"].notna().any() else 0.0
min_area = st.slider(t["min_area"], 0.0, max(max_area, 1.0), 0.0,
key="pa_area") if max_area > 0 else 0.0
filtered = view.copy()
if query:
filtered = filtered[filtered["name"].str.contains(query, case=False, na=False)]
if cat:
filtered = filtered[filtered["iucn_category"].isin(cat)]
if min_area:
filtered = filtered[filtered["area_km2"].fillna(0) >= min_area]
if filtered.empty:
st.info(t["no_data"])
else:
cols = ["name", "country", "iucn_category", "area_km2", "established",
"pressure", "neighbours_50km"]
if "effectiveness" in filtered.columns and filtered["effectiveness"].notna().any():
cols.append("effectiveness")
table = filtered[cols].sort_values("area_km2", ascending=False)
st.dataframe(table.round(1), hide_index=True, height=340,
**stretch(st.dataframe))
st.download_button(
"⬇ CSV",
data=filtered.to_csv(index=False).encode("utf-8"),
file_name="african_protected_areas_with_context.csv",
mime="text/csv", **stretch(st.download_button),
)
section(t["profile"])
names = filtered.sort_values("name")["name"].tolist()
chosen = st.selectbox(t["select_pa"], names, key="pa_choice")
row = filtered[filtered["name"] == chosen].iloc[0]
kpi_grid([
kpi(t["country"], str(row.get("country", "—")),
str(row.get("iucn_category") or "—"), "🌍"),
kpi("Area", f"{fmt(row.get('area_km2'))} km²",
f"est. {int(row['established'])}" if pd.notna(row.get("established")) else "",
"📐"),
kpi(t["pressure"], fmt(row.get("pressure")),
f"{int(row.get('neighbours_50km', 0))} PAs < 50 km", "👥"),
kpi(label_of(hd_code), fmt(row.get(hd_code)),
src.ALL_INDICATORS[hd_code]["unit"], "📊"),
])
pcol, mcol = st.columns([1, 1], gap="medium")
with pcol:
context_rows = []
for code in ["SE.SEC.ENRR", "SP.DYN.LE00.IN", "SH.DYN.MORT",
"NY.GNP.PCAP.PP.CD", "SI.POV.DDAY", "EG.ELC.ACCS.ZS",
"AG.LND.FRST.ZS", "UN:54", "UN:67", "UN:83", "UN:86",
"UN:46"]:
if code in row.index:
context_rows.append({
t["context"]: label_of(code),
"": round(float(row[code]), 1) if pd.notna(row[code]) else None,
t["unit_col"]: unit_of(code),
t["source_col"]: src.ALL_INDICATORS[code].get("source", ""),
})
st.dataframe(pd.DataFrame(context_rows), hide_index=True,
**stretch(st.dataframe))
with mcol:
same = filtered[filtered["country"] == row["country"]]
d = np.sqrt((same["lat"] - row["lat"]) ** 2 + (same["lon"] - row["lon"]) ** 2)
near = same.assign(deg=d).nsmallest(8, "deg")
near = near[near["name"] != chosen]
st.markdown(f'<div class="small-muted"><b>{t["nearby"]}</b></div>',
unsafe_allow_html=True)
st.dataframe(near[["name", "area_km2", "pressure"]].round(1),
hide_index=True, **stretch(st.dataframe))
# ---------------------------------------------------------------------------
# Page: Human conditions
# ---------------------------------------------------------------------------
elif page == t["conditions"]:
section(t["conditions"], "World Bank Open Data · UN World Population Prospects 2024")
if ind_view.empty:
st.info(t["no_data"])
else:
tabs = st.tabs([f"🎓 {t['education']}", f"🩺 {t['health']}", f"💶 {t['income']}",
f"🏘️ {t['demography']}", f"⚖️ {t['dependency']}"])
groups = [("Education", src.EDUCATION_CODES), ("Health", src.HEALTH_CODES),
("Income", src.INCOME_CODES),
("Demography", src.DEMOGRAPHY_CODES),
("Dependency", src.DEPENDENCY_CODES)]
for tab, (_, codes) in zip(tabs, groups):
with tab:
for code in codes:
latest = an.latest_values(ind_view, code)
if latest.empty:
continue
ranked = latest.sort_values("value",
ascending=src.ALL_INDICATORS[code]["good"] == "low")
top = ranked.head(20).sort_values(
"value", ascending=src.ALL_INDICATORS[code]["good"] != "low")
scale = (DIVERGING if src.ALL_INDICATORS[code]["good"] == "low"
else SEQ_GREEN)
bar = px.bar(top, x="value", y="country", orientation="h",
color="value", color_continuous_scale=scale,
title=(f"{label_of(code)} ({unit_of(code)}) · "
f"{src.ALL_INDICATORS[code].get('source', '')}"),
labels={"value": "", "country": ""})
bar.update_coloraxes(showscale=False)
bar.update_traces(
hovertemplate="<b>%{y}</b><br>%{x:,.1f}<extra></extra>",
