-
-
Notifications
You must be signed in to change notification settings - Fork 56
/
05-raster-vector.qmd
1075 lines (852 loc) · 45.7 KB
/
05-raster-vector.qmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
---
jupyter: python3
---
# Raster-vector interactions {#sec-raster-vector}
## Prerequisites {.unnumbered}
```{python}
#| echo: false
import book_options
```
```{python .content-visible when-format="pdf"}
#| echo: false
import book_options_pdf
```
This chapter requires importing the following packages:
```{python}
import os
import math
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
import shapely
import geopandas as gpd
import rasterio
import rasterio.plot
import rasterio.mask
import rasterio.features
import rasterstats
```
It also relies on the following data files:
```{python}
src_srtm = rasterio.open('data/srtm.tif')
src_nlcd = rasterio.open('data/nlcd.tif')
src_grain = rasterio.open('output/grain.tif')
src_elev = rasterio.open('output/elev.tif')
src_dem = rasterio.open('data/dem.tif')
zion = gpd.read_file('data/zion.gpkg')
zion_points = gpd.read_file('data/zion_points.gpkg')
cycle_hire_osm = gpd.read_file('data/cycle_hire_osm.gpkg')
us_states = gpd.read_file('data/us_states.gpkg')
nz = gpd.read_file('data/nz.gpkg')
src_nz_elev = rasterio.open('data/nz_elev.tif')
```
## Introduction
This chapter focuses on interactions between raster and vector geographic data models, both introduced in @sec-spatial-class.
It includes three main techniques:
- Raster cropping and masking using vector objects (@sec-raster-cropping)
- Extracting raster values using different types of vector data (@sec-raster-extraction)
- Raster-vector conversion (@sec-rasterization and @sec-spatial-vectorization)
These concepts are demonstrated using data from previous chapters, to understand their potential real-world applications.
## Raster masking and cropping {#sec-raster-cropping}
Many geographic data projects involve integrating data from many different sources, such as remote sensing images (rasters) and administrative boundaries (vectors).
Often the extent of input raster datasets is larger than the area of interest.
In this case, raster *masking*, *cropping*, or both, are useful for unifying the spatial extent of input data (@fig-raster-crop (b) and (c), and the following two examples, illustrate the difference between masking and cropping).
Both operations reduce object memory use and associated computational resources for subsequent analysis steps, and may be a necessary preprocessing step before creating attractive maps involving raster data.
We will use two layers to illustrate raster cropping:
- The `srtm.tif` raster representing elevation, in meters above sea level, in south-western Utah: a **rasterio** file connection named `src_srtm` (see @fig-raster-crop (a))
- The `zion.gpkg` vector layer representing the Zion National Park boundaries (a `GeoDataFrame` named `zion`)
Both target and cropping objects must have the same projection.
Since it is easier and more precise to reproject vector layers, compared to rasters, we use the following expression to reproject (@sec-reprojecting-vector-geometries) the vector layer `zion` into the CRS of the raster `src_srtm`.
```{python}
zion = zion.to_crs(src_srtm.crs)
```
To mask the image, i.e., convert all pixels which do not intersect with the `zion` polygon to 'No Data', we use the `rasterio.mask.mask` function.
```{python}
out_image_mask, out_transform_mask = rasterio.mask.mask(
src_srtm,
zion.geometry,
crop=False,
nodata=9999
)
```
Note that we need to choose and specify a 'No Data' value, within the valid range according to the data type.
Since `srtm.tif` is of type `uint16` (how can we check?), we choose `9999` (a positive integer that is guaranteed not to occur in the raster).
Also note that **rasterio** does not directly support **geopandas** data structures, so we need to pass a 'collection' of **shapely** geometries: a `GeoSeries` (see above) or a `list` of **shapely** geometries (see next example) both work.
The output consists of two objects.
The first one is the `out_image` array with the masked values.
```{python}
out_image_mask
```
The second one is a new transformation matrix `out_transform`.
```{python}
out_transform_mask
```
Note that masking (without cropping!) does not modify the raster extent.
Therefore, the new transform is identical to the original (`src_srtm.transform`).
Unfortunately, the `out_image` and `out_transform` objects do not contain any information indicating that `9999` represents 'No Data'.
To associate the information with the raster, we must write it to file along with the corresponding metadata.
For example, to write the masked raster to file, we first need to modify the 'No Data' setting in the metadata.
```{python}
dst_kwargs = src_srtm.meta
dst_kwargs.update(nodata=9999)
dst_kwargs
```
Then we can write the masked raster to file with the updated metadata object.
```{python}
new_dataset = rasterio.open('output/srtm_masked.tif', 'w', **dst_kwargs)
new_dataset.write(out_image_mask)
new_dataset.close()
```
Now we can re-import the raster and check that the 'No Data' value is correctly set.
```{python}
src_srtm_mask = rasterio.open('output/srtm_masked.tif')
```
The `.meta` property contains the `nodata` entry.
Now, any relevant operation (such as plotting, see @fig-raster-crop (b)) will take 'No Data' into account.
