Dataset Viewer
Auto-converted to Parquet Duplicate
grid_version
stringclasses
1 value
region_id
stringlengths
12
12
planet
stringclasses
1 value
band
int32
0
4k
column
int32
0
8.01k
lat_min
float64
-90
90
lat_max
float64
-89.96
90
lon_min
float64
-180
180
lon_max
float64
-179.96
180
center_lat
float64
-89.98
90
center_lon
float64
-179.98
180
area_km2
float64
24.5
26.2
elev_raster_px_per_deg_lon
int32
360
3.6k
elev_raster_row_start
int32
0
648k
elev_raster_row_stop
int32
162
648k
elev_raster_col_start
int32
0
1.3M
elev_raster_col_stop
int32
156
1.3M
elev_spans_two_zones
bool
2 classes
elev_south_px_per_deg_lon
int32
360
3.6k
⌀
elev_south_row_start
int32
18k
630k
⌀
elev_south_row_stop
int32
18k
630k
⌀
elev_south_col_start
int32
0
1.3M
⌀
elev_south_col_stop
int32
156
1.3M
⌀
latband5km-v1
E_0000_00000
earth
0
0
-90
-89.955237
-180
-60
-89.977619
-120
26.177141
360
647,839
648,000
0
42,692
false
null
null
null
null
null
latband5km-v1
E_0000_00001
earth
0
1
-90
-89.955237
-60
60
-89.977619
0
26.177141
360
647,839
648,000
42,692
87,026
false
null
null
null
null
null
latband5km-v1
E_0000_00002
earth
0
2
-90
-89.955237
60
180
-89.977619
120
26.177141
360
647,839
648,000
87,026
129,600
false
null
null
null
null
null
latband5km-v1
E_0001_00000
earth
1
0
-89.955237
-89.910474
-180
-140
-89.932856
-160
26.177135
360
647,678
647,839
0
14,274
false
null
null
null
null
null
latband5km-v1
E_0001_00001
earth
1
1
-89.955237
-89.910474
-140
-100
-89.932856
-120
26.177135
360
647,678
647,839
14,274
28,548
false
null
null
null
null
null
latband5km-v1
E_0001_00002
earth
1
2
-89.955237
-89.910474
-100
-60
-89.932856
-80
26.177135
360
647,678
647,839
28,548
43,371
false
null
null
null
null
null
latband5km-v1
E_0001_00003
earth
1
3
-89.955237
-89.910474
-60
-20
-89.932856
-40
26.177135
360
647,678
647,839
43,371
57,645
false
null
null
null
null
null
latband5km-v1
E_0001_00004
earth
1
4
-89.955237
-89.910474
-20
20
-89.932856
0
26.177135
360
647,678
647,839
57,645
71,919
false
null
null
null
null
null
latband5km-v1
E_0001_00005
earth
1
5
-89.955237
-89.910474
20
60
-89.932856
40
26.177135
360
647,678
647,839
71,919
86,193
false
null
null
null
null
null
latband5km-v1
E_0001_00006
earth
1
6
-89.955237
-89.910474
60
100
-89.932856
80
26.177135
360
647,678
647,839
86,193
101,016
false
null
null
null
null
null
latband5km-v1
E_0001_00007
earth
1
7
-89.955237
-89.910474
100
140
-89.932856
120
26.177135
360
647,678
647,839
101,016
115,290
false
null
null
null
null
null
latband5km-v1
E_0001_00008
earth
1
8
-89.955237
-89.910474
140
180
-89.932856
160
26.177135
360
647,678
647,839
115,290
129,600
false
null
null
null
null
null
latband5km-v1
E_0002_00000
earth
2
0
-89.910474
-89.865712
-180
-157.5
-89.888093
-168.75
24.541055
360
647,517
647,678
0
8,250
false
null
null
null
null
null
latband5km-v1
E_0002_00001
earth
2
1
-89.910474
-89.865712
-157.5
-135
-89.888093
-146.25
24.541055
360
647,517
647,678
8,250
16,170
false
null
null
null
null
null
latband5km-v1
E_0002_00002
earth
2
2
-89.910474
-89.865712
-135
-112.5
-89.888093
-123.75
24.541055
360
647,517
647,678
16,170
24,420
false
null
null
null
null
null
latband5km-v1
E_0002_00003
earth
2
3
-89.910474
-89.865712
-112.5
-90
-89.888093
-101.25
24.541055
360
647,517
647,678
24,420
32,340
false
null
null
null
null
null
latband5km-v1
E_0002_00004
earth
2
4
-89.910474
-89.865712
-90
-67.5
-89.888093
-78.75
24.541055
360
647,517
647,678
32,340
40,590
false
null
null
null
null
null
latband5km-v1
E_0002_00005
earth
2
5
-89.910474
-89.865712
-67.5
-45
-89.888093
-56.25
24.541055
360
647,517
647,678
40,590
48,510
false
null
null
null
null
null
latband5km-v1
E_0002_00006
earth
2
6
-89.910474
-89.865712
-45
-22.5
-89.888093
-33.75
24.541055
360
647,517
647,678
48,510
56,760
false
null
null
null
null
null
latband5km-v1
E_0002_00007
earth
2
7
-89.910474
-89.865712
-22.5
0
-89.888093
-11.25
24.541055
360
647,517
647,678
56,760
64,680
false
null
null
null
null
null
latband5km-v1
E_0002_00008
earth
2
8
-89.910474
-89.865712
0
22.5
-89.888093
11.25
24.541055
360
647,517
647,678
64,680
72,930
false
null
null
null
null
null
latband5km-v1
E_0002_00009
earth
2
9
-89.910474
-89.865712
22.5
45
-89.888093
33.75
24.541055
360
647,517
647,678
72,930
80,850
false
null
null
null
null
null
latband5km-v1
E_0002_00010
earth
2
10
-89.910474
-89.865712
45
67.5
-89.888093
56.25
24.541055
360
647,517
647,678
80,850
89,100
false
null
null
null
null
null
latband5km-v1
E_0002_00011
earth
2
11
-89.910474
-89.865712
67.5
90
-89.888093
78.75
24.541055
360
647,517
647,678
89,100
97,350
false
null
null
null
null
null
latband5km-v1
E_0002_00012
earth
2
12
-89.910474
-89.865712
90
112.5
-89.888093
101.25
24.541055
360
647,517
647,678
97,350
105,270
false
null
null
null
null
null
latband5km-v1
E_0002_00013
earth
2
13
-89.910474
-89.865712
112.5
135
-89.888093
123.75
24.541055
360
647,517
647,678
105,270
113,520
false
null
null
null
null
null
latband5km-v1
E_0002_00014
earth
2
14
-89.910474
-89.865712
135
157.5
-89.888093
146.25
24.541055
360
647,517
647,678
113,520
121,440
false
null
null
null
null
null
latband5km-v1
E_0002_00015
earth
2
15
-89.910474
-89.865712
157.5
180
-89.888093
168.75
24.541055
360
647,517
647,678
121,440
129,600
false
null
null
null
null
null
latband5km-v1
E_0003_00000
earth
3
0
-89.865712
-89.820949
-180
-163.636364
-89.84333
-171.818182
24.98724
360
647,356
647,517
0
5,900
false
null
null
null
null
null
latband5km-v1
E_0003_00001
earth
3
1
-89.865712
-89.820949
-163.636364
-147.272727
-89.84333
-155.454545
24.98724
360
647,356
647,517
5,900
11,800
false
null
null
null
null
null
latband5km-v1
E_0003_00002
earth
3
2
-89.865712
-89.820949
-147.272727
-130.909091
-89.84333
-139.090909
24.98724
360
647,356
647,517
11,800
17,700
false
null
null
null
null
null
latband5km-v1
E_0003_00003
earth
3
3
-89.865712
-89.820949
-130.909091
-114.545455
-89.84333
-122.727273
24.98724
360
647,356
647,517
17,700
23,600
false
null
null
null
null
null
latband5km-v1
E_0003_00004
earth
3
4
-89.865712
-89.820949
-114.545455
-98.181818
-89.84333
-106.363636
24.98724
360
647,356
647,517
23,600
29,500
false
null
null
null
null
null
latband5km-v1
E_0003_00005
earth
3
5
-89.865712
-89.820949
-98.181818
-81.818182
-89.84333
-90
24.98724
360
647,356
647,517
29,500
35,400
false
null
null
null
null
null
latband5km-v1
E_0003_00006
earth
3
6
-89.865712
-89.820949
-81.818182
-65.454545
-89.84333
-73.636364
24.98724
360
647,356
647,517
35,400
41,300
false
null
null
null
null
null
latband5km-v1
E_0003_00007
earth
3
7
-89.865712
-89.820949
-65.454545
-49.090909
-89.84333
-57.272727
24.98724
360
647,356
647,517
41,300
47,200
false
null
null
null
null
null
latband5km-v1
E_0003_00008
earth
3
8
-89.865712
-89.820949
-49.090909
-32.727273
-89.84333
-40.909091
24.98724
360
647,356
647,517
47,200
53,100
false
null
null
null
null
null
latband5km-v1
E_0003_00009
earth
3
9
-89.865712
-89.820949
-32.727273
-16.363636
-89.84333
-24.545455
24.98724
360
647,356
647,517
53,100
59,000
false
null
null
null
null
null
latband5km-v1
E_0003_00010
earth
3
10
-89.865712
-89.820949
-16.363636
0
-89.84333
-8.181818
24.98724
360
647,356
647,517
59,000
64,900
false
null
null
null
null
null
latband5km-v1
E_0003_00011
earth
3
11
-89.865712
-89.820949
0
16.363636
-89.84333
8.181818
24.98724
360
647,356
647,517
64,900
70,800
false
null
null
null
null
null
latband5km-v1
E_0003_00012
earth
3
12
-89.865712
-89.820949
16.363636
32.727273
-89.84333
24.545455
24.98724
360
647,356
647,517
70,800
76,700
false
null
null
null
null
null
latband5km-v1
E_0003_00013
earth
3
13
-89.865712
-89.820949
32.727273
49.090909
-89.84333
40.909091
24.98724
360
647,356
647,517
76,700
82,364
false
null
null
null
null
null
latband5km-v1
E_0003_00014
earth
3
14
-89.865712
-89.820949
49.090909
65.454545
-89.84333
57.272727
24.98724
360
647,356
647,517
82,364
88,264
false
null
null
null
null
null
latband5km-v1
E_0003_00015
earth
3
15
-89.865712
-89.820949
65.454545
81.818182
-89.84333
73.636364
24.98724
360
647,356
647,517
88,264
94,164
false
null
null
null
null
null
latband5km-v1
E_0003_00016
earth
3
16
-89.865712
-89.820949
81.818182
98.181818
-89.84333
90
24.98724
360
647,356
647,517
94,164
100,064
false
null
null
null
null
null
latband5km-v1
E_0003_00017
earth
3
17
-89.865712
-89.820949
98.181818
114.545455
-89.84333
106.363636
24.98724
360
647,356
647,517
100,064
105,964
false
null
null
null
null
null
latband5km-v1
E_0003_00018
earth
3
18
-89.865712
-89.820949
114.545455
130.909091
-89.84333
122.727273
24.98724
360
647,356
647,517
105,964
111,864
false
null
null
null
null
null
latband5km-v1
E_0003_00019
earth
3
19
-89.865712
-89.820949
130.909091
147.272727
-89.84333
139.090909
24.98724
360
647,356
647,517
111,864
117,764
false
null
null
null
null
null
latband5km-v1
E_0003_00020
earth
3
20
-89.865712
-89.820949
147.272727
163.636364
-89.84333
155.454545
24.98724
360
647,356
647,517
117,764
123,664
false
null
null
null
null
null
latband5km-v1
E_0003_00021
earth
3
21