marker_line_width=0)
show(style_figure(bar, 520, theme, legend_top=False), f"cond_{code}")
if not shdi.empty:
section("Subnational HDI", shdi_status.detail)
st.dataframe(shdi.head(300), hide_index=True, **stretch(st.dataframe))
else:
st.info(
"Subnational HDI is not loaded, so this page shows country averages. "
f"Set `SHDI_CSV_URL` to switch to regions — {shdi_status.detail}"
if language == "English" else
"L'IDH infranational n'est pas chargé : cette page montre des moyennes "
f"nationales. Définissez `SHDI_CSV_URL` pour passer aux régions — {shdi_status.detail}"
)
# ---------------------------------------------------------------------------
# Page: Relationships
# ---------------------------------------------------------------------------
elif page == t["relationships"]:
section(t["typology"], t["typology_hint"])
st.markdown(f'<div class="card small-muted">{t["counterfactual_note"]}</div>',
unsafe_allow_html=True)
if ind_view.empty:
st.info(t["no_data"])
else:
c1, c2, c3 = st.columns([1, 1, 1], gap="medium")
with c1:
cons_code = st.selectbox(t["conservation_ind"], CONS_CODES,
index=min(1, len(CONS_CODES) - 1),
format_func=label_with_source, key="rel_cons")
with c2:
dev_code = st.selectbox(
t["indicator"], HD_CODES,
index=HD_CODES.index(hd_code) if hd_code in HD_CODES else 0,
format_func=label_with_source, key="rel_dev")
span_ok = an.overlapping_span(ind_view, [cons_code, dev_code])
with c3:
if span_ok is None:
st.caption(t["no_overlap"])
span = None
else:
lo, hi = span_ok
years = [y for y in sorted(ind_view["year"].unique()) if lo <= y <= hi]
# Keyed on the pair so switching indicators resets the window to
# one where both series actually have observations.
span = st.select_slider(
t["period"], options=years, value=(years[0], years[-1]),
key=f"rel_span_{cons_code}_{dev_code}")
if span is None:
st.info(
f"{t['no_overlap']} — {label_of(cons_code)}: "
f"{an.year_span(ind_view, cons_code) or t['no_data']} · "
f"{label_of(dev_code)}: {an.year_span(ind_view, dev_code) or t['no_data']}"
)
else:
result = an.build_typology(
ind_view, cons_code, dev_code, int(span[0]), int(span[1]),
development_good=src.ALL_INDICATORS[dev_code]["good"],
conservation_good=src.ALL_INDICATORS[cons_code]["good"],
)
if result.frame.empty:
st.info(t["no_data"])
else:
shares = result.shares
cards = []
for rel in an.RELATIONSHIPS:
color = an.RELATIONSHIP_COLORS[rel]
cards.append(
f'<div class="rel" style="border-left:4px solid {color}">'
f'<div class="nm" style="color:{color}">{rel}</div>'
f'<div class="pc">{shares.get(rel, 0):.0f}%</div>'
f'<div class="ds">{an.RELATIONSHIP_MEANING[rel]}<br>'
f'{t["vs_li"]}: <b>{an.LI_GLOBAL_SHARES[rel]:.1f}%</b></div></div>'
)
st.markdown(f'<div class="rel-grid">{"".join(cards)}</div>',
unsafe_allow_html=True)
assoc = an.association(result)
if assoc["n"] and not pd.isna(assoc["phi"]):
phi = assoc["phi"]
direction = (t["assoc_pos"] if phi > 0.08 else
t["assoc_neg"] if phi < -0.08 else t["assoc_none"])
st.markdown(
f'<div class="card small-muted"><b>{t["assoc_title"]}</b> — '
f'{direction} <b>φ = {phi:+.2f}</b>, '
f'{t["odds"]} {assoc["odds_ratio"]:.2f}, n = {assoc["n"]}.<br>'
f'{t["assoc_note"]}</div>',
unsafe_allow_html=True,
)
section(t["quadrant"],
f"{label_of(cons_code)} × {label_of(dev_code)}, "
f"{result.start}–{result.end}")
frame = result.frame
fig = px.scatter(
frame, x="cons_oriented", y="dev_oriented", color="relationship",
hover_name="country",
color_discrete_map=an.RELATIONSHIP_COLORS,
labels={"cons_oriented": f"{t['cons_axis']}: Δ {label_of(cons_code)}",
"dev_oriented": f"{t['dev_axis']}: Δ {label_of(dev_code)}"},
)
fig.update_traces(marker=dict(size=11, line=dict(width=1,
color=palette["panel"])),
hovertemplate="<b>%{hovertext}</b><br>"
"Δ cons: %{x:,.2f}<br>Δ dev: %{y:,.2f}<extra></extra>")
fig.add_vline(x=result.conservation_reference, line_dash="dot",
line_color=palette["muted"])
fig.add_hline(y=result.development_reference, line_dash="dot",
line_color=palette["muted"])