```{python}
src_srtm_mask.meta
```
The related operation, cropping, reduces the raster extent to the extent of the vector layer:
* To crop *and* mask, we can use `rasterio.mask.mask`, same as above for masking, while setting `crop=True` (@fig-raster-crop (d))
* To just crop, *without* masking, we can derive the bounding box polygon of the vector layer, and then crop using that polygon, also combined with `crop=True` (@fig-raster-crop (c))
For the example of cropping only, the extent polygon of `zion` can be obtained as a `shapely` geometry object using `.union_all().envelope` (@fig-zion-bbox).
```{python}
#| label: fig-zion-bbox
#| fig-cap: Bounding box `'Polygon'` geometry of the `zion` layer
bb = zion.union_all().envelope
bb
```
The extent can now be used for masking.
Here, we are also using the `all_touched=True` option, so that pixels which are partially overlapping with the extent are also included in the output.
```{python}
out_image_crop, out_transform_crop = rasterio.mask.mask(
src_srtm,
[bb],
crop=True,
all_touched=True,
nodata=9999
)
```
In the case of cropping, there is no particular reason to write the result to file for easier plotting, such as in the other two examples, since there are no 'No Data' values (@fig-raster-crop (c)).
::: callout-note
As mentioned above, **rasterio** functions typically accept vector geometries in the form of `list`s of `shapely` objects. `GeoSeries` are conceptually very similar, and also accepted. However, even an individual geometry has to be in a `list`, which is why we pass `[bb]`, and not `bb`, in the above `rasterio.mask.mask` function call (the latter would raise an error).
:::
Finally, the third example is where we perform both crop and mask operations, using `rasterio.mask.mask` with `crop=True` passing `zion.geometry`.
```{python}
out_image_mask_crop, out_transform_mask_crop = rasterio.mask.mask(
src_srtm,
zion.geometry,
crop=True,
nodata=9999
)
```
When writing the result to a file, it is here crucial to update the transform and dimensions, since they were modified as a result of cropping.
Also note that `out_image_mask_crop` is a three-dimensional array (even though it has one band in this case), so the number of rows and columns are in `.shape[1]` and `.shape[2]` (rather than `.shape[0]` and `.shape[1]`), respectively.
```{python}
dst_kwargs = src_srtm.meta
dst_kwargs.update({
'nodata': 9999,
'transform': out_transform_mask_crop,
'width': out_image_mask_crop.shape[2],
'height': out_image_mask_crop.shape[1]
})
new_dataset = rasterio.open(
'output/srtm_masked_cropped.tif',
'w',
**dst_kwargs
)
new_dataset.write(out_image_mask_crop)
new_dataset.close()
```
Let's also create a file connection to the newly created file `srtm_masked_cropped.tif` in order to plot it (@fig-raster-crop (d)).
```{python}
src_srtm_mask_crop = rasterio.open('output/srtm_masked_cropped.tif')
```
@fig-raster-crop shows the original raster, and the three masking and/or cropping results.
```{python}
#| label: fig-raster-crop
#| fig-cap: Raster masking and cropping
#| layout-ncol: 2
#| fig-subcap:
#| - Original
#| - Masked
#| - Cropped
#| - Masked+Cropped
# Original
fig, ax = plt.subplots(figsize=(3.5, 3.5))
rasterio.plot.show(src_srtm, ax=ax)
zion.plot(ax=ax, color='none', edgecolor='black');
# Masked
fig, ax = plt.subplots(figsize=(3.5, 3.5))
rasterio.plot.show(src_srtm_mask, ax=ax)
zion.plot(ax=ax, color='none', edgecolor='black');
# Cropped
fig, ax = plt.subplots(figsize=(3.5, 3.5))
rasterio.plot.show(out_image_crop, transform=out_transform_crop, ax=ax)
zion.plot(ax=ax, color='none', edgecolor='black');
# Masked+Cropped
fig, ax = plt.subplots(figsize=(3.5, 3.5))
rasterio.plot.show(src_srtm_mask_crop, ax=ax)
zion.plot(ax=ax, color='none', edgecolor='black');
```
## Raster extraction {#sec-raster-extraction}
Raster extraction is the process of identifying and returning the values associated with a 'target' raster at specific locations, based on a (typically vector) geographic 'selector' object.
The reverse of raster extraction---assigning raster cell values based on vector objects---is rasterization, described in @sec-rasterization.
In the following examples, we use a package called **rasterstats**, which is specifically aimed at extracting raster values:
* To *points* (@sec-extraction-to-points) or to *lines* (@sec-extraction-to-lines), via the `rasterstats.point_query` function
* To *polygons* (@sec-extraction-to-polygons), via the `rasterstats.zonal_stats` function
### Extraction to points {#sec-extraction-to-points}
The simplest type of raster extraction is getting the values of raster cells at specific points.
To demonstrate extraction to points, we will use `zion_points`, which contains a sample of 30 locations within the Zion National Park (@fig-zion-points).
```{python}
#| label: fig-zion-points
#| fig-cap: 30-point locations within the Zion National Park, with elevation in the background
fig, ax = plt.subplots()
rasterio.plot.show(src_srtm, ax=ax)
zion_points.plot(ax=ax, color='black', edgecolor='white');
```
The following expression extracts elevation values from `srtm.tif` according to `zion_points`, using `rasterstats.point_query`.
```{python}
result1 = rasterstats.point_query(
zion_points,
src_srtm.read(1),
nodata = src_srtm.nodata,
affine = src_srtm.transform,
interpolate='nearest'
)
```
The first two arguments are the vector layer and the array with raster values.