-89.865712
-89.820949
163.636364
180
-89.84333
171.818182
24.98724
360
647,356
647,517
123,664
129,600
false
null
null
null
null
null
latband5km-v1
E_0004_00000
earth
4
0
-89.820949
-89.776186
-180
-167.142857
-89.798567
-173.571429
25.242192
360
647,195
647,356
0
4,575
false
null
null
null
null
null
latband5km-v1
E_0004_00001
earth
4
1
-89.820949
-89.776186
-167.142857
-154.285714
-89.798567
-160.714286
25.242192
360
647,195
647,356
4,575
9,333
false
null
null
null
null
null
latband5km-v1
E_0004_00002
earth
4
2
-89.820949
-89.776186
-154.285714
-141.428571
-89.798567
-147.857143
25.242192
360
647,195
647,356
9,333
13,908
false
null
null
null
null
null
latband5km-v1
E_0004_00003
earth
4
3
-89.820949
-89.776186
-141.428571
-128.571429
-89.798567
-135
25.242192
360
647,195
647,356
13,908
18,483
false
null
null
null
null
null
latband5km-v1
E_0004_00004
earth
4
4
-89.820949
-89.776186
-128.571429
-115.714286
-89.798567
-122.142857
25.242192
360
647,195
647,356
18,483
23,058
false
null
null
null
null
null
latband5km-v1
E_0004_00005
earth
4
5
-89.820949
-89.776186
-115.714286
-102.857143
-89.798567
-109.285714
25.242192
360
647,195
647,356
23,058
27,816
false
null
null
null
null
null
latband5km-v1
E_0004_00006
earth
4
6
-89.820949
-89.776186
-102.857143
-90
-89.798567
-96.428571
25.242192
360
647,195
647,356
27,816
32,391
false
null
null
null
null
null
latband5km-v1
E_0004_00007
earth
4
7
-89.820949
-89.776186
-90
-77.142857
-89.798567
-83.571429
25.242192
360
647,195
647,356
32,391
36,966
false
null
null
null
null
null
latband5km-v1
E_0004_00008
earth
4
8
-89.820949
-89.776186
-77.142857
-64.285714
-89.798567
-70.714286
25.242192
360
647,195
647,356
36,966
41,724
false
null
null
null
null
null
latband5km-v1
E_0004_00009
earth
4
9
-89.820949
-89.776186
-64.285714
-51.428571
-89.798567
-57.857143
25.242192
360
647,195
647,356
41,724
46,299
false
null
null
null
null
null
latband5km-v1
E_0004_00010
earth
4
10
-89.820949
-89.776186
-51.428571
-38.571429
-89.798567
-45
25.242192
360
647,195
647,356
46,299
50,874
false
null
null
null
null
null
latband5km-v1
E_0004_00011
earth
4
11
-89.820949
-89.776186
-38.571429
-25.714286
-89.798567
-32.142857
25.242192
360
647,195
647,356
50,874
55,632
false
null
null
null
null
null
latband5km-v1
E_0004_00012
earth
4
12
-89.820949
-89.776186
-25.714286
-12.857143
-89.798567
-19.285714
25.242192
360
647,195
647,356
55,632
60,207
false
null
null
null
null
null
latband5km-v1
E_0004_00013
earth
4
13
-89.820949
-89.776186
-12.857143
0
-89.798567
-6.428571
25.242192
360
647,195
647,356
60,207
64,782
false
null
null
null
null
null
latband5km-v1
E_0004_00014
earth
4
14
-89.820949
-89.776186
0
12.857143
-89.798567
6.428571
25.242192
360
647,195
647,356
64,782
69,357
false
null
null
null
null
null
latband5km-v1
E_0004_00015
earth
4
15
-89.820949
-89.776186
12.857143
25.714286
-89.798567
19.285714
25.242192
360
647,195
647,356
69,357
74,115
false
null
null
null
null
null
latband5km-v1
E_0004_00016
earth
4
16
-89.820949
-89.776186
25.714286
38.571429
-89.798567
32.142857
25.242192
360
647,195
647,356
74,115
78,690
false
null
null
null
null
null
latband5km-v1
E_0004_00017
earth
4
17
-89.820949
-89.776186
38.571429
51.428571
-89.798567
45
25.242192
360
647,195
647,356
78,690
83,265
false
null
null
null
null
null
latband5km-v1
E_0004_00018
earth
4
18
-89.820949
-89.776186
51.428571
64.285714
-89.798567
57.857143
25.242192
360
647,195
647,356
83,265
88,023
false
null
null
null
null
null
latband5km-v1
E_0004_00019
earth
4
19
-89.820949
-89.776186
64.285714
77.142857
-89.798567
70.714286
25.242192
360
647,195
647,356
88,023
92,598
false
null
null
null
null
null
latband5km-v1
E_0004_00020
earth
4
20
-89.820949
-89.776186
77.142857
90
-89.798567
83.571429
25.242192
360
647,195
647,356
92,598
97,173
false
null
null
null
null
null
latband5km-v1
E_0004_00021
earth
4
21
-89.820949
-89.776186
90
102.857143
-89.798567
96.428571
25.242192
360
647,195
647,356
97,173
101,748
false
null
null
null
null
null
latband5km-v1
E_0004_00022
earth
4
22
-89.820949
-89.776186
102.857143
115.714286
-89.798567
109.285714
25.242192
360
647,195
647,356
101,748
106,506
false
null
null
null
null
null
latband5km-v1
E_0004_00023
earth
4
23
-89.820949
-89.776186
115.714286
128.571429
-89.798567
122.142857
25.242192
360
647,195
647,356
106,506
111,081
false
null
null
null
null
null
latband5km-v1
E_0004_00024
earth
4
24
-89.820949
-89.776186
128.571429
141.428571
-89.798567
135
25.242192
360
647,195
647,356
111,081
115,656
false
null
null
null
null
null
latband5km-v1
E_0004_00025
earth
4
25
-89.820949
-89.776186
141.428571
154.285714
-89.798567
147.857143
25.242192
360
647,195
647,356
115,656
120,414
false
null
null
null
null
null
latband5km-v1
E_0004_00026
earth
4
26
-89.820949
-89.776186
154.285714
167.142857
-89.798567
160.714286
25.242192
360
647,195
647,356
120,414
124,989
false
null
null
null
null
null
latband5km-v1
E_0004_00027
earth
4
27
-89.820949
-89.776186
167.142857
180
-89.798567
173.571429
25.242192
360
647,195
647,356
124,989
129,600
false
null
null
null
null
null
latband5km-v1
E_0005_00000
earth
5
0
-89.776186
-89.731423
-180
-169.714286
-89.753805
-174.857143
24.681229
360
647,034
647,195
0
3,750
false
null
null
null
null
null
latband5km-v1
E_0005_00001
earth
5
1
-89.776186
-89.731423
-169.714286
-159.428571
-89.753805
-164.571429
24.681229
360
647,034
647,195
3,750
7,350
false
null
null
null
null
null
latband5km-v1
E_0005_00002
earth
5
2
-89.776186
-89.731423
-159.428571
-149.142857
-89.753805
-154.285714
24.681229
360
647,034
647,195
7,350
11,100
false
null
null
null
null
null
latband5km-v1
E_0005_00003
earth
5
3
-89.776186
-89.731423
-149.142857
-138.857143
-89.753805
-144
24.681229
360
647,034
647,195
11,100
14,850
false
null
null
null
null
null
latband5km-v1
E_0005_00004
earth
5
4
-89.776186
-89.731423
-138.857143
-128.571429
-89.753805
-133.714286
24.681229
360
647,034
647,195
14,850
18,450
false
null
null
null
null
null
latband5km-v1
E_0005_00005
earth
5
5
-89.776186
-89.731423
-128.571429
-118.285714
-89.753805
-123.428571
24.681229
360
647,034
647,195
18,450
22,200
false
null
null
null
null
null
latband5km-v1
E_0005_00006
earth
5
6
-89.776186
-89.731423
-118.285714
-108
-89.753805
-113.142857
24.681229
360
647,034
647,195
22,200
25,950
false
null
null
null
null
null
latband5km-v1
E_0005_00007
earth
5
7
-89.776186
-89.731423
-108
-97.714286
-89.753805
-102.857143
24.681229
360
647,034
647,195
25,950
29,550
false
null
null
null
null
null
latband5km-v1
E_0005_00008
earth
5
8
-89.776186
-89.731423
-97.714286
-87.428571
-89.753805
-92.571429
24.681229
360
647,034
647,195
29,550
33,300
false
null
null
null
null
null
latband5km-v1
E_0005_00009
earth
5
9
-89.776186
-89.731423
-87.428571
-77.142857
-89.753805
-82.285714
24.681229
360
647,034
647,195
33,300
37,050
false
null
null
null
null
null
latband5km-v1
E_0005_00010
earth
5
10
-89.776186
-89.731423
-77.142857
-66.857143
-89.753805
-72
24.681229
360
647,034
647,195
37,050
40,800
false
null
null
null
null
null
latband5km-v1
E_0005_00011
earth
5
11
-89.776186
-89.731423
-66.857143
-56.571429
-89.753805
-61.714286
24.681229
360
647,034
647,195
40,800
44,400
false
null
null
null
null
null
latband5km-v1
E_0005_00012
earth
5
12
-89.776186
-89.731423
-56.571429
-46.285714
-89.753805
-51.428571
24.681229
360
647,034
647,195
44,400
48,150
false
null
null
null
null
null
latband5km-v1
E_0005_00013
earth
5
13
-89.776186
-89.731423
-46.285714
-36
-89.753805
-41.142857
24.681229
360
647,034
647,195
48,150
51,900
false
null
null
null
null
null
latband5km-v1
E_0005_00014
earth
5
14
-89.776186
-89.731423
-36
-25.714286
-89.753805
-30.857143
24.681229
360
647,034
647,195
51,900
55,500
false
null
null
null
null
null
latband5km-v1
E_0005_00015
earth
5
15
-89.776186
-89.731423
-25.714286
-15.428571
-89.753805
-20.571429
24.681229
360
647,034
647,195
55,500
59,250
false
null
null
null
null
null
latband5km-v1
E_0005_00016
earth
5
16
-89.776186
-89.731423
-15.428571
-5.142857
-89.753805
-10.285714
24.681229
360
647,034
647,195
59,250
63,000
false
null
null
null
null
null
latband5km-v1
E_0005_00017
earth
5
17
-89.776186
-89.731423
-5.142857
5.142857
-89.753805
0
24.681229
360
647,034
647,195
63,000
66,600
false
null
null
null
null
null
latband5km-v1
E_0005_00018
earth
5
18
-89.776186
-89.731423
5.142857
15.428571
-89.753805
10.285714
24.681229
360
647,034
647,195
66,600
70,350
false
null
null
null
null
null
latband5km-v1
E_0005_00019
earth
5
19
-89.776186
-89.731423
15.428571
25.714286
-89.753805
20.571429
24.681229
360
647,034
647,195
70,350
74,100
false
null
null
null
null
null
latband5km-v1
E_0005_00020
earth
5
20
-89.776186
-89.731423
25.714286
36
-89.753805
30.857143
24.681229
360
647,034
647,195
74,100
77,700
false
null
null
null
null
null
latband5km-v1
E_0005_00021
earth
5
21
-89.776186
-89.731423
36
46.285714
-89.753805
41.142857
24.681229
360
647,034
647,195
77,700
81,450
false
null
null
null
null
null
End of preview. Expand in Data Studio

