Methodology documentation
How the climate risk assessments — risk, impact and vulnerability — were compiled at national and district levels for floods, soil loss and deforestation.
This page presents climate risk assessments at the national and district levels, helping users understand the key components and the level of risk across the country.
The methodologies for compiling maps showing risk, impact and vulnerability for three hazards — floods, soil loss/degradation and deforestation — affecting the eight target states are described below.
Data collection & preparation
The first step involved acquiring the spatial datasets relevant to flood risk, soil loss and deforestation:
Topography
A Digital Elevation Model (DEM) of Malawi from SRTM.
Hydrology
Major rivers of Malawi from ESRI, plus gridded annual precipitation (1981–2010) from CHIRPS (Funk et al., 2015).
Terrain layers
Flow accumulation, slope, drainage density and Topographic Wetness Index (TWI), derived from the DEM with the PCRaster plugin in QGIS.
Soils
Vectorised UN FAO datasets delineating soil types, used to estimate erosion susceptibility and infiltration capacity.
Land use / land cover
ESA CCI 300 m land cover classification maps for 1992–2020 from Copernicus.
Population
A 2020 population density raster from the Humanitarian Data Exchange (HDX).
All datasets were reprojected to UTM 36S at a 300 m × 300 m horizontal resolution to improve rendering speed and ensure consistency in the spatial analysis.
Flood risk, impact & vulnerability mapping
The methodology adapts the approach of Negese et al. (2022): flood-prone areas are mapped by combining multi-criteria decision-making (MCDM) and the analytical hierarchy process (AHP) applied to the GIS raster datasets above. Raster layers for eight flood-controlling factors were compiled in QGIS 3.26 — 300 m CHIRPS rainfall (1981–2010), the SRTM DEM, and derived slope, flow accumulation, drainage density and TWI maps, plus a distance-to-rivers raster (SAGA proximity) and ESA CCI land cover.
Each factor was reclassified from 1 (very low susceptibility) to 5 (very high). AHP (Saaty, 1987) assigned relative weights via a pairwise comparison matrix (importance classified following Negese et al., 2022), normalised by column sum and averaged. To classify each Traditional Authority (TA), the pixels in each of the five risk categories were divided by the TA land area, then thresholds were applied nationwide:
| Criterion | TA-wide flood risk |
|---|---|
| ≥ 10% of pixels are “Very high” risk | Very high |
| < 10% “Very high” + ≥ 20% “High” + < 65% “Moderate” + < 50% “Low” | High |
| < 10% “Very high” + < 20% “High” + ≥ 65% “Moderate” + < 50% “Low” | Moderate |
| < 10% “Very high” + < 20% “High” + < 65% “Moderate” + ≥ 50% “Low” | Low |
| All others | Very low |
At this initial stage, and without detailed critical-infrastructure datasets (roads, schools, hospitals…), the flood impact map is based solely on population density (HDX 2020). Impact scores were assigned from the nationwide distribution of TA-level population densities:
| Population (per km²) | Impact |
|---|---|
| ≥ 1450 | Very high |
| 350 – 1449 | High |
| 150 – 349 | Moderate |
| 75 – 149 | Low |
| < 75 | Very low |
Vulnerability maps are a simple combination of the risk and impact scores for each district (Risk + Impact):
| Risk + Impact | Vulnerability |
|---|---|
| 9, 10 | Very high |
| 7, 8 | High |
| 5, 6 | Moderate |
| 3, 4 | Low |
| 1, 2 | Very low |
Soil loss risk, impact & vulnerability mapping
The Revised Universal Soil Loss Equation (RUSLE) was used to estimate countrywide erosion: E = R × K × C × SL × P, where E is soil erosion, R the climatic erosivity index, K the soil’s resistance to erosion, C the effect of vegetation cover, SL the slope-length/topography factor and P the erosion-control measures.
In QGIS, the K factor was assigned per soil type (e.g. Cambisol, Luvisol, Lixisol) from a UN FAO soil map and rasterised; C was derived from land cover; and SL was computed from slope length and gradient with the raster calculator. Applying RUSLE yields annual soil loss (t/ha), redistributed into five erosion-risk bins. As with floods, TA-level percentages were then classified nationwide:
| Criterion | TA-wide erosion risk |
|---|---|
| < 95% “Very low”/“Low” + most other pixels “Very high” | Very high |
| < 95% “Very low”/“Low” + most other pixels “High” | High |
| < 95% “Very low”/“Low” + most other pixels “Moderate” | Moderate |
| < 95% “Very low” but ≥ 95% “Very low” or “Low” | Low |
| ≥ 95% “Very low” | Very low |
The primary input was the 2020 CCI land cover map, reclassified into six broad categories (forest, crop, shrub, wetland, urban, other). Because soil loss most affects farming, the TA-wide percentage of “crop” pixels was computed (rainfed, rainfed herbaceous, rainfed tree/shrub, irrigated and mosaic cropland) and classified nationwide:
| Cropland pixels | Impact |
|---|---|
| ≥ 80% | Very high |
| 60% – 79% | High |
| 40% – 59% | Moderate |
| 20% – 39% | Low |
| < 20% | Very low |
Vulnerability scores were calculated using the same Risk + Impact method described above for floods.
Deforestation risk, impact & vulnerability mapping
A national deforestation layer was created by overlaying the 300 m CCI land cover maps for 1992 and 2020. “Forest” pixels in each year were extracted into binary rasters (1 = forest, 0 = other) and a difference raster computed for 1992–2020. Unlike floods and soil loss, values were not binned first; instead the distribution of deforested pixels per TA was used to set five risk classes:
| Deforested pixels | Risk |
|---|---|
| ≥ 2000 | Very high |
| 1500 – 1999 | High |
| 1000 – 1499 | Moderate |
| 500 – 999 | Low |
| < 500 | Very low |
This more simplistic approach does not account for TA size or deforestation as a percentage of TA area, so it may not give a fully holistic view. However, the smallest TAs tend to be mostly urban, and the 2000-pixel threshold for “Very high” adequately captures deforestation in many mid-size TAs such as Chapananga (Chikwawa), Kapelula (Kasungu) and Kalembo (Balaka).
At this initial stage, spatial data such as biodiversity hotspots, protected areas or socioeconomic indicators of forest dependence are lacking. For now, the 2020 population-density raster was used to assess deforestation impact — rendering the deforestation impact maps identical to those for floods. This step is to be improved with additional proxy data.
Socioeconomic factors — dependence on forest resources, legal protection status and governance frameworks — were analysed to determine vulnerability levels.