# Name the quadrants in place: the 2x2 should read without the legend.
corners = [
(0.99, 0.98, "right", "top", an.SYNERGY),
(0.01, 0.98, "left", "top", an.CONSERVATION_IMPAIRED),
(0.99, 0.02, "right", "bottom", an.ECONOMY_IMPAIRED),
(0.01, 0.02, "left", "bottom", an.RETREAT),
]
for x, y, xa, ya, name in corners:
fig.add_annotation(
x=x, y=y, xref="paper", yref="paper", xanchor=xa, yanchor=ya,
text=name.upper(), showarrow=False,
font=dict(size=10, color=an.RELATIONSHIP_COLORS[name]),
opacity=0.85,
)
fig.update_layout(showlegend=False)
show(style_figure(fig, 560, theme), "typology_quadrant")
st.dataframe(
frame[["country", "relationship", "change_cons", "change_dev",
"start_year_cons", "end_year_cons"]]
.rename(columns={"change_cons": f"Δ {label_of(cons_code)}",
"change_dev": f"Δ {label_of(dev_code)}",
"start_year_cons": "from", "end_year_cons": "to"})
.sort_values("relationship").round(2),
hide_index=True, height=320, **stretch(st.dataframe),
)
# --- Protected-area level, unlocked by an upload -----------------------
section(t["upload_title"])
if "effectiveness" in pas.columns and pas["effectiveness"].notna().any():
matched = pas[pas["effectiveness"].notna()]
st.success(f"{match_report.get('matched', len(matched))} / "
f"{match_report.get('uploaded', len(matched))} {t['upload_matched']}")
x_code = st.selectbox(
t["indicator"], MAP_CODES,
index=MAP_CODES.index(hd_code) if hd_code in MAP_CODES else 0,
format_func=label_with_source, key="pa_rel_x")
plot = matched.dropna(subset=[x_code, "effectiveness"])
if plot.empty:
st.info(t["no_data"])
else:
fig = px.scatter(
plot, x=x_code, y="effectiveness", hover_name="name", color="country",
size=plot["area_km2"].fillna(plot["area_km2"].median()).clip(lower=1) ** 0.3,
size_max=20,
labels={x_code: f"{label_of(x_code)} ({src.ALL_INDICATORS[x_code]['unit']})",
"effectiveness": "Effectiveness score"},
)
corr = plot[[x_code, "effectiveness"]].corr().iloc[0, 1]
fig.update_layout(title=f"r = {corr:.2f} (n = {len(plot)})")
show(style_figure(fig, 520, theme), "pa_effectiveness_vs_hd")
if match_report.get("unmatched"):
with st.expander(f"Unmatched ({len(match_report['unmatched'])})"):
st.write(", ".join(match_report["unmatched"]))
else:
st.info(t["upload_help"])
st.dataframe(
pd.DataFrame([{"column": k, "meaning": v}
for k, v in src.EFFECTIVENESS_SCHEMA.items()]),
hide_index=True, **stretch(st.dataframe),
)
# ---------------------------------------------------------------------------
# Page: Trajectories
# ---------------------------------------------------------------------------
elif page == t["trajectories"]:
section(t["trend_title"], t["framing"])
if indicators_all.empty:
st.info(t["no_data"])
else:
default = [c for c in ["Kenya", "Tanzania", "Ethiopia", "Senegal", "Zambia"]
if c in COUNTRIES][:5]
picked = st.multiselect(t["pick_countries"], COUNTRIES,
default=default or COUNTRIES[:4], key="traj_countries")
codes = st.multiselect(t["indicator"], MAP_CODES + CONS_CODES,
default=[hd_code, "ER.LND.PTLD.ZS"],
format_func=label_with_source, key="traj_codes")
if not picked or not codes:
st.info(t["no_data"])
else:
for code in codes:
d = indicators_all[(indicators_all["indicator"] == code)
& (indicators_all["country"].isin(picked))]
if d.empty:
continue
d = d.sort_values("year")
has_projection = ("series" in d.columns
and (d["series"] == "Projection").any())