Starman - terrain features on a permanent 5 km grid

This repository holds per-cell terrain statistics on a permanent, versioned latitude-band grid (latband5km-v1) for Mars and Earth. This card describes the Earth release; the Mars files are unchanged and documented by manifests/terrain_mars_v0.2.json (Mars releases are tagged terrain-mars-v*).

Earth terrain

Per-cell terrain statistics for Earth (20,403,725 cells), computed from the Copernicus DEM GLO-30 Public and indexed by the same grid. Tables join on grid_version + region_id.

Release: tag terrain-earth-v0.2.0 (grid latband5km-v1, features terrain-v0.2). Pin it with revision="terrain-earth-v0.2.0".

Attribution (required)

These tables are derived from, and are modifications of, Copernicus DEM GLO-30. The licence's notice for adapted or modified products applies, and reads exactly:

"produced using Copernicus WorldDEM-30 © DLR e.V. 2010-2014 and © Airbus Defence and Space GmbH 2014-2018 provided under COPERNICUS by the European Union and ESA; all rights reserved".

Licence

Derived elevation products count as 'adapted or modified' so the second notice applies. Read the full EULA before redistribution.

The licence text is the EULA shipped with every source tile (INFO/eula_F.pdf in each tile folder of the copernicus-dem-30m bucket) and linked from the product's readme.html. Read it before redistributing. The bucket readme states: "GLO-30 Public and GLO-90 are available on a free basis for the general public under the terms and conditions of the Licence found [at spacedata.copernicus.eu]."

Contents

Path What
grid/planet=earth/latband5km-v1.parquet One row per Earth cell: id, bounds, centre, area
grid/planet=earth/latband5km-v1_elevation_windows.parquet The same cells plus the GLO-30 elevation samples behind each one: which source samples to read
grid/earth_global_bands.csv Band catalog: latitude edges, columns per band, area per cell
terrain/planet=earth/feature_version=terrain-v0.2/band_group=NNN.parquet Terrain features, one file per 100 bands (41 files)
manifests/terrain_earth_v0.2.json Full build manifest: source, provenance, validation, limitations
code/earth_terrain.py, code/terrain_features.py The code that produced the data, at git commit d968501

Quick start

from datasets import load_dataset

grid = load_dataset("enzeeoai/starman", "grid_earth", split="train", revision="terrain-earth-v0.2.0")
terrain = load_dataset("enzeeoai/starman", "terrain_earth", split="train", revision="terrain-earth-v0.2.0")
windows = load_dataset("enzeeoai/starman", "elevation_windows_earth", split="train", revision="terrain-earth-v0.2.0")

Joining other datasets

Every row is keyed by grid_version + region_id. Earth region_id is E_<band:04d>_<column:05d> and is permanent within a grid_version. Never join across grid versions. Earth bands are on the WGS84 ellipsoid with geodetic latitude; Mars uses its own spherical indexing reference (see the grid manifest).

Terrain table columns (terrain_earth)

Identity and provenance columns match the Mars table (grid_version, region_id, planet, band, column, quality_flags, source_product, source_version, source_checksum, sample_spacing_m, feature_version, code_version, computed_at). For Earth, source_checksum is the SHA-256 of the tile index the run read, and code_version is the SHA-256 of code/earth_terrain.py followed by code/terrain_features.py as published here. Features (float32 unless the manifest says otherwise):

Column Unit Definition
valid_frac_200m fraction (0-1) n_valid_samples_200m / n_samples_200m; NaN where n_samples_200m == 0, and 0 for a cell whose pixels all fall where no tile exists (open ocean).
elev_min metres Minimum elevation over valid analysis pixels, as height above the EGM2008 geoid (the GLO-30 vertical datum). These are Digital Surface Model heights: buildings, infrastructure and vegetation included.
elev_max metres Maximum elevation over valid analysis pixels (surface height, not bare earth).
elev_mean metres Arithmetic mean elevation over valid analysis pixels.
elev_p05 metres 5th percentile elevation (linear interpolation).
elev_p95 metres 95th percentile elevation (linear interpolation).
elev_range metres elev_max - elev_min.
elev_range_robust metres elev_p95 - elev_p05; outlier-resistant relief.
slope_mean degrees Mean of the Horn 3x3 gradient magnitude over the cell's analysis pixels with a defined slope. Both stencil spacings come from true WGS84 ground distance between block centres - east-west from unwrapped longitudes, so the +-180 seam is exact, north-south from the meridional radius at that latitude; neither is a fixed 200 m.
slope_median degrees 50th percentile of the per-pixel Horn slope.
slope_std degrees Population standard deviation of the per-pixel Horn slope.
slope_p90 degrees 90th percentile of the per-pixel Horn slope.
slope_p99 degrees 99th percentile of the per-pixel Horn slope.
slope_frac_gt_2 fraction (0-1) Fraction of the cell's analysis pixels with a defined Horn slope strictly greater than 2 degrees.
slope_frac_gt_5 fraction (0-1) Same, strictly greater than 5 degrees.
slope_frac_gt_10 fraction (0-1) Same, strictly greater than 10 degrees.
slope_frac_gt_20 fraction (0-1) Same, strictly greater than 20 degrees.
slope_frac_gt_30 fraction (0-1) Same, strictly greater than 30 degrees.
roughness_detrended_std metres Standard deviation of elevation residuals after removing a per-cell least-squares plane fitted in local metres (east, north) on the WGS84 ellipsoid. Removes the regional tilt, so the number describes texture rather than setting.
local_relief_1km metres Cell mean of (max - min) elevation within a 5x5 analysis-pixel window (~1000 m across); a window containing any NaN pixel is excluded.
ew_samples_averaged count Native columns averaged east-west so analysis pixels stay near-square on the ground: 200 m divided by the true ground width of one native column at the band centre. The native column width follows the GLO-30 latitude zone, so this is not 1/cos(lat) alone; it runs from 6 at the equator to 1652 in the northernmost band of this release.
ns_samples_averaged count Native rows averaged north-south, from 200 m divided by the ground height of one 1 arc-second row. It is 6 or 7 everywhere on Earth, and is the same for every cell in a band. Earth blocks in both directions; Mars blocks only east-west.
water_mask_frac fraction (0-1) Native pixels whose GLO-30 Water Body Mask code is 1 (ocean), 2 (lake) or 3 (river), over n_native_samples; code 0 is land. NaN rather than 0 where n_native_samples == 0, or where a DEM tile has no mask beside it (quality flag wbm_missing). Appended, never substituted: no other column excludes water pixels, so elev_*, slope_*, roughness and relief cover every valid pixel, water included, exactly as on Mars.

Renamed in v0.2

terrain-v0.2 is a pure rename: not one value changes, and every cell, region_id and band edge is exactly what terrain-earth-v0.1.0 published. Only the column names moved, so that each says what it measures rather than how it is stored. Everything not in this table kept its name, and the window file itself was renamed from latband5km-v1_windows.parquet (config grid_earth_windows) to latband5km-v1_elevation_windows.parquet (config elevation_windows_earth).

Where v0.1 v0.2
Terrain table n_pixels n_samples_200m
Terrain table n_valid n_valid_samples_200m
Terrain table n_source_pixels n_native_samples
Terrain table n_source_valid n_valid_native_samples
Terrain table valid_fraction valid_frac_200m
Terrain table ew_block_factor ew_samples_averaged
Terrain table analysis_spacing_m sample_spacing_m
Terrain table source_sha256 source_checksum
Terrain table (Earth only) ns_block_factor ns_samples_averaged
Terrain table (Earth only) water_fraction water_mask_frac
Elevation windows src_tile_columns elev_raster_px_per_deg_lon
Elevation windows src_row_min elev_raster_row_start
Elevation windows src_row_max elev_raster_row_stop
Elevation windows src_col_min elev_raster_col_start
Elevation windows src_col_max elev_raster_col_stop
Elevation windows src_zone_split elev_spans_two_zones
Elevation windows src2_tile_columns elev_south_px_per_deg_lon
Elevation windows src2_row_min elev_south_row_start
Elevation windows src2_row_max elev_south_row_stop
Elevation windows src2_col_min elev_south_col_start
Elevation windows src2_col_max elev_south_col_stop
resolve_region keys dem_url elevation_url
resolve_region keys wbm_url water_mask_url

Two of these are worth a sentence. valid_frac_200m says which count it is based on - it is measured on 200 m analysis samples, not on native ones, which is the single most misread thing in v0.1. And water_mask_frac names the Water Body Mask it comes from, which keeps it apart from WorldCover's water_frac.

