How Tech is saving lives by predicting floods and drought in Kenya

Posted on September 13, 2021

Climate change has had significant effects on the natural and social components of global systems. Changes in extreme weather and climate events have had significant impacts and are among the most serious challenges for all human societies today.

 In recent decades, an unprecedented number of extreme climate events have occurred in different parts of the world, such as unusually hot days and nights, fewer unusually cold days and nights, while heavy rainfall has also become more frequent and intense. In addition, more extreme and frequent floods and droughts have been occurring and remain anticipated in the future. All of these factors exert a major influence at regional scales, potentially increasing the risk of disaster for both human and environmental systems.

Feasible regional climate scenarios are necessary to efficiently adapt to climate change and to mitigate the risk of disaster, especially when extreme weather and high impact climatic events are concerned. Confidence in projecting changes in the direction and magnitude of climate extremes depends on many factors, for example, the type of extreme, the region, and season, the amount and quality of observational data, the understanding of underlying processes, and the reliability of model simulations, especially at the regional scale.

East Africa continues to experience unprecedented records of floods and drought, associated with extreme climatic conditions, which have had devastating effects on socioeconomic activities. Given the vulnerabilities to extreme weather events in the region, mitigation and adaptation strategies are necessary as part of long-term plans to help address such challenges. An important aspect of mitigating the impacts of drought is effective methods for predicting future flood and drought events.

Kenya lies within the tropical climatic zone that is faced with different extreme weather conditions in different areas. The Northern region exhibits dry conditions with central and western Kenya being wet areas receiving high rainfall.

Strathmore University, through its Research and Innovation Centre, @ILabAfrica in collaboration with Centre for Intellectual Property and Information Technology Law (CIPIT) have been working on a project aimed at enabling prediction of floods and drought in the country to enable the implementation of mitigation measures.

The project team evaluated existing physical and data-driven regional models focusing on spatial and temporal scales, assumptions, outputs, and efficacy in their prediction of climate extremes, proposing a tool to map regions and collect all the climate details on selected regions to enable farmers and other stakeholders make data-driven climate decisions.

The project is aimed at modeling the use modern technologies of machine learning algorithms, Artificial intelligence (AI) and data analytics in the correct prediction of drought and floods before they happen.

In the project deployment, Narok and Ahero were chosen for their prevalence to floods while Wajir and Turkana were selected because of the continued drought occurrences in the areas.

Jeremy Gachanja, Data analyst at @iLabAfrica faults foreign models for wrongly getting weather predictions in Kenya and Africa and large, citing that the new project provides more accurate models of forecasting African extreme climates.

“Global Climatic Models (GCMs) which are developed abroad do not do an adequate job when it comes to forecasting weather conditions in Kenya, specifically in Narok, Wajir, Turkana and Ahero. However, the application of machine learning algorithms in weather forecasting yields better results than those from GCMs. Therefore, in order to adequately forecast weather in Kenya and in Africa, machine learning algorithms would be the best place to start” says Mr. Gachanja.

“In the modelling for extreme weather conditions in Kenya, we used two approaches, Outlier detection modeling and use of Artificial Neural Networks (ANN) in the modeling for floods and droughts respectively. For floods the outlier based detection models was preferred as our choice of regions (Narok and Ahero) are mostly affected by flash floods and as such we wanted to see whether extreme positive outliers may be the cause of the floods in Narok and Ahero.

“For drought however, it cannot be such a drastic occurrence, as such we sought to use an index to determine whether or not a drought would occur and model future occurrences using that index. We used the Standard Precipitation and Evapotranspiration Index (SPEI) that is commonly used as the drought index for predicting floods. The ANN came in to predict the prospective droughts in our regions (Wajir and Turkana).” Said Mr. Alvin Igobwa, Business Intelligence Analyst at @iLabAfrica and lead researcher in the project.

According to Mr. Igobwa, policy leaders, government and local authorities will now be in a better position to make evidence driven decisions derived from data obtained through the project, this will enable for better preparedness in the wake of floods and drought situations.

Advances in technology have demonstrated data-driven models as promising tools for forecasting atmospheric, climatic, and hydrological variables. Data-driven models such as machine learning/artificial intelligence algorithms use historical climate patterns and data attributes such as recognized climate indices to make climate projections by learning the linear or non-linear relationship between predictors and the outcome of interests. The advantages of data-driven models over physical climate models are the explanations and more relevant information.  

Drought and Flood occurrences have huge social consequences to individuals and communities, the immediate effects of flooding include loss of human life, property damage, crop destruction, livestock loss and deterioration of health conditions due to waterborne diseases. A technology driven approach in the management of these two disasters will go a long way towards enabling a more stable human life with minimal interruptions in the wake of disaster occurrences as they will be well prepared to handle them.