The `nodata` and `affine` arguments are used to align the array values into the CRS, and to correctly treat 'No Data' flags.
Finally, the `interpolate` argument controls the way that the cell values are assigned to the point; `interpolate='nearest'` typically makes more sense, as opposed to the other option `interpolate='bilinear'` which is the default.
Alternatively, we can pass a raster file path to `rasterstats.point_query`, in which case `nodata` and `affine` are not necessary, as the function can understand those properties directly from the raster file.
```{python}
result2 = rasterstats.point_query(
zion_points,
'data/srtm.tif',
interpolate='nearest'
)
```
Either way, the resulting object is a `list` of raster values, corresponding to `zion_points`.
For example, here are the elevations of the first five points.
```{python}
result1[:5]
```
To get a `GeoDataFrame` with the original points geometries (and other attributes, if any), as well as the extracted raster values, we can assign the extraction result into a new column.
As you can see, both approaches give the same result.
```{python}
zion_points['elev1'] = result1
zion_points['elev2'] = result2
zion_points
```
The function supports extracting from just one raster band at a time.
When passing an array, we can read the required band (as in, `.read(1)`, `.read(2)`, etc.).
When passing a raster file path, we can set the band using the `band_num` argument (the default being `band_num=1`).
### Extraction to lines {#sec-extraction-to-lines}
Raster extraction is also applicable with line selectors.
The typical line extraction algorithm is to extract one value for each raster cell touched by a line.
However, this particular approach is not recommended to obtain values along the transects, as it is hard to get the correct distance between each pair of extracted raster values.
For line extraction, a better approach is to split the line into many points (at equal distances along the line) and then extract the values for these points using the 'extraction to points' technique (@sec-extraction-to-points).
To demonstrate this, the code below creates (see @sec-vector-data for recap) `zion_transect`, a straight line going from northwest to southeast of the Zion National Park.
```{python}
coords = [[-113.2, 37.45], [-112.9, 37.2]]
zion_transect = shapely.LineString(coords)
print(zion_transect)
```
The utility of extracting heights from a linear selector is illustrated by imagining that you are planning a hike.
The method demonstrated below provides an 'elevation profile' of the route (the line does not need to be straight), useful for estimating how long it will take due to long climbs.
First, we need to create a layer consisting of points along our line (`zion_transect`), at specified intervals (e.g., `250`).
To do that, we need to transform the line into a projected CRS (so that we work with true distances, in $m$), such as UTM.
This requires going through a `GeoSeries`, as **shapely** geometries have no CRS definition nor concept of reprojection (see @sec-vector-layer-from-scratch).
```{python}
zion_transect_utm = gpd.GeoSeries(zion_transect, crs=4326).to_crs(32612)
zion_transect_utm = zion_transect_utm.iloc[0]
```
The printout of the new geometry shows this is still a straight line between two points, only with coordinates in a projected CRS.
```{python}
print(zion_transect_utm)
```
Next, we need to calculate the distances, along the line, where points are going to be generated.
We do this using `np.arange`.
The result is a numeric sequence starting at `0`, going up to line `.length`, in steps of `250` ($m$).
```{python}
distances = np.arange(0, zion_transect_utm.length, 250)
distances[:7] ## First 7 distance cutoff points
```
The distance cutoffs are used to sample ('interpolate') points along the line.
The **shapely** `.interpolate` method is used to generate the points, which then are reprojected back to the geographic CRS of the raster (EPSG:`4326`).
```{python}
#| code-overflow: wrap
zion_transect_pnt = [zion_transect_utm.interpolate(d) for d in distances]
zion_transect_pnt = gpd.GeoSeries(zion_transect_pnt, crs=32612) \
.to_crs(src_srtm.crs)
zion_transect_pnt
```
Finally, we extract the elevation values for each point in our transect and combine the information with `zion_transect_pnt` (after 'promoting' it to a `GeoDataFrame`, to accommodate extra attributes), using the point extraction method shown earlier (@sec-extraction-to-points).
We also attach the respective distance cutoff points `distances`.
```{python}
result = rasterstats.point_query(
zion_transect_pnt,
src_srtm.read(1),
nodata = src_srtm.nodata,
affine = src_srtm.transform,
interpolate='nearest'
)
zion_transect_pnt = gpd.GeoDataFrame(geometry=zion_transect_pnt)
zion_transect_pnt['dist'] = distances
zion_transect_pnt['elev'] = result
zion_transect_pnt
```
The information in `zion_transect_pnt`, namely the `'dist'` and `'elev'` attributes, can now be used to draw an elevation profile, as illustrated in @fig-zion-transect.
```{python}
#| label: fig-zion-transect
#| fig-cap: Extracting a raster values profile to line
#| layout-ncol: 2
#| fig-subcap:
#| - Raster and a line transect
#| - Extracted elevation profile
# Raster and a line transect
fig, ax = plt.subplots()
rasterio.plot.show(src_srtm, ax=ax)
gpd.GeoSeries(zion_transect).plot(ax=ax, color='black')
zion.plot(ax=ax, color='none', edgecolor='white');
# Elevation profile
fig, ax = plt.subplots()
zion_transect_pnt.set_index('dist')['elev'].plot(ax=ax)
ax.set_xlabel('Distance (m)')
ax.set_ylabel('Elevation (m)');
```
### Extraction to polygons {#sec-extraction-to-polygons}
The final type of geographic vector object for raster extraction is polygons.