From a cell to its elevation samples (elevation_windows_earth)

The terrain table is an aggregate: it says how many elevation samples stand behind a cell, never which ones. grid/planet=earth/latband5km-v1_elevation_windows.parquet is the pointer layer that closes that gap. It carries the twelve columns of the plain grid table plus, for each of the 20,403,725 cells, the rectangle of native GLO-30 samples the extractor read. Use it when the 200 m statistics are not precise enough and you want the 30 m samples themselves. The ranges are half-open, so start/stop follows Python slicing (stop is excluded), and the elev_raster_ prefix keeps these columns apart from the elev_* statistics.

Column What
elev_raster_px_per_deg_lon Elevation samples per degree of longitude in this cell's GLO-30 latitude zone (360, 720, 1200, 1800, 2400 or 3600)
elev_raster_row_start, elev_raster_row_stop Half-open row range in the zone's virtual mosaic, row 0 at latitude 90
elev_raster_col_start, elev_raster_col_stop Half-open column range, column 0 at longitude -180, in a mosaic 360 * elev_raster_px_per_deg_lon wide
elev_spans_two_zones True for the 28,052 cells that straddle two pixel-width zones; only then are the elev_south_* columns non-null
elev_south_* The southern segment of a straddling cell: its own samples per degree and its own row and column range

Three things about this table are easy to get wrong:

  • The window is geometry; whether the tiles exist is not. GLO-30 has no open-ocean tiles, so 12,994,896 cells carry a perfectly valid window over samples that are not published anywhere. A populated window is not a promise that data exists there — check n_native_samples in the terrain table, which is zero for exactly those cells.
  • Ten bands own samples in two mosaics. Bands 111, 223, 446, 669, 892, 3108, 3331, 3554, 3777 and 3889 cross a zone boundary, so a cell there has a northern segment at one pixel width and a southern segment at another. Reading only elev_raster_* silently drops up to half the cell. elev_spans_two_zones is the flag that says to read elev_south_* as well.
  • ew_samples_averaged and ns_samples_averaged describe the larger segment, not the window. They are about how the reduction was done; these columns are about which pixels went into it. For a split cell the two statements are about different halves, which is why the flag matters.

Mosaic coordinates are not tile filenames. code/earth_terrain.py ships the resolver, which keeps the pointer layer and the extractor from drifting apart:

import earth_terrain as et                     # code/earth_terrain.py from this repo
from planetary_grid import PlanetBandGrid      # code/planetary_grid.py, published with the Mars release

grid = PlanetBandGrid.build("earth")
# Reading the bucket is anonymous, but not credential-free: set AWS_NO_SIGN_REQUEST=YES,
# or GDAL raises InvalidCredentials. GLO-30 declares no nodata value, hence nodata=None.
source = et.EarthSource(
    root="/vsis3/copernicus-dem-30m",
    tiles=et.load_tile_index("https://copernicus-dem-30m.s3.amazonaws.com/tileList.txt"),
    nodata=None,
)
reads = et.resolve_region(grid, source, "E_2084_04949")
for read in reads:
    print(read["elevation_url"], read["row_off"], read["col_off"], read["n_rows"], read["n_cols"])
    # Copernicus_DSM_COG_10_N03_00_E042_00_DEM.tif 646 2736 163 168
    # read["water_mask_url"] is the matching Water Body Mask, or None where the tile has none.

# The windows reproduce the aggregate exactly: this sums to 27,384, which is
# n_native_samples for E_2084_04949 in the terrain table.
sum(r["n_rows"] * r["n_cols"] for r in reads)

The index you pass decides which tiles are considered to exist. Every row records the SHA-256 of the index this release read in source_checksum, so if you need to reproduce a cell exactly rather than merely re-read it today, compare that hash against the tileList.txt you fetch.

resolve_region returns one entry per tile a cell touches, already clipped to the window inside that tile, and an empty list for a cell with no existing tile — the honest answer for open ocean. The native pixels themselves are not republished here: they stay in the public copernicus-dem-30m bucket under the licence below.

Limitations

  • Digital Surface Model. GLO-30 includes buildings, infrastructure and vegetation, so the elevations are surface heights, not bare earth.
  • No tile, no statistics. Open-ocean tiles do not exist in GLO-30 and some tiles are withheld from the public instance; cells with no source tile carry no elevation statistics. Check n_valid_native_samples and quality_flags before use.
  • The manifest manifests/terrain_earth_v0.2.json records the validation verdict, per-feature checks and the complete list of known limitations of this release.

Source and terms

Source: Copernicus DEM GLO-30 Public (Copernicus DEM GLO-30 Public (COP-DEM_GLO-30, derived from the COP-DEM_GLO-30-DGED source)). Provider per the EULA: Airbus Defence and Space GmbH (provider), Copernicus programme / ESA (licensor). Source tile index SHA-256 recorded in every row: tileindex-sha256:10604e3052c98a09e9216f1a8f0a555a04148419757575f783d4937fd44316dc.

Mars

Mars grid and terrain tables live under grid/planet=mars/ and terrain/planet=mars/, unchanged by this release. Their provenance, source credit and terms are in manifests/terrain_mars_v0.2.json. The Mars documentation is reproduced in full below, so that publishing Earth does not take it off this page.

Per-cell terrain statistics for all of Mars (5,797,982 cells), computed from the USGS MOLA-HRSC blended 200 m DEM and indexed by a permanent, versioned latitude-band grid (latband5km-v1). The grid and the terrain features are separate tables that join on grid_version + region_id, so later datasets can be added to the same cells without touching these files.

Release: tag terrain-mars-v0.3.0 (grid latband5km-v1, features terrain-v0.2). Pin it with revision="terrain-mars-v0.3.0".

Contents

Path What
grid/planet=mars/latband5km-v1.parquet One row per Mars cell: id, bounds, centre, area
grid/mars_global_bands.csv Band catalog: latitude edges, columns per band, area per cell
grid/grid_manifest.json Grid definition and statistics (also lists the Earth reference, which has no table here)
terrain/planet=mars/feature_version=terrain-v0.2/band_group=NNN.parquet Terrain features, one file per 100 bands (22 files)
manifests/terrain_mars_v0.2.json Full build manifest: source, provenance, validation, measured artifacts, limitations
code/planetary_grid.py, code/mars_terrain.py, code/terrain_features.py The code that produced the data, at git commit d968501

Quick start

from datasets import load_dataset

grid = load_dataset("enzeeoai/starman", "grid_mars", split="train", revision="terrain-mars-v0.3.0")
terrain = load_dataset("enzeeoai/starman", "terrain_mars", split="train", revision="terrain-mars-v0.3.0")
import pandas as pd

base = "hf://datasets/enzeeoai/starman@terrain-mars-v0.3.0"
grid = pd.read_parquet(f"{base}/grid/planet=mars/latband5km-v1.parquet")
terrain = pd.read_parquet(f"{base}/terrain/planet=mars/feature_version=terrain-v0.2/band_group=010.parquet")
cells = terrain.merge(
    grid.drop(columns=["planet", "band", "column"]), on=["grid_version", "region_id"], validate="1:1")

Joining other datasets

Every row in every table of this repo is keyed by grid_version + region_id. region_id is M_<band:04d>_<column:05d> (for example M_1067_01234) and is permanent within a grid_version: a cell's bounds and area never change. A different grid_version is a different set of cells - never join across versions. A new dataset should add <name>/planet=mars/... files with those two columns, and ship the grid_version it was built on. The grid table is also the way to get coordinates for any other table: join on the key and read center_lat / center_lon or the bounds.

Grid table columns (grid_mars)

Column Type Unit Meaning
grid_version string - latband5km-v1
region_id string - Permanent cell id M_<band>_<column>
planet string - mars
band int32 - Latitude band, 0 (south pole) to 2133 (north pole)
column int32 - Longitude column within the band, west to east from -180
lat_min, lat_max float64 degrees Band edges, planetocentric latitude on the 3,396,190 m Mars sphere
lon_min, lon_max float64 degrees east Cell edges in [-180, 180]
center_lat, center_lon float64 degrees Midpoint of the bounds
area_km2 float64 km^2 Cell area on the sphere (about 25 km^2, varies slightly by band)

Bands have equal meridional arc length (~5 km). Each band is divided into round(circumference / 5 km) equal-longitude columns, so the number of columns shrinks toward the poles. PlanetBandGrid.lookup(lat, lon) in code/planetary_grid.py maps a point to its cell; the table was checked against it for random points.

Terrain table columns (terrain_mars)

Identity and provenance:

Column Type Meaning
grid_version, region_id, planet, band, column Same as the grid table
quality_flags string Semicolon-separated advisory flags, empty when clean (see Limitations)
source_product, source_version, source_checksum string The DEM that was read
sample_spacing_m float32, m Ground spacing of the analysis grid (200)
feature_version string terrain-v0.2
code_version string SHA-256 of code/mars_terrain.py followed by code/terrain_features.py, as published here
computed_at string UTC timestamp of the computation

Pixel counts - read this before using valid_frac_200m:

  • n_samples_200m, n_valid_samples_200m count analysis pixels: the DEM after east-west block averaging (below). An analysis pixel is ew_samples_averaged native pixels averaged along a row. At the equator that is one native pixel; near the poles it is hundreds.
  • n_native_samples, n_valid_native_samples count native 200 m DEM pixels. n_native_samples summed over all cells equals the raster's width x height exactly (every source pixel belongs to exactly one cell).
  • valid_frac_200m = n_valid_samples_200m / n_samples_200m is measured on analysis pixels. Where ew_samples_averaged > 1 a single nodata native pixel is averaged away, so use n_valid_native_samples / n_native_samples for native coverage.
  • n_samples_200m and n_valid_samples_200m are stored as float32 (whole numbers); n_native_samples and n_valid_native_samples are int64.

