Sliderule data I/O#
This notebook will highlight analyzing various coincident elevation measurements. We will find regions with and use slideruleearth.io service to retrieve ICESat-2 and GEDI point elevation measurements.
Note
Keep in mind, these measurements are from different sensor types, close in time, but not at exactly the same time. All measurements also have uncertainties, so we do not expect perfect agreement.
import coincident
from mpl_toolkits.axes_grid1.inset_locator import inset_axes
import matplotlib.pyplot as plt
import numpy as np
# For testing OR if you have access to a private cluster set that here
# https://slideruleearth.io/web/rtd/developer_guide/articles/private_clusters.html#private-clusters
# sliderule.init(verbose=True, organization="uw", desired_nodes=5, bypass_dns=True, time_to_live=60)
/home/docs/checkouts/readthedocs.org/user_builds/coincident/checkouts/latest/src/coincident/io/download.py:25: TqdmExperimentalWarning: Using `tqdm.autonotebook.tqdm` in notebook mode. Use `tqdm.tqdm` instead to force console mode (e.g. in jupyter console)
from tqdm.autonotebook import tqdm
%matplotlib inline
# %config InlineBackend.figure_format = 'retina'
Identify a primary dataset#
Start by loading a full resolution polygon of a 3DEP LiDAR workunit which has a known start_datetime and end_datatime:
workunit = "CO_WestCentral_2019"
df_wesm = coincident.search.wesm.read_wesm_csv()
gf_lidar = coincident.search.wesm.load_by_fid(
df_wesm[df_wesm.workunit == workunit].index
)
gf_lidar
| workunit | workunit_id | project | project_id | start_datetime | end_datetime | ql | spec | p_method | dem_gsd_meters | ... | seamless_category | seamless_reason | lpc_link | sourcedem_link | metadata_link | geometry | collection | datetime | dayofyear | duration | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | CO_WestCentral_2019 | 175984 | CO_WestCentral_2019_A19 | 175987 | 2019-08-21 | 2019-09-19 | QL 2 | USGS Lidar Base Specification 1.3 | linear-mode lidar | 1.0 | ... | Meets | Meets 3DEP seamless DEM requirements | https://rockyweb.usgs.gov/vdelivery/Datasets/S... | https://prd-tnm.s3.amazonaws.com/index.html?pr... | https://prd-tnm.s3.amazonaws.com/index.html?pr... | MULTIPOLYGON (((-106.13571 38.4146, -106.1702 ... | 3DEP | 2019-09-04 12:00:00 | 247 | 29 |
1 rows × 33 columns
Search secondary datasets#
Provide a list that will be searched in order. The list contains tuples of dataset aliases and the temporal pad in days to search before the primary dataset start and end dates
secondary_datasets = [
("gedi", 40), # +/- 40 days from lidar
("icesat-2", 60), # +/- 60 days from lidar
]
gf_gedi, gf_is2 = coincident.search.cascading_search(
gf_lidar,
secondary_datasets,
min_overlap_area=30, # km^2
)
Get ICESat-2 ATL06 Data#
We’ve identified 7 granules of icesat-2 data to examine, but there is no need to work with the entire granule, which spans a huge geographic extent. Instead we’ll retrieve a subset of elevation values in the area of interest for each granule.
# Hone in on single granule as a pandas series
i = 0
granule_gdf = gf_is2.iloc[[i]]
granule = gf_is2.iloc[i]
granule_gdf.plot() # needs to be a geodataframe
plt.title(granule.id);
# The icesat-2 track trends N-S, and is crossed by multiple GEDI tracks, resulting in the crosshatched appearance
data_atl08 = coincident.io.sliderule.subset_atl08(
granule_gdf,
include_worldcover=True,
)
print(len(data_atl08))
data_atl08.iloc[0]
43
n_te_photons 103
segment_snowcover 1
region 6
gt 10
h_te_uncertainty 21.347189
solar_elevation 52.006649
segment_id_beg 787273
canopy_h_metrics [0.7434082, 1.0415039, 1.592041, 2.715088, 3.7...