Like lines, polygons tend to return many raster values per vector geometry.
For continuous rasters (@fig-raster-extract-to-polygon (a)), we typically want to generate summary statistics for raster values per polygon, for example to characterize a single region or to compare many regions.
The generation of raster summary statistics, by polygons, is demonstrated in the code below using `rasterstats.zonal_stats`, which creates a list of summary statistics (in this case a list of length 1, since there is just one polygon).
```{python}
result = rasterstats.zonal_stats(
zion,
src_srtm.read(1),
nodata = src_srtm.nodata,
affine = src_srtm.transform,
stats = ['mean', 'min', 'max']
)
result
```
::: callout-note
`rasterstats.zonal_stats`, just like `rasterstats.point_query` (@sec-extraction-to-points), supports raster input as file paths, rather than arrays plus `nodata` and `affine` arguments.
:::
Transformation of the `list` to a `DataFrame` (e.g., to attach the derived attributes to the original polygon layer), is straightforward with the `pd.DataFrame` constructor.
```{python}
pd.DataFrame(result)
```
Because there is only one polygon in the example, a `DataFrame` with a single row is returned.
However, if `zion` was composed of more than one polygon, we would accordingly get more rows in the `DataFrame`.
The result provides useful summaries, for example that the maximum height in the park is `2661` $m$ above see level.
Note the `stats` argument, where we determine what type of statistics are calculated per polygon.
Possible values other than `'mean'`, `'min'`, and `'max'` include:
- `'count'`---The number of valid (i.e., excluding 'No Data') pixels
- `'nodata'`---The number of pixels with 'No Data'
- `'majority'`---The most frequently occurring value
- `'median'`---The median value
See the documentation of `rasterstats.zonal_stats` for the complete list.
Additionally, the `rasterstats.zonal_stats` function accepts user-defined functions for calculating any custom statistics.
To count occurrences of categorical raster values within polygons (@fig-raster-extract-to-polygon (b)), we can use masking (@sec-raster-cropping) combined with `np.unique`, as follows.
```{python}
out_image, out_transform = rasterio.mask.mask(
src_nlcd,
zion.geometry.to_crs(src_nlcd.crs),
crop=False,
nodata=src_nlcd.nodata
)
counts = np.unique(out_image, return_counts=True)
counts
```
According to the result, for example, the value `2` ('Developed' class) appears in `4205` pixels within the Zion polygon.
@fig-raster-extract-to-polygon illustrates the two types of raster extraction to polygons described above.
```{python}
#| label: fig-raster-extract-to-polygon
#| fig-cap: Sample data used for continuous and categorical raster extraction to a polygon
#| layout-ncol: 2
#| fig-subcap:
#| - Continuous raster
#| - Categorical raster
# Continuous raster
fig, ax = plt.subplots()
rasterio.plot.show(src_srtm, ax=ax)
zion.plot(ax=ax, color='none', edgecolor='black');
# Categorical raster
fig, ax = plt.subplots()
rasterio.plot.show(src_nlcd, ax=ax, cmap='Set3')
zion.to_crs(src_nlcd.crs).plot(ax=ax, color='none', edgecolor='black');
```
<!-- jn: what is the state of plotting categorical rasters? can it read the color palette from a file? -->
<!-- md: admittedly I've never used this functionality in either R or Python... If you have a sample data file I'll be happy to experiment with it. -->
## Rasterization {#sec-rasterization}
Rasterization is the conversion of vector objects into their representation in raster objects.
Usually, the output raster is used for quantitative analysis (e.g., analysis of terrain) or modeling.
As we saw in @sec-spatial-class, the raster data model has some characteristics that make it conducive to certain methods.
Furthermore, the process of rasterization can help simplify datasets because the resulting values all have the same spatial resolution: rasterization can be seen as a special type of geographic data aggregation.
The **rasterio** package contains the `rasterio.features.rasterize` function for doing this work.
To make it happen, we need to have the 'template' grid definition, i.e., the 'template' raster defining the extent, resolution and CRS of the output, in the `out_shape` (the output dimensions) and `transform` (the transformation matrix) arguments of `rasterio.features.rasterize`.
In case we have an existing template raster, we simply need to query its `.shape` and `.transform`.
On the other hand, if we need to create a custom template, e.g., covering the vector layer extent with specified resolution, there is some extra work to calculate both of these objects (see next example).
As for the vector geometries and their associated values, the `rasterio.features.rasterize` function requires the input vector shapes in the form of an iterable object of `geometry,value` pairs, where:
- `geometry` is the given geometry (**shapely** geometry object)
- `value` is the value to be 'burned' into pixels coinciding with the geometry (`int` or `float`)
Furthermore, we define how to deal with multiple values burned into the same pixel, using the `merge_alg` parameter.
The default `merge_alg=rasterio.enums.MergeAlg.replace` means that 'later' values replace 'earlier' ones, i.e., the pixel gets the 'last' burned value.
The other option `merge_alg=rasterio.enums.MergeAlg.add` means that burned values are summed, i.e., the pixel gets the sum of all burned values.
When rasterizing lines and polygons, we also have the choice between two pixel-matching algorithms.
