Assessment of Rainfall and Climate Change Patterns in the City of Kigali via Machine Learning and IPCC Codex Models and their Impact on Heavy Storms
In one sentence. Genetic-fuzzy GENOFIS model beats ANFIS and SVM for Kigali rainfall prediction using 1991–2023 gauge data, combined with downscaled climate-change scenarios.
Abstract
Abstract Rwanda has been negatively affected by a changing climate, exacerbated by human impact on land and water resources. In most parts of the country the rainfall pattern has changed over the last decades resulting in both enhanced flooding and water shortage / scarcity in much of the country, especially in the Capital City of Kigali and peripheries which is the main economic hub of the country with strong links to the East African region. Changes in precipitation have affected agricultural production, hydropower production, and water supplies, and has been a result of increased heavy storms, and flash floods in the city. This study developed a new predictive model rainfall patterns in the City of Kigali (CoK) in the Republic of Rwanda using evolutionary methodologies that apply machine learning techniques of Fuzzy Inference Systems (FIS) trained via Genetic Algorithms, Neuro Network Systems and a comparative Support Vector Machine tool, and assessment downscaled climate change combinations with predicted rainfall patterns. The models were calibrated and validated using measured rainfall data in the City of Kigali from 1991 through 2023. The model results show the developed Geno Fuzzy Inference System (GENOFIS) model performed better than the Adaptive Neuro Fuzzy Inference System (ANFIS) and Support Vector Machine (SVM) models. The Coefficient of Efficiency (CE), and Root Mean Square Error (RMSE) were used as diagnostic measures for model performance evaluation. Models generated with GENOFIS are therefore recommended for rainfall and related prediction pattern in the City of Kigali for climate change adaptation and resilience policy and planning.
Labels
- Topic
- Local & station forecasting, Climate analysis & attribution
- Type
- New model or method
- Method
- Classical ML (trees, GP, linear, kernels)
- Time scale
- Climate projections / centennial
- Domain
- Local / station
- Data
- Station / in-situ, CMIP / climate model output
- Status
- Research
- Identifiers
- DOI:10.21203/rs.3.rs-3491099/v1 OpenAlex:W4388200869
- Added
- 2026-10-22
Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.