segment_cover 25
worldcover.time_ns [2021-06-30T00:00:00.000000000]
terrain_slope 0.045475
rgt 211
h_te_median 2198.79834
h_canopy 14.040283
n_ca_photons 40
worldcover.fileid [0]
h_mean_canopy 7.448916
spot 1
worldcover.value Grassland
n_seg_ph 200
h_canopy_uncertainty 21.788818
h_max_canopy 14.15625
h_min_canopy 0.504883
srcid 3
segment_landcover 20
cycle 4
canopy_openness 4.628767
geometry POINT (-107.73212432861328 38.65806579589844)
Name: 2019-07-11 21:56:57.438899968, dtype: object
# Plot this data and overlay footprint
fig, ax = plt.subplots(figsize=(4, 5))
plt.scatter(
x=data_atl08.geometry.x,
y=data_atl08.geometry.y,
c=data_atl08.h_canopy,
s=10,
)
granule_gdf.dissolve().boundary.plot(ax=ax, color="magenta")
cb = plt.colorbar()
ax.set_title("ATL08 Canopy Height")
cb.set_label("h_canopy (m)")
# ATL06 corresponds to land surface elevation
data_is2 = coincident.io.sliderule.subset_atl06(
granule_gdf,
include_worldcover=True,
)
print(len(data_is2))
data_is2.iloc[0]
284
tide_ocean NaN
region 6
gt 50
n_fit_photons 17
w_surface_window_final 4.128994
atl06_quality_summary 0
worldcover.time_ns [2021-06-30T00:00:00.000000000]
rgt 211
r_eff 0.02509
seg_azimuth -174.518005
bsnow_h NaN
h_robust_sprd 0.688166
sigma_geo_h 5.002733
worldcover.fileid [0]
h_li_sigma 0.467223
dh_fit_dx 0.01674
x_atc 15767759.895834
spot 5
y_atc 4675.225098
worldcover.value Tree cover
srcid 2
bsnow_conf 127
cycle 4
h_li 2245.366943
segment_id 787226
geometry POINT (-107.80409390031208 38.67234499367456)
Name: 2019-07-11 21:56:57.269883392, dtype: object
# Plot this data and overlay footprint
fig, ax = plt.subplots(figsize=(4, 5))
plt.scatter(
x=data_is2.geometry.x,
y=data_is2.geometry.y,
c=data_is2.h_li,
s=10,
)
granule_gdf.dissolve().boundary.plot(ax=ax, color="magenta")
cb = plt.colorbar()
ax.set_title("ATL06 Land Surface Elevation")
cb.set_label("h_li (m)")
Note
sliderule gets data from the envelope of multipolygon, not only data within intersecting GEDI tracks. Along-track gaps are where there is missing data.
Spatial join: nearest points#
NOTE: we will not worry about the difference in time of acquisition between adjacent points for now
def get_nearest_points(gf_reference, gf_compare, max_distance=100):
"""Get nearest points across two geodataframes within a maximum distance in meters.
NOTE: reprojects dataframes to UTM and returns dataframe point in UTM
"""
gf_left_utm = gf_reference.to_crs(gf_reference.estimate_utm_crs())
gf_right_utm = gf_compare.to_crs(gf_compare.estimate_utm_crs())
nearest_points = gf_left_utm.sjoin_nearest(
gf_right_utm,
how="left",
max_distance=max_distance,
distance_col="distances",
)
return nearest_points
# ATL08 has h_te_median as an estimate of median terrain elevation.
# ATL06 has h_li as the land surface elevation which we expect to be close but not exactly the same
# based on algorithm differences and point spacing.
nearest_points = get_nearest_points(data_atl08, data_is2, max_distance=100)
nearest_points.head()
fig, ax = plt.subplots(figsize=(5, 4))
(nearest_points["h_te_median"] - nearest_points["h_li"]).plot.hist(bins=10, ax=ax)
plt.title("ATL08 h_te_median - ATL06 h_li")
plt.xlabel("Elevation Difference (m)");
Get GEDI Data#
# Here, we specify aoi=granule_gdf to only return GEDI data
# in our icesat-2 track of interest
data_gedi = coincident.io.sliderule.subset_gedi02a(
gf_gedi,
aoi=granule_gdf,
include_worldcover=True,
)
data_gedi.head(2)
| orbit | track | elevation_lm | solar_elevation | shot_number | flags | worldcover.time_ns | worldcover.fileid | elevation_hr | worldcover.value | beam | sensitivity | srcid | geometry | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| time_ns | ||||||||||||||
| 2019-08-13 00:26:33.160042496 | 3775 | 3020 | 1773.083374 | 18.861452 | 37750100300143430 | 130 | [2021-06-30T00:00:00.000000000] | [17179869184] | 1776.642456 | Grassland | 1 | 0.925909 | 121 | POINT (-107.69263 38.75743) |