The default, `all_touched=False`, implies pixels that are selected by Bresenham's line algorithm[^bresenham] (for lines) or pixels whose center is within the polygon (for polygons).
The other option `all_touched=True`, as the name suggests, implies that all pixels intersecting with the geometry are matched.
[^bresenham]: [https://en.wikipedia.org/wiki/Bresenham%27s_line_algorithm](https://en.wikipedia.org/wiki/Bresenham%27s_line_algorithm)
Finally, we can set the `fill` value, which is the value that 'unaffected' pixels get, with `fill=0` being the default.
How the `rasterio.features.rasterize` function works with all of these various parameters will be made clear in the next examples.
The geographic resolution of the 'template' raster has a major impact on the results: if it is too low (cell size is too large), the result may miss the full geographic variability of the vector data; if it is too high, computational times may be excessive.
There are no simple rules to follow when deciding an appropriate geographic resolution, which is heavily dependent on the intended use of the results.
Often the target resolution is imposed on the user, for example when the output of rasterization needs to be aligned to an existing raster.
Depending on the input data, rasterization typically takes one of two forms which we demonstrate next:
- in *point* rasterization (@sec-rasterizing-points), we typically choose how to treat multiple points: either to summarize presence/absence, point count, or summed attribute values (@fig-rasterize-points)
- in *line* and *polygon* rasterization (@sec-rasterizing-lines-and-polygons), there are typically no such 'overlaps' and we simply 'burn' attribute values, or fixed values, into pixels coinciding with the given geometries (@fig-rasterize-lines-polygons)
### Rasterizing points {#sec-rasterizing-points}
To demonstrate point rasterization, we will prepare a 'template' raster that has the same extent and CRS as the input vector data `cycle_hire_osm_projected` (a dataset on cycle hire points in London, illustrated in @fig-rasterize-points (a)) and a spatial resolution of 1000 $m$.
To do that, we first take our point layer and transform it to a projected CRS.
```{python}
cycle_hire_osm_projected = cycle_hire_osm.to_crs(27700)
```
Next, we calculate the `out_shape` and `transform` of the template raster.
To calculate the transform, we combine the top-left corner of the `cycle_hire_osm_projected` bounding box with the required resolution (e.g., 1000 $m$).
```{python}
bounds = cycle_hire_osm_projected.total_bounds
res = 1000
transform = rasterio.transform.from_origin(
west=bounds[0],
north=bounds[3],
xsize=res,
ysize=res
)
transform
```
To calculate the `out_shape`, we divide the x-axis and y-axis extent by the resolution, taking the ceiling of the results.
```{python}
rows = math.ceil((bounds[3] - bounds[1]) / res)
cols = math.ceil((bounds[2] - bounds[0]) / res)
shape = (rows, cols)
shape
```
Finally, we are ready to rasterize.
As mentioned above, point rasterization can be a very flexible operation: the results depend not only on the nature of the template raster, but also on the pixel 'activation' method, namely the way we deal with multiple points matching the same pixel.
To illustrate this flexibility, we will try three different approaches to point rasterization (@fig-rasterize-points (b)-(d)).
First, we create a raster representing the presence or absence of cycle hire points (known as presence/absence rasters).
In this case, we transfer the value of `1` to all pixels where at least one point falls in.
In the **rasterio** framework, we use the `rasterio.features.rasterize` function, which requires an iterable object of `geometry,value` pairs.
In this first example, we transform the point `GeoDataFrame` into a `list` of `shapely` geometries and the (fixed) value of `1`, using list comprehension, as follows.
The first five elements of the `list` are hereby printed to illustrate its structure.
```{python}
g = [(g, 1) for g in cycle_hire_osm_projected.geometry]
g[:5]
```
The list of `geometry,value` pairs is passed to `rasterio.features.rasterize`, along with the `out_shape` and `transform` which define the raster template.
The result `ch_raster1` is an `ndarray` with the burned values of `1` where the pixel coincides with at least one point, and `0` in 'unaffected' pixels.
Note that `merge_alg=rasterio.enums.MergeAlg.replace` (the default) is used here, which means that a pixel gets `1` when one or more points fall in it, or keeps the original `0` value otherwise.
```{python}
ch_raster1 = rasterio.features.rasterize(
shapes=g,
out_shape=shape,
transform=transform
)
ch_raster1
```
In our second variant of point rasterization, we count the number of bike hire stations.
To do that, we use the fixed value of `1` (same as in the last example), but this time combined with the `merge_alg=rasterio.enums.MergeAlg.add` argument.
That way, multiple values burned into the same pixel are *summed*, rather than replaced keeping last (which is the default).
The new output, `ch_raster2`, shows the number of cycle hire points in each grid cell.
```{python}
g = [(g, 1) for g in cycle_hire_osm_projected.geometry]
ch_raster2 = rasterio.features.rasterize(
shapes=g,
out_shape=shape,
transform=transform,
merge_alg=rasterio.enums.MergeAlg.add
)
ch_raster2
```
The cycle hire locations have different numbers of bicycles described by the capacity variable, raising the question, what is the capacity in each grid cell?
To calculate that, in our third point rasterization variant we sum the field (`'capacity'`) rather than the fixed values of `1`.
This requires using a more complex list comprehension expression, where we also (1) extract both geometries and the attribute of interest, and (2) filter out 'No Data' values, which can be done as follows.
