The Burundi climate and fragility ASA developed a national level multi-risk modelling, including modelling of floods, landslides, droughts, and soil erosions hazards induced by intense precipitation, temperature effects, and natural climate variability. To do so, three input layers were developed by the study:

a) Modelling of climate and fragility hotspots on Burundi’s collines

b) Modelling of land degradation risks on Burundi’s collines

c) Lastly, modelling of fluvial and pluvial flood risk on Burundi’s collines

Put together; the three risk layers helped us to constitute a national-level map of Burundi’s multi-risk hotspots where urgent investments into resilience are needed.

a) Modelling climate and fragility risks in Burundi

Climate Centre et al. (2021) applied a compound risk analysis to create national maps of climate-fragility hotspots in Burundi, using an enhanced INFORM index method (Marin-Ferrer et al., 2017) in which the risk data is analysed based on four standard risk dimensions: hazard, vulnerability, exposure, and coping capacity. INFORM has been developed by the Joint Research Centre of the European Commission (see Climate Centre technical report for detailed methodology).

Introduction to INFORM - the Index for Risk Management.

INFORM is a global, open-source risk assessment for humanitarian crises and disasters.

b) Modelling conflict and climatic risk in Burundi

The analysis of conflict plays an essential role in the overall understanding of vulnerability across Burundi. There is a complex dynamic between vulnerability, conflict, and climatic hazards. All these components need to be explored to draw a holistic picture of the vulnerability of people living in the country. To identify locations and hot spots of high numbers of people with vulnerabilities exacerbated through conflict, a conflict risk analysis has been conducted using georeferenced historical conflict event data (UCDP 1989–2020; ACLED 2020–2021) as well as conflict hotspot mapping. Together with the results of a brief grey literature study and six (6) semi-structured interviews, this information has been used to identify the historical and current conflict situation and locations as well as potential future conflict trends and locations (Climate Centre et al. (2021).

c) Modelling land degradation risks in Burundi

Stanford Natural Capital Project et al. (2021) applied spatially explicit models to estimate erosion and landslide hazards nationally. Two Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) models were utilized, namely Sediment Delivery Ratio (SDR) model and Seasonal Water Yield (SWY) model as well as the factor of safety (Selby 1993) and gravitational process path (Wichmann 2017) approaches to estimate soil erosion, sedimentation rates, and identify areas of high-medium-low landslide risk, under 2020 conditions and with projections to 2050 (see Stanford Natural Capital Project technical report for methodology details).

The InVEST Sediment Retention Model explained

d) Assessing flood risks in Burundi

Lastly, the study developed a national-level rapid flood risk assessment based on a spatial multi-criteria evaluation (SMCE) approa ch based on the analytical hierarchy process (AHP) method and geographic information system (GIS) techniques, using globally available earth observation datasets. The SMCE method evaluates and integrates multiple layers to inform a flood risk modeling process. The input map layers included in this geospatial modeling are digital elevation model (DEM), precipitation (p), potential evapotranspiration (pet), normalized difference vegetation index (NDVI), compound topographic index (CTI), grided population layer, and accessibility map (i.e., representing the distance to major cities). These methods have been widely applied to inform decision-making processes in many developing countries, where there is limited or no data on flood extent. This approach aims not to deliver a “final and optimal solution” to flood risk mitigation but rather to provide high-level details on population, infrastructure, and landscape features’ relative exposure and vulnerability to fluvial and pluvial flood risks in order to inform decision-makers with information on where to prioritize interventions. Table 1 highlights flood risk severity and vulnerability levels of major factors conditioning flood risk.

Table 1. Flood risk severity & vulnerability levels for major factors conditioning to flood risk

	Risk Severity
Exposure & Vulnerability Level	Lowest	Low	Moderate	High	Highest
Risk Index	1	2	3	4	5
NDVI (index)	>0.637	0.637-0.527	0.527-0.416	0.416-0.305	<0.305
Elevation (m)	<1155	1155-1543	1543-1931	1931-2319	>2319
Precipitation (mm)	<115.7	115.7-140.8	140.8-165.9	165.9-191.0	>191.0
Potential Evapotranspiration (mm)	>119.5	119.5-110.5	110.5-101.6	101.6-92.7	<92.7
Slope (%)	<15	15.2-30.4	30.4-45-6	45.6-60.8	>60.8
Compound Topographic Index	<902.4	902.4-1391.8	1391.8-1881.2	1881.2-2370.6	>2370.6
Travel time (minutes)/ Mobility access	<211.4	211.4-422.8	422.8-634.2	634.2-845.6	>845.6
Population (people/100m)	<168	168-335	335-502	502-669	>836

e) Putting pieces together: mapping multi-hazard hotspots on Burundi’s colline landscapes

Finally, using a multi-hazard risk mapping approach, we overlaid individual climate-induced hazard layers threatening any given colline based on the three complementary scientific diagnostics of climate/fragility, land degradation, and flood modelling risks. Multi-hazard risk analysis helps policymakers and investors to decide where to channel funding and attention to making development efforts risk-informed.

An important methodological caveat: Locally available data were unavailable during phase 1 of the Burundi climate and fragility ASA for validating or calibrating the models used for the global diagnostic. The relative index scoring used in the two studies are relatively robust for making decisions at a country level. However, when prioritizing and designing investment programs, including landscape management interventions in specific collines , and recommending specific NBS interventions will require, at minimum, model validation, including calibration of biophysical models or, at best, ground-truthing in the prioritized areas. As indicated in the ASA’s timeline, the modeling results will be validated through ground-truthing in phase 2 of the project, including through a national stakeholder high-level validation of modeling results in mid-October 2021 and subsequent community-level validation workshops at the colline-level.