Features (all float32):

Column Unit Definition
valid_frac_200m fraction (0-1) n_valid_samples_200m / n_samples_200m; NaN where n_samples_200m == 0.
elev_min metres Minimum elevation over valid analysis pixels, relative to the Mars areoid datum of the source product.
elev_max metres Maximum elevation over valid analysis pixels.
elev_mean metres Arithmetic mean elevation over valid analysis pixels.
elev_p05 metres 5th percentile elevation (linear interpolation).
elev_p95 metres 95th percentile elevation (linear interpolation).
elev_range metres elev_max - elev_min.
elev_range_robust metres elev_p95 - elev_p05; outlier-resistant relief.
slope_mean degrees Mean of the Horn 3x3 gradient magnitude over the cell. The east-west spacing of each stencil is taken from the unwrapped block-centre longitudes, so the +-180 seam is handled exactly; north-south spacing is the constant 200 m row pitch.
slope_median degrees 50th percentile of the per-pixel Horn slope.
slope_std degrees Population standard deviation of the per-pixel Horn slope.
slope_p90 degrees 90th percentile of the per-pixel Horn slope.
slope_p99 degrees 99th percentile of the per-pixel Horn slope.
slope_frac_gt_2 fraction (0-1) Fraction of the cell's valid analysis pixels with Horn slope strictly greater than 2 degrees.
slope_frac_gt_5 fraction (0-1) Fraction of the cell's valid analysis pixels with Horn slope strictly greater than 5 degrees.
slope_frac_gt_10 fraction (0-1) Fraction of the cell's valid analysis pixels with Horn slope strictly greater than 10 degrees.
slope_frac_gt_20 fraction (0-1) Fraction of the cell's valid analysis pixels with Horn slope strictly greater than 20 degrees.
slope_frac_gt_30 fraction (0-1) Fraction of the cell's valid analysis pixels with Horn slope strictly greater than 30 degrees.
roughness_detrended_std metres Standard deviation of elevation residuals after removing a per-cell least-squares plane fitted in local metres (east, north). Removes the regional tilt so the number describes texture, not setting.
local_relief_1km metres Cell mean of (max - min) elevation within a 5x5 analysis-pixel window (~1000 m across at the band centre).
ew_samples_averaged count k = max(1, floor(1/cos(lat_centre) + 0.5)): native columns averaged east-west before analysis, so analysis pixels stay near-square on the ground.

Renamed in v0.2

terrain-v0.2 is a pure rename: not one value changes, and every cell, region_id and band edge is exactly what terrain-mars-v0.1.0 published. Only the column names moved, so that each says what it measures rather than how it is stored. Everything not in this table kept its name.

Where v0.1 v0.2
Terrain table n_pixels n_samples_200m
Terrain table n_valid n_valid_samples_200m
Terrain table n_source_pixels n_native_samples
Terrain table n_source_valid n_valid_native_samples
Terrain table valid_fraction valid_frac_200m
Terrain table ew_block_factor ew_samples_averaged
Terrain table analysis_spacing_m sample_spacing_m
Terrain table source_sha256 source_checksum

Migrating is a rename and nothing else - the v0.1 tags remain, and pd.read_parquet(...).rename(columns=...) on a v0.1 file reproduces a v0.2 file's columns.

Method

  • Source. One 200 m/pixel simple-cylindrical GeoTIFF (Mars 2000 sphere, planetocentric latitude, east longitude). Each latitude band is read once as a strip of full-width pixel rows plus a two-row halo.
  • Analysis spacing 200 m. North-south spacing is the native 200 m row pitch. East-west pixel width on the ground shrinks with cos(latitude), so rows are block-averaged east-west to keep analysis pixels near-square.
  • East-west block averaging, one k per band. k = max(1, floor(1/cos(lat_centre) + 0.5)), evaluated once at the band's centre latitude and used for every row of the band's read window so the three rows of a Horn stencil stay column-aligned. Where the per-row factor would differ inside the band the cell is flagged mixed_block_factor. The factor is recorded per cell as ew_samples_averaged.
  • Horn 3x3. Slope is the magnitude of the Horn (1981) gradient on the block-averaged grid, in degrees, using local metric spacing.
  • Pixel-centre assignment. An analysis pixel belongs to the cell that contains its block-centre longitude (and whose band contains its row-centre latitude). No pixel is split between cells, so cells hold 450 to 950 analysis pixels depending on band.
  • True dx at the +-180 degree seam. The source raster does not close the longitude ring: its east edge is 0.00155 degrees (about 0.46 pixel, ~92 m) short of +180. The east-west stencil spacing is therefore taken from the unwrapped block-centre longitudes, not assumed to be 2k pixels, so the short final block and the gap are absorbed exactly. The global maps show no discontinuity at the seam.
  • Detrended roughness removes a per-cell least-squares plane in local metres before taking the standard deviation. Local relief is the cell mean of max - min in a 5 x 5 analysis-pixel window.
  • Validation. 50 cells drawn uniformly at random were recomputed with an independent reference (polygon containment on pixel centres, pure-Python Horn and relief loops, numpy.linalg.lstsq detrending, numpy.percentile). All features agree to within 0.5 ULP of float32 storage; the manifest records the per-feature errors.

Limitations

  • Effective resolution is coarser than 200 m where HRSC is absent. The product blends ~50 m HRSC DEMs (about 44% of Mars at the time the product page was written) onto the ~463 m MOLA DEM, then samples to 200 m. Outside HRSC coverage the data are MOLA-limited, so 200 m features there are smoother than the nominal spacing suggests. The source has no coverage mask, so this release cannot say which cells are which.
  • Swath footprints are visible in texture. HRSC footprints appear as straight-edged rectangles. Elevation is continuous across them but slope and roughness are not: measured ratios inside vs outside two northern-plains footprints were 3.2x and 4.1x for slope_mean. This is in the source, not introduced by the pipeline. A model trained on these features can learn footprint boundaries instead of geology.
  • Elevation is context only. Heights are the source values, passed through unchanged, in metres. The datum is the areoid as inferred by us from pole values and the Olympus Mons maximum; the product's label does not state it, nor does it state the vertical unit (metres is also an inference from the value range). Do not mix these elevations with data on a different datum without checking.
  • polar_distortion (3,618 cells): band centre |latitude| > 88 degrees. A 5 km cell there is far wider than the 200 m source can support, ew_samples_averaged exceeds 28 (it reaches 1359 in band 0) and features are heavily smoothed. The two polar-cap bands also lose stencil rows to the raster edge. Treat as advisory.
  • mixed_block_factor (31,724 cells, including every polar_distortion cell): the per-row block factor is not constant across the band, so the band-centre k is only approximate for some rows.
  • Do not compare across a block-factor step. k changes with latitude and more averaging lowers slope and roughness. Band-to-band change at the 102 k transitions is 6.4% (median) vs 1.6% elsewhere, visible as faint stripes between |lat| 78 and 88. Stratify on ew_samples_averaged.
  • Other quality_flags values (low_coverage, no_pixels, degenerate_block_grid) are defined but occur in no cell of this release.
  • The source's 39,957 nodata pixels all sit in its last column (+180 edge); 720 cells have valid_frac_200m between 0.96 and 0.997 because of them.
  • 22 cells on the Elysium / Cerberus plains have exactly zero elev_range: the source is a single constant Int16 value there (relief under the 1 m quantisation).
  • Cells are not equal area and not square: width in metres varies within a band and area is about 25 km^2.
  • Features are float32; percentiles for polar cells with few valid pixels are coarse.
  • code/mars_terrain.py and code/terrain_features.py are the versions that wrote these rows (d968501). The feature maths lives in the shared module, which Earth uses too, and code_version is the SHA-256 of the driver followed by it - so a changed feature definition cannot ship under an unchanged stamp. The values themselves are carried across from terrain-mars-v0.1.0 unmodified.

Source, credit and terms

Derived from Mars MGS MOLA - MEX HRSC Blended DEM Global 200m, v2, distributed by the USGS Astrogeology Science Center (Astropedia): https://astrogeology.usgs.gov/search/map/mars_mgs_mola_mex_hrsc_blended_dem_global_200m. The page's recommended citation is: Fergason, R. L., Hare, T. M., & Laura, J. (2018). HRSC and MOLA Blended Digital Elevation Model at 200m v2. Astrogeology PDS Annex, U.S. Geological Survey. The product combines NASA Mars Global Surveyor MOLA elevations with ESA Mars Express HRSC elevation models. Source file SHA-256: e8bfe4c3b9dee9d5fdb8c4d49798869654e552e7b154712c9e12d6d117464419. The page describes v2 as an interim product that will be superseded.

License: not verified. We could not confirm the USGS terms for this product. The Astropedia page could not be fetched by our tooling, the product label (.lbl) has no rights or credit field, and no license text was found elsewhere. USGS-authored works are generally in the public domain in the United States, but this product includes ESA/DLR HRSC-derived data whose terms we have not checked. The license: other in this card's metadata means exactly that: terms unverified, not "open". Before reusing or redistributing, read the terms on the Astropedia page and the HRSC team's data policy, and credit the source above. Our own code in code/ carries no separate license in this release.

The source product name and version in the build manifest come from the file name; the label does not record them.

Phase C partial native features — 2026-11-14

The pinned partial release added 50 Parquets covering selected dunes, channels, roughness, Earth geomorphometry, Mars climate, EMIT abundance, provisional TES composition and regional Atacama ERA5 temperature. This is a partial release with source gaps and held comparisons; publication does not enable matcher features.

See the release inventory and restrictions and file manifest. All seven requested Copernicus groups are included for both geomorphometry and K–H. Existing terrain releases and previously published layers are preserved. Use tag phase-c-partial-20261114-r2 for this snapshot.

Core production features

Field-specific restrictions remain in effect. The historical subset section above describes the pinned partial release.

Current Starman excludes the provisional Mars TES composition layer. TES albedo and thermal inertia cover all 22 Mars groups.

Mars valley networks (mars_valleys-v0.2)

The Alemanno and Orofino map supplies digitized Martian fluvial systems. The layer reports cell channel length per cell area (drain_dens_km_km2), the maximum inferred Strahler order (max_order), and centre-to-channel distance (dist_channel_km). Lines are generalized to 460 m, endpoints snapped within 1 km, and the network pruned at 20 km. Flow orientation is inferred. Small and buried valleys may be absent, and the global effects of topology preparation remain incompletely validated; treat the output as context rather than an unrestricted Earth comparison.

The layer has 5,797,982 grid rows across 22 groups. Stored clipped lengths sum to 427,244.300 km. In 400 stratified checks across five groups, the largest independent clipped-length error was 0.0000195 km and all orders matched. Independent reconciliation of the oriented graph matched pruning masks at 10, 20 and 40 km; at 20 km it retained 24,442 of 60,844 edges. These checks do not validate all snapping and orientation decisions. Source: Alemanno and Orofino, global map of Martian fluvial systems, Zenodo DOI 10.5281/zenodo.1051038, CC BY-SA 4.0; non-commercial use approved by Enzo on 2026-11-13. Version identifier: mars_valleys-mars-v0.2.0.