You are invited to run the separate parts to see how this works; the important point is that, in the end, we get the list `g` with the `geometry,value` pairs to be burned, only that the `value` is now variable, rather than fixed, among points.
```{python}
g = [(g, v) for g, v in cycle_hire_osm_projected[['geometry', 'capacity']] \
.dropna(subset='capacity')
.to_numpy() \
.tolist()]
g[:5]
```
Now we rasterize the points, again using `merge_alg=rasterio.enums.MergeAlg.add` to sum the capacity values per pixel.
```{python}
ch_raster3 = rasterio.features.rasterize(
shapes=g,
out_shape=shape,
transform=transform,
merge_alg=rasterio.enums.MergeAlg.add
)
ch_raster3
```
The result `ch_raster3` shows the total capacity of cycle hire points in each grid cell.
The input point layer `cycle_hire_osm_projected` and the three variants of rasterizing it `ch_raster1`, `ch_raster2`, and `ch_raster3` are shown in @fig-rasterize-points.
```{python}
#| label: fig-rasterize-points
#| fig-cap: Original data and three variants of point rasterization
#| layout-ncol: 2
#| fig-subcap:
#| - Input points
#| - Presence/Absence
#| - Point counts
#| - Summed attribute values
# Input points
fig, ax = plt.subplots()
cycle_hire_osm_projected.plot(column='capacity', ax=ax);
# Presence/Absence
fig, ax = plt.subplots()
rasterio.plot.show(ch_raster1, transform=transform, ax=ax);
# Point counts
fig, ax = plt.subplots()
rasterio.plot.show(ch_raster2, transform=transform, ax=ax);
# Summed attribute values
fig, ax = plt.subplots()
rasterio.plot.show(ch_raster3, transform=transform, ax=ax);
```
### Rasterizing lines and polygons {#sec-rasterizing-lines-and-polygons}
Another dataset based on California's polygons and borders (created below) illustrates rasterization of lines.
There are three preliminary steps.
First, we subset the California polygon.
```{python}
california = us_states[us_states['NAME'] == 'California']
california
```
Second, we 'cast' the polygon into a `'MultiLineString'` geometry, using the `.boundary` property that `GeoSeries` and `DataFrame`s have.
```{python}
california_borders = california.boundary
california_borders
```
Third, we create the `transform` and `shape` describing our template raster, with a resolution of `0.5` degree, using the same approach as in @sec-rasterizing-points.
```{python}
bounds = california_borders.total_bounds
res = 0.5
transform = rasterio.transform.from_origin(
west=bounds[0],
north=bounds[3],
xsize=res,
ysize=res
)
rows = math.ceil((bounds[3] - bounds[1]) / res)
cols = math.ceil((bounds[2] - bounds[0]) / res)
shape = (rows, cols)
shape
```
Finally, we rasterize `california_borders` based on the calculated template's `shape` and `transform`.
When considering line or polygon rasterization, one useful additional argument is `all_touched`.
By default it is `False`, but when changed to `True`---all cells that are touched by a line or polygon border get a value.
Line rasterization with `all_touched=True` is demonstrated in the code below (@fig-rasterize-lines-polygons, left).
We are also using `fill=np.nan` to set 'background' values to 'No Data'.
```{python}
california_raster1 = rasterio.features.rasterize(
[(g, 1) for g in california_borders],
out_shape=shape,
transform=transform,
all_touched=True,
fill=np.nan,
dtype=np.float64
)
```
Compare it to polygon rasterization, with `all_touched=False` (the default), which selects only raster cells whose centroids are inside the selector polygon, as illustrated in @fig-rasterize-lines-polygons (right).
```{python}
california_raster2 = rasterio.features.rasterize(
[(g, 1) for g in california.geometry],
out_shape=shape,
transform=transform,
fill=np.nan,
dtype=np.float64
)
```
To illustrate which raster pixels are actually selected as part of rasterization, we also show them as points.
This also requires the following code section to calculate the points, which we explain in @sec-spatial-vectorization.
```{python}
height = california_raster1.shape[0]
width = california_raster1.shape[1]
cols, rows = np.meshgrid(np.arange(width), np.arange(height))
x, y = rasterio.transform.xy(transform, rows, cols)
x = np.array(x).flatten()
y = np.array(y).flatten()
z = california_raster1.flatten()
geom = gpd.points_from_xy(x, y, crs=california.crs)
pnt = gpd.GeoDataFrame(data={'value':z}, geometry=geom)
pnt
```
@fig-rasterize-lines-polygons shows the input vector layer, the rasterization results, and the points `pnt`.
```{python}
#| label: fig-rasterize-lines-polygons
#| fig-cap: Examples of line and polygon rasterization
#| layout-ncol: 2
#| fig-subcap:
#| - Line rasterization w/ `all_touched=True`
#| - Polygon rasterization w/ `all_touched=False`
# Line rasterization
fig, ax = plt.subplots()
rasterio.plot.show(california_raster1, transform=transform, ax=ax, cmap='Set3')
gpd.GeoSeries(california_borders).plot(ax=ax, edgecolor='darkgrey', linewidth=1)
pnt.plot(ax=ax, color='black', markersize=1);
# Polygon rasterization
fig, ax = plt.subplots()
rasterio.plot.show(california_raster2, transform=transform, ax=ax, cmap='Set3')
california.plot(ax=ax, color='none', edgecolor='darkgrey', linewidth=1)
pnt.plot(ax=ax, color='black', markersize=1);
```
## Spatial vectorization {#sec-spatial-vectorization}
Spatial vectorization is the counterpart of rasterization (@sec-rasterization).
