poligrain quick tour - OpenSense Software

This is a very brief overview of features available in the poligrain package. A more comprehensive overview of can be found in the poligrain documentation.

import matplotlib.pyplot as plt

import poligrain as plg

Explore example datasets¶

We provide the option to download subsets of existing open datasets with opportunistic sensor data. Currently, the following open datasets are supported:

OpenMRG: CML, rain gauge and radar data in Gothenburg, Sweden (Andersson et al. (2022))
OpenRainER: CML and rain gauge data in Emilia-Romagna, Italy (Covi et al. (2026))
Amsterdam PWS data: PWS data in Amsterdam, Netherlands (https://data.4tu.nl/articles/dataset/Rainfall_observations_datasets_from_Personal_Weather_Stations/12703250)
OpenMesh: CML, PWS and ASOS reference station data in New York City, USA (Jacoby et al. (2025))

Below we briefly show how to load a subset of the OpenMRG dataset.

(
    ds_rad,
    ds_cmls,
    ds_gauges_municp,
    ds_gauge_smhi,
) = plg.example_data.load_openmrg(data_dir="example_data", subset="8d")

The CML data is provided in a xarray.Dataset following the OpenSense naming conventions defined by (Fencl et al., 2023).

Note that this example dataset has already been processed and the rain rates R per CML are stored. Normally are raw CML dataset would only contain RSL and TSL data.

ds_cmls

Plot point, line and grid data on a map¶

One key feature of poligrain is the ability to plot point, line and grid data on a map. Plotting points and grids is straightforward with matplotlib. A fast and flexible way to plot line data was missing before.

Below we show an example of how to plot the rainfall accumulatoin of CMLs, rain gauges and radar on a map.

plg.plot_map.plot_plg(
    da_cmls=ds_cmls.isel(sublink_id=0).R.resample(time="1h").mean().sum(dim="time"),
    da_grid=ds_rad.R.resample(time="1h").mean().sum(dim="time"),
    da_gauges=ds_gauges_municp.rainfall_amount.sum(dim="time"),
    vmin=0,
    vmax=120,
);

Project coordinates¶

We project the coordinates to UTM zone 32N (EPSG:32832), suitable for this dataset from Gothenburg in Sweden, to have correct spatial measurements and distances.

Note that the source projection in project_point_coordinates defaults to WGS84 (EPSG:4326). Our lon-lat coordinates are WGS84, so we do not have to set the source projection explicitly.

epsg_str = "EPSG:25832"

# Project coordinates for rain gauges
ds_gauges_municp.coords["x"], ds_gauges_municp.coords["y"] = plg.spatial.project_point_coordinates(
    ds_gauges_municp.lon, ds_gauges_municp.lat, target_projection=epsg_str
)

# Project coordinates for CMLs
(
    ds_cmls.coords["site_0_x"],
    ds_cmls.coords["site_0_y"],
) = plg.spatial.project_point_coordinates(
    ds_cmls.site_0_lon, ds_cmls.site_0_lat, epsg_str
)
(
    ds_cmls.coords["site_1_x"],
    ds_cmls.coords["site_1_y"],
) = plg.spatial.project_point_coordinates(
    ds_cmls.site_1_lon, ds_cmls.site_1_lat, target_projection=epsg_str
)

# Project coordinates for radar grid
(
    ds_rad.coords["x_grid"],
    ds_rad.coords["y_grid"],
) = plg.spatial.project_point_coordinates(
    ds_rad.lon,
    ds_rad.lat,
    target_projection=epsg_str,
)

Calculate distances between points and lines¶

point-to-point distances¶

closest_neigbors = plg.spatial.get_closest_points_to_point(
    ds_points=ds_gauges_municp,
    ds_points_neighbors=ds_gauges_municp,
    max_distance=20e3,
    n_closest=12,
)

closest_neigbors.distance.plot();

point-to-line distances¶

closest_neigbors = plg.spatial.get_closest_points_to_line(
    ds_cmls, ds_gauges_municp, max_distance=20e3, n_closest=10,
)

closest_neigbors.distance.plot();

cml_id = 10089
cml_id = 10098
cml_id = 10073

plg.plot_map.plot_plg(
    da_cmls=ds_cmls.sel(cml_id=cml_id),
    use_lon_lat=False,
)

gauge_id = closest_neigbors.sel(cml_id=cml_id).dropna(dim='n_closest').neighbor_id.values
gauge_distance = closest_neigbors.sel(cml_id=cml_id).dropna(dim='n_closest').distance.values

sc = plt.scatter(
    ds_gauges_municp.sel(id=gauge_id).x, 
    ds_gauges_municp.sel(id=gauge_id).y,
    c=gauge_distance, 
    cmap="viridis",
    vmin=1.4e3,
    vmax=5.1e3,
)
plt.colorbar(sc, label="Distance to CML [m]");