Mars cave candidates (MGC3)

This site list contains 1,062 Mars cave candidates from the Mars Global Cave Candidate Catalog PDS4 bundle. Coordinates are east-positive longitude and planetocentric latitude, with region_id assigned on latband5km-v1. All identifiers are unique and all coordinates resolve to their stored regions. Candidates are not confirmed caves, and priority is an image-targeting rank rather than confidence. There is no matched Earth catalog; these rows supply query-site context and are never scored.

Source: Cushing, Mars Global Cave Candidate Catalog, PDS4 bundle urn:nasa:pds:mars_mro.odyssey_multi_cavecatalog_cushing_2016, v1.0. Licence text was not found in the bundle; public USGS/PDS access is assumed but unverified. Non-commercial use approved by Enzo on 2026-11-13. Version identifier: mgc3_caves-mars-v0.1.0.

MGD3 modeled dune fields (Mars; mgd3_dunes-v0.2)

The Mars Global Digital Dune Database provides dune-field centroids and areas. Each field is represented by an equivalent-area disc on the Mars sphere. dune_frac is the fraction of a cell covered by the union of discs with a counted class; dune_type is the dominant shared type; dist_dune_km is the distance from the cell centre to the nearest counted disc. Sand sheets and other types are excluded. The categorical mixed-class split is approximate, and the modeled boundaries are not measured outlines. valid_frac gives the fraction of a cell within the survey latitude range; wholly outside cells have missing features and outside_survey flags.

There are 5,797,982 rows across 22 groups on latband5km-v1. The full grid, value ranges and missing survey coverage were checked. Independent 50-cell samples gave maximum fraction error 0.000381 against 0.02 and maximum distance error 0.00138 km against 0.25 km. Comparisons to Earth require reviewed frozen reference populations; this layer is a modeled-footprint proxy.

Source: Mars Dune Digital Database Merged, Gullikson, Hayward and Titus, ScienceBase DOI 10.5066/P9ORPZKL; public domain per the source metadata, approved for non-commercial use by Enzo on 2026-11-13. Version identifier: mgd3_dunes-mars-v0.2.0.

Robbins crater context (robbins_craters-v0.1)

Robbins and Hynek's Mars crater database supplies counts within 50 km of each cell centre, density per million square kilometres, the largest diameter in that radius, and distance to the nearest modeled crater rim. The radius uses the Mars sphere; only craters at least 1 km in diameter are counted. The output is Mars-only context and is never scored. The source has known errors and its licence text is unverified; non-commercial use was approved by Enzo on 2026-11-13.

The dataset contains 5,797,982 rows across 22 groups. Source: Robbins and Hynek (2012), DOI 10.1029/2011JE003966, author-site 20121016 database. Version identifier: robbins_craters-mars-v0.1.0. Full-grid and independent reference results are in the layer manifest.

tes_albedo (mars), tes_albedo-v0.2

Source: TES bolometric albedo global mosaic, 7410 m (USGS Astropedia copy; Christensen et al. 2001) (unversioned (no label on disk)); https://planetarymaps.usgs.gov/mosaic/Mars_MGS_TES_Albedo_mosaic_global_7410m.tif. Licence: USGS catalog: public domain; cite Christensen et al. 2001, doi:10.1029/2000JE001370..

5,797,982 grid rows in 22 base Parquets; 0 extra Parquets. Version identifier tes_albedo-mars-v0.2.0. Grid: latband5km-v1.

Column Type Unit Comparison Meaning
albedo float32 unitless planet_specific TES bolometric (~0.3-2.9 um) Lambert albedo, NaN-aware bilinear at the cell centroid from the 7.41 km mosaic. The MODIS Earth product has different spectral and illumination support.
src_res_km float32 km planet_specific Pixel size of the source raster at the equator. Coverage field, never scored. Earth twin: src_res_km (coverage; 5.55 km).
interp_flag bool flag planet_specific True when a NaN source pixel was left out of the bilinear weights. Coverage field, never scored. Earth twin: interp_flag (coverage).

Known limitations:

  • TES bolometric albedo (~0.3-2.9 um) has different spectral and angular support from MODIS white-sky shortwave albedo (0.3-5 um) on Earth.
  • Single epoch, 7.41 km pixels; direct USGS catalog page returned 403, but indexed catalog metadata and the ASU dataset page agree on longitude and citation.
  • Full-planet QA passed. Original Parquet source metadata retains historical unverified source wording; current source citations and catalog references are recorded in the layer manifest.

tes_ti (mars), tes_ti-v0.2

Source: TES Thermal Inertia Maps (TIMAP), Putzig & Mellon 2007 (MGS-M-TES-5-TIMAP-V1.0 (2007-01-30)); https://pds-geosciences.wustl.edu/mgs/mgs-m-tes-5-timap-v1/mgst_9001/data/. Licence: NASA PDS; cite Putzig, Mellon, Heet and Ward, MGS-M-TES-5-TIMAP-V1.0, 2009; see dataset profile for terms..

5,797,982 grid rows in 22 base Parquets; 22 extra Parquets. Version identifier tes_ti-mars-v0.2.0. Grid: latband5km-v1.

Column Type Unit Comparison Meaning
ti_tiu_mean float32 J m-2 K-1 s-1/2 (tiu) planet_specific Mean night-side TES thermal inertia of the pixels whose centres fall in the cell (TI = 0 no-data pixels excluded).
ti_tiu_std float32 J m-2 K-1 s-1/2 (tiu) planet_specific Population standard deviation of the night TI pixels in the cell.
ti_tiu_p10 float32 J m-2 K-1 s-1/2 (tiu) planet_specific 10th percentile of the night TI pixels in the cell.
ti_tiu_p50 float32 J m-2 K-1 s-1/2 (tiu) planet_specific Median of the night TI pixels in the cell.
ti_tiu_p90 float32 J m-2 K-1 s-1/2 (tiu) planet_specific 90th percentile of the night TI pixels in the cell.
ti_day_tiu_mean float32 J m-2 K-1 s-1/2 (tiu) planet_specific Mean day-side TES thermal inertia over the same pixels (TI = 0 excluded).
ti_interp_frac float32 fraction planet_specific Share of the valid night TI pixels (TI > 0) that were infilled by interpolation (mask 0). Coverage field, never scored. Earth twin: none (coverage).
n_px int32 pixels planet_specific Number of 20 px/deg pixels whose centres fall in the cell. Coverage field, never scored. Earth twin: none (coverage).
valid_frac float32 fraction planet_specific Share of the cell pixels with night TI > 0. Coverage field, never scored. Earth twin: valid_frac (coverage).
ti_pct (extra ti_pct) float32 percentile (0-100) within_planet Percentile rank (0-100, ties share the mid rank) of ti_tiu_p50 within the whole-planet distribution of Mars cells with a valid TI and ti_interp_frac < 0.5. Cells outside that reference set are ranked against it too and carry the interpolated flag. The Earth ATI rank uses a separate property and reference mask.

Known limitations:

  • TES TI map release is 2007. 6.4% of pixels in the day map and 7.6% in the night map are interpolated (area-weighted).
  • TI = 0 marks no data (polar rows and gaps); it is NaN here, not zero inertia.
  • TI = 4999 appears to be an undocumented cap (113 day and 140 night pixels at the cap).
  • Cells are summarised by pixel-centre assignment of 2.96 km pixels; cells with no pixel centre use the pixel under the cell centre (flag centroid_px).
  • Seasonal frost near the poles biases TI; cells poleward of 80 degrees carry the polar flag.
  • Earth ATI is an independent thermal-response proxy. Cross-planet rank equivalence is unvalidated.
  • Full-planet QA and independent percentile check passed. Cross-planet TI/ATI equivalence is unvalidated. Original Parquet source metadata retains historical unverified attribution wording; current source citations and catalog references are recorded in the layer manifest.

Visual inspection: high northern latitude diagonal stripes and bright polar band appear in independent source-pixel averages as well as output. They remain source-data limitations, not established geophysical features.

emit_minerals (earth), emit_minerals-v0.2

Source: EMITL3ASA (002); https://doi.org/10.5067/EMIT/EMITL3ASA.002. Licence: Freely Distributed (file attribute; LP DAAC data-use policy, no further terms found).

20,403,725 grid rows in 41 base Parquets; 0 extra Parquets. Version identifier emit_minerals-earth-v0.2.0. Grid: latband5km-v1.

