Toolkit for focal site multi-scale studies in Python.
(C) OpenStreetMap contributors, tiles style by Humanitarian OpenStreetMap Team hosted by OpenStreetMap France
Compute multi-scale spatial predictors:
import contextily as cx
import geopandas as gpd
import matplotlib.pyplot as plt
from sklearn import ensemble
import focalpy
species_richness_filepath = "data/bird-richness.gpkg"
buildings_gdf_filepath = "data/buildings.gpkg"
tree_canopy_filepath = "data/tree-canopy.tif"
buffer_dists = [100, 250, 500]
species_gdf = gpd.read_file(species_richness_filepath)
y_col = "n.species" # species richness
fa = focalpy.FocalAnalysis(
[buildings_gdf_filepath, tree_canopy_filepath],
species_gdf,
buffer_dists,
[
"compute_vector_features",
"compute_raster_features",
],
feature_col_prefixes=["building", "tree"],
feature_methods_args={
"compute_vector_features": [{"area": "sum"}],
},
feature_methods_kwargs={
"compute_raster_features": {"stats": "sum"},
},
)
fa.features_df.head()| building_area_sum_100 | building_area_sum_250 | building_area_sum_500 | tree_sum_100 | tree_sum_250 | tree_sum_500 | |
|---|---|---|---|---|---|---|
| 0 | 13069.227511 | 60218.251616 | 207368.012055 | 2016.0 | 14875.0 | 61452.0 |
| 1 | 7439.635337 | 41645.546860 | 131432.855040 | 1331.0 | 15760.0 | 84520.0 |
| 2 | 8962.495280 | 54251.129360 | 146157.281494 | 2385.0 | 16725.0 | 79704.0 |
| 3 | 8001.653873 | 29735.393494 | 102803.559740 | 2512.0 | 22892.0 | 95945.0 |
| 4 | 10447.531020 | 39405.263870 | 110922.947475 | 3886.0 | 19860.0 | 99111.0 |
# target area (for region-wide prediction/extrapolation)
study_area_filepath = "data/study-area.gpkg"
grid_res = 500
# train a model and spatially extrapolate it
model = ensemble.GradientBoostingRegressor().fit(fa.features_df, species_gdf[y_col])
pred_da = fa.predict_raster(model, study_area_filepath, grid_res, pred_label=y_col)
# plot the field data and predicted raster
fig, ax = plt.subplots()
cmap = "BuGn"
vmin = min(pred_da.min().item(), species_gdf[y_col].min())
vmax = max(pred_da.max().item(), species_gdf[y_col].max())
pred_da.plot(ax=ax, alpha=0.7, vmin=vmin, vmax=vmax, cmap=cmap)
species_gdf.plot(y_col, ax=ax, edgecolor="k", vmin=vmin, vmax=vmax, cmap=cmap)
ax.set_axis_off()
cx.add_basemap(ax, crs=species_gdf.crs, attribution=False)
(C) OpenStreetMap contributors, tiles style by Humanitarian OpenStreetMap Team hosted by OpenStreetMap France
See the user guide and the API documentation for more details on the features of focalpy.
Like many other geospatial Python packages, focalpy requires many base C libraries that cannot be installed with pip. Accordingly, the best way to install focalpy is to use conda/mamba, i.e., in a given conda environment, run:
# or mamba install -c conda-forge geopandas
conda install -c conda-forge geopandasWithin the same conda environment, you can then install focalpy using pip:
pip install https://github.com/martibosch/focalpy/archive/main.zip- This package was created with the martibosch/cookiecutter-geopy-package project template.
- Huais, P. Y. (2024). Multilandr: An r package for multi-scale landscape analysis. Landscape Ecology, 39(8), 140.