fig = go.Figure()
for i, c in enumerate(picked):
sub = d[d["country"] == c]
if sub.empty:
continue
colour = CATEGORICAL[i % len(CATEGORICAL)]
if has_projection:
obs = sub[sub["series"] != "Projection"]
proj = sub[sub["series"] == "Projection"]
else:
obs, proj = sub, sub.iloc[0:0]
if not obs.empty:
fig.add_trace(go.Scatter(
x=obs["year"], y=obs["value"], mode="lines", name=c,
legendgroup=c, line=dict(color=colour, width=2.6),
hovertemplate=f"<b>{c}</b><br>%{{x}} · %{{y:,.1f}}<extra></extra>"))
if not proj.empty:
bridge = pd.concat([obs.tail(1), proj]) if not obs.empty else proj
fig.add_trace(go.Scatter(
x=bridge["year"], y=bridge["value"], mode="lines", name=c,
legendgroup=c, showlegend=obs.empty,
line=dict(color=colour, width=2.2, dash="dash"),
hovertemplate=(f"<b>{c}</b> ({t['projection']})"
"<br>%{x} · %{y:,.1f}<extra></extra>")))
fig.update_layout(
title=f"{label_of(code)} ({unit_of(code)}) · "
f"{src.ALL_INDICATORS[code].get('source', '')}",
hovermode="x unified", xaxis_title="", yaxis_title="")
fig = style_figure(fig, 420, theme)
if has_projection:
fig.add_vline(x=src.UN_LAST_ESTIMATE_YEAR + 0.5, line_dash="dot",
line_color=palette["muted"], line_width=1.4)
fig.add_annotation(x=src.UN_LAST_ESTIMATE_YEAR + 1, y=1.0, yref="paper",
yanchor="bottom", xanchor="left",
text=t["projection"], showarrow=False,
font=dict(color=palette["muted"], size=10))
show(fig, f"traj_{code}")
# ---------------------------------------------------------------------------
# Page: Data & method
# ---------------------------------------------------------------------------
elif page == t["data"]:
section(t["status"])
with st.spinner(t["loading_map"]):
_, bnd_status = load_boundaries(0)
for s in statuses + [bnd_status]:
dot = "ok" if (s.ok and not s.from_cache) else ("cache" if s.ok else "bad")
st.markdown(
f'<div class="status-row"><span class="dot {dot}"></span>'
f'<span class="nm">{s.name}</span>'
f'<span class="dt">{s.badge} · {s.rows:,} {t["rows"].lower()} · {s.detail}</span>'
f'<span class="dt">{s.licence}</span></div>',
unsafe_allow_html=True,
)
section(t["coverage"], t["coverage_hint"])
cov = an.coverage(indicators_all)
if cov.empty:
st.info(t["no_data"])
else:
cov = cov.assign(
Indicator=[label_of(c) for c in cov["indicator"]],
Source=[src.ALL_INDICATORS.get(c, {}).get("source", "") for c in cov["indicator"]],
)[["indicator", "Indicator", "Source", "countries", "first", "last", "points"]]
st.dataframe(
cov.rename(columns={"indicator": "Code", "countries": t["countries"],
"first": t["first_year"], "last": t["last_year"],
"points": t["observations"]}),
hide_index=True, height=300, **stretch(st.dataframe),
)
section("Method")
with st.expander("Sources and licences", expanded=True):
rows = []
for code, meta in src.ALL_INDICATORS.items():
rows.append({"Code": code, "Indicator": meta["label"],
"Unit": meta["unit"], "Group": meta["group"],
"Source": meta.get("source", ""),
"Direction": {"high": "higher is better",
"low": "lower is better"}.get(meta.get("good"), "context")})
st.dataframe(pd.DataFrame(rows), hide_index=True, height=330,
**stretch(st.dataframe))
st.markdown(
"""
- **Protected areas** — Wikidata (CC0) by default: name, country, coordinates, reported
area, year established, IUCN category. Set `PROTECTED_PLANET_TOKEN` in the Space
secrets to pull the authoritative WDPA from Protected Planet instead.