Column Type Unit Comparison Meaning
min_frac_feoxide float32 1 planet_specific Held former enrichment fraction for iron oxides (hematite + goethite); all values missing pending independent threshold validation at L3 support.
min_abund_feoxide float32 mass fraction 0-1 planet_specific iron oxides (hematite + goethite): sum of member median-grain-size abundances, bilinear at the cell centroid (containing cell if a neighbour has no data; see interp_flag).
min_abund_feoxide_lo float32 mass fraction 0-1 planet_specific iron oxides (hematite + goethite): sum of the members' UQ low-grain-size layers, same interpolation as min_abund_feoxide. Grain-size envelope, not a confidence interval.
min_abund_feoxide_hi float32 mass fraction 0-1 planet_specific iron oxides (hematite + goethite): sum of the members' UQ high-grain-size layers, same interpolation as min_abund_feoxide. Grain-size envelope, not a confidence interval.
min_frac_al_clay float32 1 planet_specific Held former enrichment fraction for Al clays (kaolinite + montmorillonite + illite/muscovite); all values missing pending independent threshold validation at L3 support.
min_abund_al_clay float32 mass fraction 0-1 planet_specific Al clays (kaolinite + montmorillonite + illite/muscovite): sum of member median-grain-size abundances, bilinear at the cell centroid (containing cell if a neighbour has no data; see interp_flag).
min_abund_al_clay_lo float32 mass fraction 0-1 planet_specific Al clays (kaolinite + montmorillonite + illite/muscovite): sum of the members' UQ low-grain-size layers, same interpolation as min_abund_al_clay. Grain-size envelope, not a confidence interval.
min_abund_al_clay_hi float32 mass fraction 0-1 planet_specific Al clays (kaolinite + montmorillonite + illite/muscovite): sum of the members' UQ high-grain-size layers, same interpolation as min_abund_al_clay. Grain-size envelope, not a confidence interval.
min_frac_femg_clay float32 1 planet_specific Held former enrichment fraction for Fe/Mg clays (chlorite + vermiculite); all values missing pending independent threshold validation at L3 support.
min_abund_femg_clay float32 mass fraction 0-1 planet_specific Fe/Mg clays (chlorite + vermiculite): sum of member median-grain-size abundances, bilinear at the cell centroid (containing cell if a neighbour has no data; see interp_flag).
min_abund_femg_clay_lo float32 mass fraction 0-1 planet_specific Fe/Mg clays (chlorite + vermiculite): sum of the members' UQ low-grain-size layers, same interpolation as min_abund_femg_clay. Grain-size envelope, not a confidence interval.
min_abund_femg_clay_hi float32 mass fraction 0-1 planet_specific Fe/Mg clays (chlorite + vermiculite): sum of the members' UQ high-grain-size layers, same interpolation as min_abund_femg_clay. Grain-size envelope, not a confidence interval.
min_frac_sulfate float32 1 planet_specific Held former enrichment fraction for sulfate (gypsum); all values missing pending independent threshold validation at L3 support.
min_abund_sulfate float32 mass fraction 0-1 planet_specific sulfate (gypsum): sum of member median-grain-size abundances, bilinear at the cell centroid (containing cell if a neighbour has no data; see interp_flag).
min_abund_sulfate_lo float32 mass fraction 0-1 planet_specific sulfate (gypsum): sum of the members' UQ low-grain-size layers, same interpolation as min_abund_sulfate. Grain-size envelope, not a confidence interval.
min_abund_sulfate_hi float32 mass fraction 0-1 planet_specific sulfate (gypsum): sum of the members' UQ high-grain-size layers, same interpolation as min_abund_sulfate. Grain-size envelope, not a confidence interval.
min_frac_carbonate float32 1 planet_specific Held former enrichment fraction for carbonates (calcite + dolomite); all values missing pending independent threshold validation at L3 support.
min_abund_carbonate float32 mass fraction 0-1 planet_specific carbonates (calcite + dolomite): sum of member median-grain-size abundances, bilinear at the cell centroid (containing cell if a neighbour has no data; see interp_flag).
min_abund_carbonate_lo float32 mass fraction 0-1 planet_specific carbonates (calcite + dolomite): sum of the members' UQ low-grain-size layers, same interpolation as min_abund_carbonate. Grain-size envelope, not a confidence interval.
min_abund_carbonate_hi float32 mass fraction 0-1 planet_specific carbonates (calcite + dolomite): sum of the members' UQ high-grain-size layers, same interpolation as min_abund_carbonate. Grain-size envelope, not a confidence interval.
interp_flag string categorical 'bilinear' when all 4 neighbouring L3ASA cells have data, 'nearest' when the containing cell was used instead; null where there is no L3ASA data.
has_emit bool planet_specific True when any EMIT L2BMIN granule footprint (starts before the guide cutoff) overlaps the cell under the V002 guide scenario. Cloud-blind acquisition footprint.
cov_frac_emit float32 1 planet_specific Fraction of the cell area inside the union of cutoff-filtered EMIT L2BMIN footprints under the V002 guide scenario (cloud-blind).
emit_l3_valid bool planet_specific True when part of the cell overlaps a finite 0.5-degree L3ASA aggregate. This is regional support.
src_res_km float32 km planet_specific Native resolution of the L3ASA source (0.5 deg, about 55 km).

Known limitations:

  • EMIT L3ASA is a 0.5 degree product masked for vegetation, water, cloud and urban cover: only about 9% of its cells have data, between about 52.5N and 51S. Everything else is NaN (no_emit_data or outside_emit_domain).
  • min_frac_* is held pending independent validation of the enrichment thresholds at L3 support.
  • min_abund_*_lo / _hi sum the members' grain-size uncertainty layers; they bound the grain-size assumption only.
  • has_emit / cov_frac_emit are footprint-based (L2BMIN granule polygons started before 2023-07-31 under the guide scenario). The exact L3 contributor list and native valid area are unknown.
  • Earth only in v0.1: no Mars twin (the Mars OMEGA / MOCAAS sources are unavailable).
  • All min_frac_* fields remain unpopulated. cov_frac_emit measures the cloud-blind acquisition footprint. Grain-size variants bound the grain-size assumption.

Global cell-weighted quantiles and terrain-validity proxy coverage are in the manifest; the proxy is not a verified land mask. All five abundance baselines lie within their grain-size variant envelope in finite output cells.

worldcover (Earth), worldcover-v0.1

Source: ESA WorldCover 10 m 2021 v200 (x4 overview) (v200 (2021)); s3://esa-worldcover/v200/2021/map/. Licence: CC BY 4.0; "(c) ESA WorldCover project 2021 / Contains modified Copernicus Sentinel data (2021) processed by ESA WorldCover consortium".

20,403,725 rows across 41 groups; 6,158,972 have valid class fractions. Version identifier worldcover-earth-v0.1.0. Grid latband5km-v1.

Column Type Unit Comparison Meaning
tree_frac float32 fraction planet_specific Fraction of valid pixels in class 10 (tree)
shrub_frac float32 fraction planet_specific Fraction of valid pixels in class 20 (shrub)
grass_frac float32 fraction planet_specific Fraction of valid pixels in class 30 (grass)
crop_frac float32 fraction planet_specific Fraction of valid pixels in class 40 (crop)
built_frac float32 fraction planet_specific Fraction of valid pixels in class 50 (built)
bare_frac float32 fraction planet_specific Fraction of valid pixels in class 60 (bare). Earth-only; no Mars value is inferred.
snow_frac float32 fraction planet_specific Fraction of valid pixels in class 70 (snow)
water_frac float32 fraction planet_specific Fraction of valid pixels in class 80 (water)
wetland_frac float32 fraction planet_specific Fraction of valid pixels in class 90 (wetland)
mangrove_frac float32 fraction planet_specific Fraction of valid pixels in class 95 (mangrove)
moss_frac float32 fraction planet_specific Fraction of valid pixels in class 100 (moss)
lc_major int16 code categorical Majority WorldCover class code (ties go to the lowest code; 0 = no valid pixel)
surface_not_ground bool flag planet_specific (tree + mangrove + built) fraction >= SURFACE_NOT_GROUND_THRESHOLD (0.20), provisional canopy/roof indicator; use only if explicitly requested. False with no valid pixels is unknown, not bare ground
n_px int32 count planet_specific x4-overview pixels read whose centre is in the cell
valid_frac float32 fraction planet_specific Valid pixels over the pixels the cell would hold with full tile coverage

Known limitations:

  • Read from the x4 overview (about 40 m): every overview pixel is one pixel of its 4 x 4 parent block (nearest-style), so fractions differ from full resolution (S-A: up to 0.03 for built-up in a city).
  • Open sea is coded 0 (nodata) in some places and 80 in others; class 80 is not ocean and code 0 is not "outside the tile". valid_frac is low where sea is coded 0.
  • No Antarctica; the north limit is 82.75 N (flag no_tile).
  • Class 70 (snow and ice) and 100 (moss and lichen) are reported as the product codes them.
  • Every row has flag overview_resampled.
  • surface_not_ground uses a 0.20 threshold that has not been tuned against Copernicus bias yet.
  • The 20% tree+mangrove+built indicator remains provisional; use only when explicitly requested. False with no valid pixels is unknown.
  • The native-vs-x4 calibration covered 702 stratified cells; 27 cutoff-enriched flags changed. Independent labeled review remains outstanding.
  • Seven retained groups have a prior code hash: only comments and metadata descriptions differ from the current module. Both source versions and per-group code hashes are recorded in the manifest.
  • At 20%, stored float32 fractions can round across the threshold; differing reconstructed flags lie within 8.2e-9 of 0.20.

Water content in the upper 1 m of the Martian regolith from TGO FREND (Zenodo 6282972)

Version identifier: frend_weh-mars-v0.1.0. Grid: latband5km-v1, 22 band groups, 5,797,982 cells. Grid checks passed. See the layer manifest for full details.

Primary field ice_wt_pct_frend: 4,442,732 finite values; p01 1.9498, median 4.6611, p99 11.0222; range 1.38295–21.8325. Complete per-column distributions, units, flags, source references, file paths and SHA-256s are in the manifest.

Water-equivalent hydrogen is smoothed about 12° FWHM, only ±50° latitude, and its absolute level is tied to GRS. No absolute Earth pair.

Source: https://doi.org/10.5281/zenodo.6282972. Use and attribution: CC BY 4.0 (Zenodo record); cite Malakhov et al. 2022, JGR Planets, doi:10.1029/2022JE007258.

Across 4,440,330 paired cells FREND−GRS mean is +0.9683 percentage points and Pearson r is 0.7505; differing support, spatial dependence and shared calibration limit interpretation. The cell-weighted Solis output box mean is 2.907 wt%, versus the 2.78 wt% source calibration anchor; these use different supports and weighting. See the layer manifest for details.

Circum-Arctic Map of Permafrost and Ground-Ice Conditions (GGD318)

Version identifier: ggd318-earth-v0.1.0. Grid: latband5km-v1, 41 band groups, 20,403,725 cells. Grid checks passed. See the layer manifest for full details.

Primary field ice_wt_pct: 2,808,530 finite values; p01 0, median 0, p99 16.9249; range 0–16.9249. Complete per-column distributions, units, flags, source references, file paths and SHA-256s are in the manifest.

Class-derived ground-ice mass uses assumed ice and soil densities and open-ended class representatives; it is not an absolute counterpart to GRS or FREND.

Source: https://doi.org/10.7265/skbg-kf16. Use and attribution: No explicit licence; citation required: Brown, J., O. Ferrians, J. A. Heginbottom, and E. Melnikov. 2002. Circum-Arctic Map of Permafrost and Ground-Ice Conditions, Version 2. NSIDC DAAC, doi:10.7265/skbg-kf16.

Named-site QA includes Yakutsk, Canadian Shield and southern Canada; Yakutsk native center polygon is low-ice Clf, while the output represents gridded area sampling. See the layer manifest for details.

Mars Odyssey GRS 5-element concentrations, H2O (ODY-M-GRS-5-ELEMENTS-V1.0, product v2.0)

Version identifier: grs_h2o-mars-v0.1.0. Grid: latband5km-v1, 22 band groups, 5,797,982 cells. Grid checks passed. See the layer manifest for full details.

Primary field ice_wt_pct: 4,870,219 finite values; p01 2.0743, median 3.89934, p99 6.47659; range 1.64562–7.397. Complete per-column distributions, units, flags, source references, file paths and SHA-256s are in the manifest.

Smoothed H2O includes bound water and covers a different depth/material from GGD318. Latitude convention is assumed; no absolute Earth pair.

Source: https://pds-geosciences.wustl.edu/missions/odyssey/grs_elements.html. Use and attribution: PDS public data, no licence text; non-commercial use accepted (Enzo, 2026-11-13). Cite Boynton et al. 2007, JGR 112, E12S99.