Get radar path-averaged rainfall rates along CMLs¶

When comparing CML data with radar data we need to intersect the CML paths with the radar grid. poligrain has functions to perform this line-grid intersection efficiently using its GridAtLines class. An equivalent for point data exists with the GridAtPoints class so that point and line data can be used in the same way.

We use the radar rainfall rates and get the path-averages along the CMLs using the GridAtLines class which returns a xarray.DataArray with the original CML IDs and coordinates and the radar data time stamps.

get_grid_at_lines = plg.spatial.GridAtLines(
    da_gridded_data=ds_rad,
    ds_line_data=ds_cmls,
)

radar_along_cml = get_grid_at_lines(da_gridded_data=ds_rad.R)
radar_along_cml

Plot the 1h mean rainfall rates of the radar and the CML path-averages for selected time steps.

da_R_1h = ds_rad.R.resample(time='1h').mean()
da_radar_along_cml_1h = radar_along_cml.resample(time='1h').mean()

fig, axs = plt.subplots(1, 3, figsize=(12, 4), sharex=True, sharey=True)

for i, t_str in enumerate(['2015-07-29T06:00:00', '2015-07-29T07:00:00', '2015-07-29T08:00:00']):
    plt.sca(axs[i])
    plg.plot_map.plot_plg(
        da_grid=da_R_1h.sel(time=t_str),
        da_cmls=da_radar_along_cml_1h.sel(time=t_str),
        vmin=0,
        vmax=10,
        cmap='YlGnBu',
        ax=axs[i],
        kwargs_cmls_plot={"edge_color": "r"},
    )
plt.tight_layout()

Plot the path-averaged radar rain rate time series and the total path loss tl (transmitted minus received signal level) for selected CMLs.

for i_cml in [1, 130, 312]:
    fig, axs = plt.subplots(2, 1, figsize=(12, 4), sharex=True)
    da_radar_along_cml_1h.isel(cml_id=i_cml).plot(ax=axs[0], label='radar rain rate along CML', color='k')
    
    rsl = ds_cmls.isel(cml_id=i_cml, sublink_id=0).rsl
    tsl = ds_cmls.isel(cml_id=i_cml, sublink_id=0).tsl
    tl = tsl - rsl
    tl.plot(ax=axs[1], label='CML total path loss')

    axs[0].set_title(f"CML ID: {ds_cmls.cml_id.values[i_cml]}")
    axs[1].set_title("")
    axs[0].set_xlabel("")
    axs[0].set_ylabel("Path averaged\nradar rain rate in mm/h")
    axs[1].set_ylabel("CML Total Path Loss")
    plt.show()

References¶

Andersson, J. C. M., Olsson, J., van de Beek, R. (C. Z. ), & Hansryd, J. (2022). OpenMRG: Open data from Microwave links, Radar, and Gauges for rainfall quantification in Gothenburg, Sweden. Earth System Science Data, 14(12), 5411–5426. 10.5194/essd-14-5411-2022
Covi, E., Roversi, G., & Nebuloni, R. (2026). OpenRainER: an open-source dataset for studying the opportunistic sensing of rainfall in Emilia-Romagna, Italy. 10.5194/essd-2026-160
Jacoby, D., Yu, S., Hu, Q., Hine, Z., Johnson, R., Ostrometzky, J., Kadota, I., Zussman, G., & Messer, H. (2025). OpenMesh: Wireless Signal Dataset for Opportunistic Urban Weather Sensing in New York City. 10.5194/essd-2025-238
Fencl, M., Nebuloni, R., C. M. Andersson, J., Bares, V., Blettner, N., Cazzaniga, G., Chwala, C., Colli, M., de Vos, L., El Hachem, A., Galdies, C., Giannetti, F., Graf, M., Jacoby, D., Victor Habi, H., Musil, P., Ostrometzky, J., Roversi, G., Sapienza, F., … Zheng, X. (2024). Data formats and standards for opportunistic rainfall sensors. Open Research Europe, 3, 169. 10.12688/openreseurope.16068.2