It involves converting spatially continuous raster data into spatially discrete vector data such as points, lines, or polygons.
There are three standard methods to convert a raster to a vector layer, which we cover next:
- Raster to polygons (@sec-raster-to-polygons)---converting raster cells to rectangular polygons, representing pixel areas
- Raster to points (@sec-raster-to-points)---converting raster cells to points, representing pixel centroids
- Raster to contours (@sec-raster-to-contours)
Let us demonstrate all three in the given order.
### Raster to polygons {#sec-raster-to-polygons}
The `rasterio.features.shapes` gives access to raster pixels as polygon geometries, along with the associated raster values.
The returned object is a generator (see note in @sec-spatial-subsetting-raster), yielding `geometry,value` pairs.
For example, the following expression returns a generator named `shapes`, referring to the pixel polygons.
```{python}
shapes = rasterio.features.shapes(rasterio.band(src_grain, 1))
shapes
```
We can generate all shapes at once into a `list` named `pol` with `list(shapes)`.
```{python}
pol = list(shapes)
```
Each element in `pol` is a `tuple` of length 2, containing the GeoJSON-like `dict`---representing the polygon geometry and the value of the pixel(s) which comprise the polygon.
For example, here is the first element of `pol`.
```{python}
pol[0]
```
::: callout-note
Note that, when transforming a raster cell into a polygon, five-coordinate pairs need to be kept in memory to represent its geometry (explaining why rasters are often fast compared with vectors!).
:::
To transform the `list` coming out of `rasterio.features.shapes` into the familiar `GeoDataFrame`, we need few more steps of data reshaping.
First, we apply the `shapely.geometry.shape` function to go from a `list` of GeoJSON-like `dict`s to a `list` of `shapely` geometry objects.
The `list` can then be converted to a `GeoSeries` (see @sec-vector-layer-from-scratch).
```{python}
geom = [shapely.geometry.shape(i[0]) for i in pol]
geom = gpd.GeoSeries(geom, crs=src_grain.crs)
geom
```
The values can also be extracted from the `rasterio.features.shapes` result and turned into a corresponding `Series`.
```{python}
values = [i[1] for i in pol]
values = pd.Series(values)
values
```
Finally, the two can be combined into a `GeoDataFrame`, hereby named `result`.
```{python}
result = gpd.GeoDataFrame({'value': values, 'geometry': geom})
result
```
The polygon layer `result` is shown in @fig-raster-to-polygons.
```{python}
#| label: fig-raster-to-polygons
#| fig-cap: '`grain.tif` converted to a polygon layer'
result.plot(column='value', edgecolor='black', legend=True);
```
As highlighted using `edgecolor='black'`, neighboring pixels sharing the same raster value are dissolved into larger polygons.
The `rasterio.features.shapes` function unfortunately does not offer a way to avoid this type of dissolving.
One [suggestion](https://gis.stackexchange.com/questions/455980/vectorizing-all-pixels-as-separate-polygons-using-rasterio#answer-456251) is to add unique values between `0` and `0.9999` to all pixels, convert to polygons, and then get back to the original values using `np.floor`.
### Raster to points {#sec-raster-to-points}
To transform a raster to points, we can use the `rasterio.transform.xy` function.
As the name suggests, the function accepts row and column indices, and transforms them into x- and y-coordinates (using the raster's transformation matrix).
For example, the coordinates of the top-left pixel can be calculated passing the `(row,col)` indices of `(0,0)`.
```{python}
src = rasterio.open('output/elev.tif')
rasterio.transform.xy(src.transform, 0, 0)
```
::: callout-note
Keep in mind that the coordinates of the top-left pixel (`(-1.25, 1.25)`), as calculated in the above expression, refer to the pixel *centroid*.
Therefore, they are not identical to the raster origin coordinates (`(-1.5,1.5)`), as specified in the transformation matrix, which are the coordinates of the top-left edge/corner of the raster (see @fig-raster-to-points).
```{python}
src.transform
```
:::
To generalize the above expression to calculate the coordinates of *all* pixels, we first need to generate a grid of all possible row/column index combinations.
This can be done using `np.meshgrid`, as follows.
```{python}
height = src.shape[0]
width = src.shape[1]
cols, rows = np.meshgrid(np.arange(width), np.arange(height))
```
We now have two arrays, `rows` and `cols`, matching the shape of `elev.tif` and containing the corresponding row and column indices.
```{python}
rows
```
```{python}
cols
```
These can be passed to `rasterio.transform.xy` to transform the indices into point coordinates, accordingly stored in lists of arrays `x` and `y`.
```{python}
x, y = rasterio.transform.xy(src.transform, rows, cols)
```
```{python}
x
```
```{python}
y
```
Typically we want to work with the points in the form of a `GeoDataFrame` which also holds the attribute(s) value(s) as point attributes.
To get there, we can transform the coordinates as well as any attributes to 1-dimensional arrays, and then use methods we are already familiar with (@sec-vector-layer-from-scratch) to combine them into a `GeoDataFrame`.