| 2019-08-13 00:26:33.168306944 | 3775 | 3020 | 1772.384277 | 18.861008 | 37750100300143431 | 130 | [2021-06-30T00:00:00.000000000] | [17179869184] | 1777.179688 | Grassland | 1 | 0.935042 | 121 | POINT (-107.69212 38.7571) |
# Plot this data and overlay footprint
fig, ax = plt.subplots(figsize=(4, 5))
plt.scatter(
x=data_gedi.geometry.x,
y=data_gedi.geometry.y,
c=data_gedi.elevation_hr,
s=10,
)
granule_gdf.dissolve().boundary.plot(ax=ax, color="magenta")
cb = plt.colorbar()
cb.set_label("elevation_hr (m)")
NOTE Like ICESat-2, it’s possible to have points falling outside the estimated footprints
# find nearest IS2 point to each GEDI
close_gedi = get_nearest_points(data_gedi, data_is2, max_distance=100)
close_gedi = close_gedi[close_gedi["distances"].notna()]
fig, ax = plt.subplots()
plt.scatter(close_gedi.h_li, close_gedi.elevation_lm)
ax.axline((0, 0), slope=1, color="k", transform=ax.transAxes)
plt.xlabel("ATL06 elevation (m)")
plt.ylabel("GEDI L2A elevation_lm (m)")
plt.title(f"{len(close_gedi)} points within 100m")
# Add statistics and inset histogram to plot
inset_ax = inset_axes(ax, width="40%", height="30%", borderpad=2, loc="lower right")
residuals = close_gedi.h_li - close_gedi.elevation_lm
mean_residual = np.mean(residuals)
median_residual = np.median(residuals)
std_residual = np.std(residuals)
textstr = "\n".join(
(
f"Mean: {mean_residual:.2f}",
f"Median: {median_residual:.2f}",
f"Std: {std_residual:.2f}",
)
)
props = dict(boxstyle="round", facecolor="wheat", alpha=0.5)
ax.text(
0.05,
0.95,
textstr,
transform=ax.transAxes,
fontsize=10,
verticalalignment="top",
bbox=props,
)
_ = inset_ax.hist(residuals, bins=50, range=(-20, 20), color="gray")
inset_ax.axvline(0, color="k", linestyle="-", linewidth=1);
Sample 3DEP#
Sample 3DEP 1m DEM at the subset of GEDI Points
# Warning: this relies on TNM API, which is often down :(
close_gedi_3dep = coincident.io.sliderule.sample_raster(
close_gedi.to_crs(4326), # reproject to lat/lon for sampling
asset_key="usgs3dep-1meter-dem",
# Restrict to only DEMs derived from specific WESM LiDAR project
project_name=gf_lidar["project"].iloc[0],
)
/home/docs/checkouts/readthedocs.org/user_builds/coincident/checkouts/latest/.pixi/envs/docs/lib/python3.14/site-packages/geopandas/array.py:411: UserWarning: Geometry is in a geographic CRS. Results from 'sjoin_nearest' are likely incorrect. Use 'GeoSeries.to_crs()' to re-project geometries to a projected CRS before this operation.
warnings.warn(
# Add 3DEP elevation values to existing dataframe
close_gedi["elevation_3dep"] = close_gedi_3dep.value.values
# Geometries of nearest IS2 points
close_is2 = data_is2.loc[close_gedi.time_ns_right]
# Sample 3DEP at the subset of GEDI Points
close_is2_3dep = coincident.io.sliderule.sample_raster(
close_is2,
asset_key="usgs3dep-1meter-dem",
project_name=gf_lidar["project"].iloc[0],
)
/home/docs/checkouts/readthedocs.org/user_builds/coincident/checkouts/latest/.pixi/envs/docs/lib/python3.14/site-packages/geopandas/array.py:411: UserWarning: Geometry is in a geographic CRS. Results from 'sjoin_nearest' are likely incorrect. Use 'GeoSeries.to_crs()' to re-project geometries to a projected CRS before this operation.
warnings.warn(
close_is2["elevation_3dep"] = close_is2_3dep.value.values
plt.axvline(0, color="k", linestyle="dashed", linewidth=1)
diff_3dep = close_is2.h_li.values - close_is2.elevation_3dep
label = f"ICESat-2 (median={np.nanmedian(diff_3dep):.2f}, mean={np.nanmean(diff_3dep):.2f}, std={np.nanstd(diff_3dep):.2f}, n={len(diff_3dep)})"
plt.hist(diff_3dep, range=[-5, 5], bins=100, color="m", alpha=0.5, label=label)
diff_3dep = close_gedi.elevation_lm.values - close_gedi.elevation_3dep
label = f"GEDI2A (median={np.nanmedian(diff_3dep):.2f}, mean={np.nanmean(diff_3dep):.2f}, std={np.nanstd(diff_3dep):.2f}, n={len(diff_3dep)})"
plt.hist(diff_3dep, range=[-5, 5], bins=100, color="c", alpha=0.5, label=label)
plt.xlabel("elevation difference (m)")
plt.title("Altimeter Elevation - 3DEP 1m DEM")
plt.xlim(-5, 5)
plt.legend(loc="upper left");