Eighteen longitude-seam cells match periodic independent interpolation within 2.35e-7 wt%. Across 4,440,330 paired cells FREND−GRS mean is +0.9683 percentage points and Pearson r is 0.7505; differing support, spatial dependence and shared calibration limit interpretation. See the layer manifest for details.

ESA Permafrost CCI: permafrost extent (PFR), Northern Hemisphere (+ Antarctica if present)

Version identifier: permafrost_cci-earth-v0.1.0. Grid: latband5km-v1, 41 band groups, 20,403,725 cells. Grid checks passed. See the layer manifest for full details.

Primary field ice_frac: 1,138,702 finite values; p01 0, median 0.580197, p99 1; range 0–1. Complete per-column distributions, units, flags, source references, file paths and SHA-256s are in the manifest.

PFR is modeled area extent, not ice mass or SWIM consistency; 255 is missing, zero is valid, and both northern and Antarctic products were used.

Source: https://catalogue.ceda.ac.uk/uuid/d235665772ec4b558e9a89ac85595e71. Use and attribution: ESA CCI Data Policy: free and open access; cite the dataset DOI 10.5285/d235665772ec4b558e9a89ac85595e71.

Named-site QA includes Siberia, Tibet and Kansas; the Kansas output is outside coverage. See the layer manifest for details.

Subsurface Water Ice Mapping (SWIM) 2.0 data products

Version identifier: swim2-mars-v0.1.0. Grid: latband5km-v1, 22 band groups, 5,797,982 cells. Grid checks passed. See the layer manifest for full details.

Primary field ice_frac: 5,019,784 finite values; p01 0, median 0.5, p99 1; range 0–1. Complete per-column distributions, units, flags, source references, file paths and SHA-256s are in the manifest.

Positive shallow consistency is model evidence, not a probability or permafrost extent. The documented +1 km source cutoff remains unresolved; cell elevation is only a proxy.

Source: https://swim.psi.edu/SWIM2Products.php. Use and attribution: No licence text on swim.psi.edu; non-commercial use accepted (Enzo, 2026-11-13).

Named-site and longitude-seam QA: equatorial ice_frac 1 represents positive weak source consistency (mean 0.1083), not certain ice occurrence. Ten seam cells match wrapped source windows within 2.98e-8. See the layer manifest for details.

Geomorphometry Earth

41 groups, 20,403,725 exact latband5km-v1 cells. TRI and TPI use metres; geomorphon fractions and elevation coverage are dimensionless. Eight-neighbour Riley TRI, 51×51-pixel TPI, and 3 km geomorphon sightlines use the terrain-v0.2 200 m equalized analysis grid. Flatness is 1° and skip is zero.

All groups pass exact-grid, range, fraction-sum, majority and quality-flag checks. Independent source-cell arithmetic and full distributions are recorded in the manifest. Geomorphons use the documented ten-class count table; GRASS 498-pattern parity remains unverified. Earth DSM canopy/roofs, Mars source resolution, zone/planet-edge NaN halos and polar distortion limit interpretation. Same-name computational twins do not establish physical comparability.

Source: Terrain texture at 200 m (Mars: MOLA-HRSC Blended DEM; Earth: Copernicus GLO-30). Blended DEM terms not recorded (approved for non-commercial use); Copernicus DEM licence ("produced using Copernicus WorldDEM-30" notice). Non-commercial use approved by Enzo 2026-11-13. Version identifier: geomorphometry-earth-v0.1.0. Original per-group source and code hashes are preserved, with an exact source snapshot and reuse evidence.

Geomorphometry Mars

22 groups, 5,797,982 exact latband5km-v1 cells. TRI and TPI use metres; geomorphon fractions and elevation coverage are dimensionless. Eight-neighbour Riley TRI, 51×51-pixel TPI, and 3 km geomorphon sightlines use the terrain-v0.2 200 m equalized analysis grid. Flatness is 1° and skip is zero.

All groups pass exact-grid, range, fraction-sum, majority and quality-flag checks. Independent source-cell arithmetic and full distributions are recorded in the manifest. Geomorphons use the documented ten-class count table; GRASS 498-pattern parity remains unverified. Earth DSM canopy/roofs, Mars source resolution, zone/planet-edge NaN halos and polar distortion limit interpretation. Same-name computational twins do not establish physical comparability.

Source: Terrain texture at 200 m (Mars: MOLA-HRSC Blended DEM; Earth: Copernicus GLO-30). Blended DEM terms not recorded (approved for non-commercial use); Copernicus DEM licence ("produced using Copernicus WorldDEM-30" notice). Non-commercial use approved by Enzo 2026-11-13. Version identifier: geomorphometry-mars-v0.1.0. Original per-group source and code hashes are preserved, with an exact source snapshot and reuse evidence.

SoilGrids v0.2 Earth

41 groups, 20,403,725 exact latband5km-v1 cells. ISRIC SoilGrids 2.0 1000 m 0–5 cm inputs; units are volume/mass percent and µm. Native texture-derived grain diameter is Earth-only and has no validated thermal-grain counterpart.

Independent 96-cell source checks passed with maximum absolute error 1.484e-5 across the checked statistics. The projection-domain fix excludes 276,449 pixels with invalid inverse/forward Homolosine mapping; accepted round trips differ by less than 4.66e-8 m. All 41 output hashes match private HF checkpoints. Missing ocean/ice/Antarctic coverage remains explicit.

Version identifier: soilgrids-earth-v0.2.0. Earlier v0.1 output is excluded after the projection-domain correction. CC BY 4.0; cite ISRIC and Poggio et al. (2021), with exact source metadata in the manifest.

Additional runbook checks

Independent native 250 m versus retained 1000 m means: 32 cells across Atacama and Amazon, four source properties, all 128 pairs valid. Maximum absolute difference 3.307 percentage points; the aggregation kernel and edge masks remain undocumented. The production mask excludes all cells south of 60°S. Canonical terrain coverage and ocean-only counts, with explicit proxy limitations, are in the manifest.

Atacama 1° box: 444/444 texture-covered cells; median fine-earth grain 196.24 µm, coarse fragments 28.30 vol %, sand/clay 65.88/12.13 %. These are descriptive source predictions, not local ground truth.

Amazon 1° box: 469/484 texture-covered cells; median fine-earth grain 19.67 µm, coarse fragments 2.89 vol %, sand/clay 21.09/33.80 %. These are descriptive source predictions, not local ground truth.

Sahara 1° box: 444/444 texture-covered cells; median fine-earth grain 63.67 µm, coarse fragments 33.60 vol %, sand/clay 45.81/24.40 %. These are descriptive source predictions, not local ground truth.

HydroRIVERS v0.3

Release tag hydrorivers-earth-v0.3.0. The 41 Earth tables carry channel density per land area, maximum recomputed Strahler order and distance to the nearest retained channel on the 5 km grid. Use for Earth context. Cross-planet scoring is unapproved. Antarctica is outside source coverage. The source includes ice-sheet artifacts and lake reaches.

Source: HydroRIVERS v1, Lehner and Grill (2013), DOI 10.1002/hyp.9740. HydroSHEDS v1 © World Wildlife Fund, Inc. (2006–2022), used under the HydroSHEDS v1 licence. WWF has not evaluated these derived statistics and provides no warranty for accuracy, completeness, currency or suitability. The source contributors and full notice are listed in Exhibit B of the licence. Licence · Layer manifest

Earth dunes mapped distance v0.4

Release tag earth-dunes-distance-earth-v0.4.0. The 41 tables provide WGS84 distance bounds to retained GSDP30 dune footprints for 20,403,725 Earth 5 km grid cells. Every interval spans at most 5 m. The source coverage flag identifies incomplete observation at the cell center. Use the distance supplement for Earth context. Fraction and type data are separate.

Derived from GSDP30 v1 by Zhijia Zheng, Xiuyuan Zhang, Jiajun Li, Eslam Ali, Jinsongdi Yu and Shihong Du (2024). GSDP30 source · Layer manifest

Mars Climate Database v0.2.1

The 22 Parquet tables contain derived Mars Climate Database v6.2 summaries for 5,797,982 Mars grid cells. These are coarse climate model estimates regridded to 5 km by the v0.2 producer. Numerical and grid checks passed. Matched five-site annual temperature checks, shortwave plausibility and the pressure scale-height approximation remain open. The release supports non-commercial research and model context. Credit the Mars Climate Database, LMD / Mars Planetary Climate Model, and keep LMD informed of usage. Source and terms · Manifest

Earth mapped dune fraction and type v0.4

The 41 tables provide GSDP30 dune fractions, dominant types and mapped support for 20,403,725 Earth grid cells. The release incorporates 72 independently checked coverage corrections, including two cells whose zero fractions became available. Use the published earth-dunes-distance-earth-v0.4.0 tables for certified distance bounds and join by region_id. Source gaps and unresolved patch boundaries can leave values unavailable.

Derived from GSDP30 v1 by Zhijia Zheng, Xiuyuan Zhang, Jiajun Li, Eslam Ali, Jinsongdi Yu and Shihong Du (2024), CC BY 4.0. GSDP30 source · Manifest

Earth MODIS 2017–2021 climatology

The four table sets each cover 20,403,725 Earth grid cells in 41 groups. The sets provide land-surface temperature, albedo and NDVI, apparent thermal inertia (ATI), and a within-Earth bare-cell ATI percentile. Join on grid_version and region_id. Missing or ocean measurements remain null or NaN. The nighttime LST estimate for the 2021 Atacama source pixel is 276.27 K using the equal-weight mean of MYD11C3 and MYD21C3. The site estimate is limited to that pixel and year. The five-year grids use MYD11C3 for LST, and the retrieval disagreement remains unexplained.

Manifest · Atacama site estimate · Verification · NASA MYD11C3 · NASA MCD43C3 · NASA MYD13C2 · NASA MYD21C3

Earth and Mars K–H roughness v0.2.1

This release contains 41 Earth tables (20,403,725 grid cells) and 22 Mars tables (5,797,982 grid cells). Join on grid_version and region_id. Some cells have missing roughness where source support is absent. The release preserves the v0.2 producer values and flags. The longest-baseline field rough_9p6 uses 9.6 km on Earth and the published 9.2 km map label on Mars. Mars map registration and shot selection, the baseline difference and cross-planet normalization remain open. Use the fields as provisional planet-specific context and filters. Cross-planet scoring requires further validation.

Mars is derived from Mikhail Kreslavsky's MOLA K–H roughness maps, CC BY 4.0. Earth is derived from Copernicus DEM GLO-30. The attribution in the Earth terrain section applies. Mars source · Earth source · Manifest

Downloads last month
171