```{python}
x = np.array(x).flatten()
y = np.array(y).flatten()
z = src.read(1).flatten()
geom = gpd.points_from_xy(x, y, crs=src.crs)
pnt = gpd.GeoDataFrame(data={'value':z}, geometry=geom)
pnt
```
This 'high-level' workflow, like many other **rasterio**-based workflows covered in the book, is a commonly used one but lacking from the package itself.
From the user's perspective, it may be a good idea to wrap the workflow into a function (e.g., `raster_to_points(src)`, returning a `GeoDataFrame`), to be re-used whenever we need it.
@fig-raster-to-points shows the input raster and the resulting point layer.
```{python}
#| label: fig-raster-to-points
#| fig-cap: Raster and point representation of `elev.tif`
#| layout-ncol: 2
#| fig-subcap:
#| - Input raster
#| - Points
# Input raster
fig, ax = plt.subplots()
pnt.plot(column='value', legend=True, ax=ax)
rasterio.plot.show(src_elev, ax=ax);
# Points
fig, ax = plt.subplots()
pnt.plot(column='value', legend=True, edgecolor='black', ax=ax)
rasterio.plot.show(src_elev, alpha=0, ax=ax);
```
Note that 'No Data' pixels can be filtered out from the conversion, if necessary (see @sec-distance-to-nearest-geometry).
### Raster to contours {#sec-raster-to-contours}
Another common type of spatial vectorization is the creation of contour lines, representing lines of continuous height or temperatures (*isotherms*), for example.
We will use a real-world digital elevation model (DEM) because the artificial raster `elev.tif` produces parallel lines (task for the reader: verify this and explain why this happens).
*Plotting* contour lines is straightforward, using the `contour=True` option of `rasterio.plot.show` (@fig-raster-contours1).
```{python}
#| label: fig-raster-contours1
#| fig-cap: Displaying raster contours
fig, ax = plt.subplots()
rasterio.plot.show(src_dem, ax=ax)
rasterio.plot.show(
src_dem,
ax=ax,
contour=True,
levels=np.arange(0,1200,50),
colors='black'
);
```
Unfortunately, **rasterio** does not provide any way of extracting the contour lines in the form of a vector layer, for uses other than plotting.
There are two possible workarounds:
1. Using `gdal_contour` on the command line (see below), or through its Python interface **osgeo**
2. Writing a custom function to export contour coordinates generated by, e.g., **matplotlib** or **skimage**
We demonstrate the first approach, using `gdal_contour`.
Although we deviate from the Python-focused approach towards more direct interaction with GDAL, the benefit of `gdal_contour` is the proven algorithm, customized to spatial data, and with many relevant options.
Both the `gdal_contour` program (along with other GDAL programs) and its **osgeo** Python wrapper, should already be installed on your system since GDAL is a dependency of **rasterio**.
Using the command line pathway, generating 50 $m$ contours of the `dem.tif` file can be done as follows.
```{python}
#| eval: false
os.system('gdal_contour -a elev data/dem.tif output/dem_contour.gpkg -i 50.0')
```
Like all GDAL programs (also see `gdaldem` example in @sec-focal-operations), `gdal_contour` works with files.
Here, the input is the `data/dem.tif` file and the result is exported to the `output/dem_contour.gpkg` file.
To illustrate the result, let's read the resulting `dem_contour.gpkg` layer back into the Python environment.
Note that the layer contains an attribute named `'elev'` (as specified using `-a elev`) with the contour elevation values.
```{python}
contours1 = gpd.read_file('output/dem_contour.gpkg')
contours1
```
@fig-raster-contours2 shows the input raster and the resulting contour layer.
```{python}
#| label: fig-raster-contours2
#| fig-cap: Contours of the `dem.tif` raster, calculated using the `gdal_contour` program
fig, ax = plt.subplots()
rasterio.plot.show(src_dem, ax=ax)
contours1.plot(ax=ax, edgecolor='black');
```
## Distance to nearest geometry {#sec-distance-to-nearest-geometry}
Calculating a raster of distances to the nearest geometry is an example of a 'global' raster operation (@sec-global-operations-and-distances).
To demonstrate it, suppose that we need to calculate a raster representing the distance to the nearest coast in New Zealand.
This example also wraps many of the concepts introduced in this chapter and in previous chapters, such as raster aggregation (@sec-raster-agg-disagg), raster conversion to points (@sec-raster-to-points), and rasterizing points (@sec-rasterizing-points).
For the coastline, we will dissolve the New Zealand administrative division polygon layer and 'extract' the boundary as a `'MultiLineString'` geometry (@fig-nz-coastline). Note that `.dissolve(by=None)` (@sec-vector-attribute-aggregation) calls `.union_all` on all geometries (i.e., aggregates everything into one group), which is what we want to do here.
```{python}
#| label: fig-nz-coastline
#| fig-cap: New Zealand coastline geometry
coastline = nz.dissolve().to_crs(src_nz_elev.crs).boundary.iloc[0]
coastline
```
For a 'template' raster, we will aggregate the New Zealand DEM, in the `nz_elev.tif` file, to 5 times coarser resolution.
The code section below follows the aggregation example in @sec-raster-agg-disagg.
```{python}
factor = 0.2
# Reading aggregated array
r = src_nz_elev.read(1,
out_shape=(