- **Education, health, income, conservation cover** — World Bank Open Data (CC BY 4.0),
1990–2024.
- **Demography and dependency** — UN Population Division, *World Population Prospects
2024*, median variant. Read from `unpopulation_data*.csv` committed at the repo root:
the same file, the same indicator ids and therefore the same numbers as the Africa
Population Atlas. Set `UN_WPP_CSV` to point elsewhere.
- **Subnational HDI** — Global Data Lab. No stable open endpoint; set `SHDI_CSV_URL`
to a direct link or a file in the repo to move from country averages to regions.
- **Boundaries** — geoBoundaries (CC BY 4.0).
- **Effectiveness scores** — supplied by you, e.g. Canopy PACE.
"""
)
with st.expander("Rules this dashboard keeps", expanded=True):
st.markdown(
f"""
1. **Provenance is never inferred.** Every indicator carries its source, and every
selector prints it, so a World Bank series and a UN series are never mistaken for
one another in the same chart.
2. **Projections never enter an analysis.** WPP 2024 runs to 2100; everything after
**{src.UN_LAST_ESTIMATE_YEAR}** is a median-variant projection. Those rows are
filtered out of the relationship typology, the change calculations, the map and the
protected-area profiles. They appear only on *Trajectories*, drawn dashed and behind
a marked line.
3. **One variant only.** Only the WPP median variant is loaded; high, low and
probabilistic variants are dropped at read time.
4. **The reference is stated, not hidden.** The 2×2 uses the African median change over
the selected period, because matched counterfactuals are not reconstructible from
open country series. That is a benchmark, not a causal claim.
5. **Missing is missing.** A connector that fails leaves its features disabled and says
so on this page. Nothing is filled in with modelled or invented values.
"""
)
with st.expander("Known limits"):
st.markdown(
"""
- The pressure index is built from **national** density, child dependency, rural share and
protected-area clustering. It is a screening proxy, not a count of people living
within 10 km of a park. Doing that properly needs gridded population (WorldPop or GHSL)
clipped to a buffer around each boundary polygon — the natural next step, and the reason
synthetic or modelled population surfaces are worth looking at.
- Country averages hide exactly the subnational variation the question is about. The
subnational HDI connector is the single highest-value thing to switch on.
- Wikidata coverage of protected areas is uneven between countries; WDPA via a Protected
Planet token is the authoritative alternative.
- No nightlight or land-cover change series is used here, so this dashboard describes
**conditions and co-movement**, not the causal park effect Li et al. estimate.
"""
)
section(t["detailed"], t["detailed_hint"])
for block in methods.sections(language):
with st.expander(f"{block['icon']} {block['title']}", expanded=False):
st.markdown(block["body"])
with st.expander("Citation"):
st.markdown(
"""
Li, B.V., Wu, S., Pimm, S.L. & Cui, J. (2024). The synergy between protected area
effectiveness and economic growth. *Current Biology* 34, 1–14.
https://doi.org/10.1016/j.cub.2024.05.044
Dashboard: Diouf, I. (2026). *Conservation & Human Development in Africa*.
"""
)
# ---------------------------------------------------------------------------
# Signature
# ---------------------------------------------------------------------------
st.markdown(
f"""
<div class="signature">
<div class="sig-name">{AUTHOR['name']}</div>
<div class="sig-affil">{AUTHOR['affiliation']}</div>
<div class="sig-links">
{''.join(f'<a href="{u}" target="_blank" rel="noopener">'
f'<span class="ic">{i}</span>{n}</a>'
for i, n, u in AUTHOR['links'])}
</div>
<div class="sig-meta">
© 2026 · <b>Conservation & Human Development in Africa</b><br>
Wikidata · World Bank Open Data · geoBoundaries · framed on Li et al. (2024)
</div>
</div>
""",
unsafe_allow_html=True,
)
"""Africa Conservation & Human Development — an exploratory dashboard.
