Climate Maps

This pipeline maps annual climate indicators onto Rwanda’s sectors. It cleans ERA5-Land climate values keyed by health facility, aligns their names to the admin3 boundaries, joins the two, and renders a binned choropleth. The full source is analysis/spatial-analysis/Climate_maps.ipynb.

The end result is a two-panel choropleth of annual mean temperature and total precipitation:

The data

The climate workbook (Climate data.xls) holds ERA5-Land variables per health facility, with the facility’s location described by the DHIS2 org-unit hierarchy. Only five variables are carried forward:

The org-unit levels map onto the administrative hierarchy — orgunitlevel3 is the district and orgunitlevel5 is the sector, which is the join key for admin3.

Pipeline

import pandas as pd
 
cl_df = pd.read_excel("Climate data.xls", engine="xlrd", skiprows=1)

cl_df["District"] = cl_df["orgunitlevel3"].str.split(" ").str[0]
cl_df["Sector"] = cl_df["orgunitlevel5"].str.split(r"[ _-]").str[0].str.capitalize()

cl_df["humidity"] = cl_df["RF001 Relative humidity (ERA5-Land)"]
cl_df["Average_temperature"] = cl_df["TMP001 Air temperature (ERA5-Land)"]
cl_df["Max_temperature"] = cl_df["TMP002 Max air temperature (ERA5-Land)"]
cl_df["Min_temperature"] = cl_df["TMP005 Min temperature (ERA5-Land)"]
cl_df["Rainfall"] = cl_df["RF003 Precipitation (ERA5-Land)"]
 
climate_data = cl_df[["periodname", "Sector", "District", "Average_temperature",
                      "Max_temperature", "Min_temperature", "Rainfall", "humidity"]]
climate_data["periodname"] = pd.to_datetime(climate_data["periodname"], format="%B %Y")

merged = pd.merge(admin3, climate_data, on="Sector", how="inner")

import geopandas as gpd
 
agg = merged[merged["periodname"].dt.year == year_to_plot]
g = (agg.groupby("adm_name", as_index=False)
        .agg({"geometry": "first",
              "Average_temperature": "mean",
              "Rainfall": "sum"}))
gdf = gpd.GeoDataFrame(g, geometry="geometry", crs=merged.crs)

Name standardization

The single most error-prone step is matching climate sector names to the shapefile. Spelling drifts between systems, and an unmatched key silently drops a sector from the inner join. Known fixes are applied explicitly:

# Shapefile side
admin3["ADM3_EN"].replace("Mageregere", "Mageragere", inplace=True)
admin3["ADM3_EN"].replace("Shyrongi", "Shyorongi", inplace=True)
admin3["Sector"] = admin3["ADM2_EN"] + " " + admin3["ADM3_EN"]
 
# Climate side
climate_data["Sector"].replace("Ririma", "Rilima", inplace=True)

Binning and rendering

The choropleth uses fixed, meaningful class breaks rather than a continuous ramp so the map reads at a glance. Temperature and precipitation each get their own bins and colormap (YlOrRd for temperature, Blues for precipitation):

temp_bins = [0, 15, 17, 19, 22, gdf["Average_temperature"].max()]
temp_labels = ["0-15", "15-17", "17-19", "19-22", "> 22"]
 
precip_bins = [0, 200, 300, 500, 800, gdf["Rainfall"].max()]
precip_labels = ["0-200", "200-300", "300-500", "500-800", "> 800"]

Each panel plots the value column with its colormap, overlays the admin1–admin3 boundaries for context, and draws a patch legend from the bin colors. See the notebook’s final cell for the complete two-panel figure.