Built around the question Liesbet Steer put after Li et al. (2024): do the human
conditions around protected areas explain whether conservation and local
development pull together or against each other?
"""
from __future__ import annotations
import inspect
import os
from string import Template
import numpy as np
import pandas as pd
import plotly.express as px
import plotly.graph_objects as go
import streamlit as st
import streamlit.components.v1 as components
import analysis as an
import methods
import sources as src
st.set_page_config(
page_title="Africa Conservation & Human Development",
page_icon="🌍",
layout="wide",
initial_sidebar_state="auto",
menu_items={
"About": (
"Exploratory dashboard on protected areas and the human conditions "
"around them in Africa. Framed on Li et al. (2024), *The synergy "
"between protected area effectiveness and economic growth*, "
"Current Biology 34."
)
},
)
AUTHOR = {
"name": "Dr. Ibrahima Diouf",
"affiliation": "UCAD–ESP/LPAO-SF · Université de Labé–FST",
"links": [
("🌐", "Website", "https://scientific-hub-lab.github.io/Dr-Ibrahima-Diouf/"),
("🤗", "Hugging Face", "https://huggingface.co/idiouf"),
("🆔", "ORCID", "https://orcid.org/0000-0002-3157-0925"),
("in", "LinkedIn", "https://www.linkedin.com/in/dr-ibrahima-diouf-610872108/"),
("🎓", "Google Scholar",
"https://scholar.google.com/citations?user=phsMm7sAAAAJ&hl=fr"),
("🔗", "This Space",
"https://huggingface.co/spaces/idiouf/"
"African-Conservation-Human-Development-Observatory"),
],
}
SEQ_BLUE = ["#EEF4FB", "#D2E3F4", "#AACBE9", "#7FB0DC", "#5493CB",
"#2F76B4", "#1B5B95", "#114374", "#0A2C51"]
SEQ_GREEN = ["#F0F7F2", "#D3E9DA", "#A9D5BA", "#7CBE99", "#51A57A",
"#33885E", "#216A48", "#154E34", "#0C3322"]
DIVERGING = ["#8C2D3F", "#C05B5F", "#E39A88", "#F5D0BC", "#F2F2EF",
"#C4DCE4", "#8CBCD1", "#4E90B5", "#1F5F8F"]
CATEGORICAL = ["#2F6FED", "#0E9F9B", "#E8873D", "#A64CB0", "#3FA34D",
"#D6455D", "#7B6FE8", "#B08A24", "#3FA3D9", "#7A8896"]
PALETTES = {
"Dark": {
"bg": "#070F1A", "panel": "#0D1E30", "panel2": "#122840", "border": "#1D3A55",
"border_soft": "#162E45", "text": "#F2F7FC", "muted": "#9AB0C6",
"accent": "#4A9BFF", "accent_soft": "rgba(74,155,255,.15)",
"positive": "#3FBF8F", "negative": "#F2778D",
"shadow": "0 10px 30px rgba(0,0,0,.35)",
"hero1": "#0B2C52", "hero2": "#14503C", "hero3": "#5C2350",
"map_style": "carto-darkmatter",
},
"Light": {
"bg": "#F2F6FA", "panel": "#FFFFFF", "panel2": "#F6F9FD", "border": "#D8E3EF",
"border_soft": "#E6EDF5", "text": "#0F2439", "muted": "#54687E",
"accent": "#1668CE", "accent_soft": "rgba(22,104,206,.10)",
"positive": "#12805C", "negative": "#B93049",
"shadow": "0 8px 24px rgba(15,36,57,.08)",
"hero1": "#0B3A69", "hero2": "#1B6B4F", "hero3": "#7A2359",
"map_style": "carto-positron",
},
}
# ---------------------------------------------------------------------------
# Translations
# ---------------------------------------------------------------------------
TEXT = {
"English": {
"title": "Conservation & Human Development in Africa",
"subtitle": ("Where protected areas sit, what human conditions surround them, "
"and whether the two move together."),
"framing": ("Framed on Li et al. (2024), Current Biology — protected areas and "
"neighbouring communities fall into four relationships: synergy, "
"economy impaired, conservation impaired, retreat."),
"nav": "Sections",
"overview": "Overview", "areas": "Protected areas", "conditions": "Human conditions",