At this moment in May 2024, severe flooding is taking place in Kenya’s lowlands, causing significant casualties and livelihood destruction. Due to climate change, the volatility of weather patterns in Kenya has increased and brought with it a regular pattern of severe flooding in wet season and catastrophic drought in the dry season. Flooding is becoming increasingly more devastating to communities,
while at the same time, during dry season when the river beds are dry, there is insufficient groundwater for the local farmers to use.
There are a number of side weirs scattered along the river networks meant to divert water flow. However, they have been placed and organized with little consideration and forethought. Hydropower dams can also be leveraged to participate in improved water management. Finally, the design and use of makeshift reservoirs for water storage can be done cheaply and with local materials.
With active cooperation from Kenya’s Ministry of Water, and data gathering, processing, and community engagement from Lukenya University, Innovation Kenya, and IBM Research - Nairobi, a team of computational mathematicians and water engineers in the EU wish to simulate, model and ultimately optimize the water management practices in Kenya state of the art techniques
in scientifically informed machine learning.
Staff at Lukenya University will perform regular camera, depth, and ground soil measurements. IBM Research will provide processed satellite imagery data.
Research Staff at Czech Technical University in Prague and the Italian National Research Lab will develop CFD solvers for both river flow (Shallow Water), groundwater (Richards' equations), and peridynamic variants thereof (to faithfully model flooding - grade turbulence). These will be calibrated with System Identification and Uncertainty Quantification using the state of the art algorithms in Physics Informed Machine Learning.
The intention is to obtain an accurate model of 1) the yearly cycle of flooding to drought, in order to understand the circumstances that maximize groundwater presence during dry season, and 2) the Athi and Kambu network of rivers in the lowlands of Kenya, as governed by their sources in Ngong Hill, Kenya.
With this, we intend to perform engineering design optimization / operations research, to 1) optimize the placement of side weirs in the river, 2) incorporate hydropower dam decision and control into the water management practice, 3) work with farmers to design local material sourced simple reservoir storage devices and strategically store water and simulate optimal water storage and release. The target objective would be remaining groundwater in dry season and a risk measure on the volume of overflow in wet season.
Lukenya University will perform community outreach to understand the most critical considerations with respect to groundwater presence and other ethnographic variations as far as resource acquisition and use. The specific local needs of farmers, and preferences as far as agricultural and agroforestry products will be central to the groundwater optimization.
The ultimate project deliverable is a complete in silico description of the modeling and OR. If the deliverable is satisfactory, the team will then pursue additional funding to secure technician staff time and equipment for the physical engineering in Kenya, as well as additional research in order to develop a data driven real-time control scheme for the water network using the reservoir management. We hope that Manifund's impact market incentive structure will assist in in providing the proper information to investors in regards to this project's performance and thus its potential in that second phase.
1) Research Staff Time (Senior and Postdoctoral) at Czech Technical University
2) Research Staff Time (Senior and Postdoctoral) at Italian National Research Council, Institute of Applied Mathematics
3) Staff time at Lukenya University for Performing Field Measurements and Community Outreach
4) Research Staff time at IBM Research - Nairobi for processing satellite imagery data
5) Staff Time for Innovation Kenya in Assuring Streamlined Logistics through Dialogue with the Ministry of Water
Hiring and staff allocation will depend on the raised funds, that is whether the minimal $100 or full $500k is raised, i.e., we will try to make do with what we get, but appreciate funding for the most possible aggregate brainpower devoted to the project.
Vyacheslav Kungurtsev - Department of Computer Science, Faculty of Electrical Engineering, Czech Technical University in Prague. Researcher in optimization and computational mathematics with 30+ publications in the field.
Fabio DiFonzo and others - the Institute of Water Research (IRSA-CNR), and Institute of Applied Mathematics (IAC-CNR), from the Italian National Research Council, which has actively simulated and performed control for groundwater and river networks in Italy, California, and other locations.
Tana Athi River Development Authority (TARDA) - Runs the river networks of interest in Kenya, as affiliated with the Ministry of Water
Kennedy Mugo - Innovation Kenya - performs ethnographic research to assist FDI in Kenya, with a number of previous industrial collaborations.
Judith Wafula and others - Lukenya University - has strong links with the smallholder farmer community, and is currently undertaking an initiative to plant 10 million trees in the country.
Julien Kuehnert - IBM Research - Nairobi, has developed a comprehensive tool for using satellite imagery to predict biomass and carbon balance in regions of Kenya.
Given the inherent complexity of the system involved, it is clear that there must be a minimal amount of modeling computational precision, and a minimal amount of high quality information from data, that would be necessary to accurately fit the model to data (for instance, an infinitely powerful computer simulating Navier Stokes fit through a row of real time sensors would probably be effective, but impossible). A priori it is difficult to adequately estimate what this minimum is. Poor prediction accuracy will be the indication that we have not reached this minimum. If experimentation along an assortment of tools in CFD and machine learning does not yield success then the extensive experience of the Italian Water engineers will help develop a simpler empirical if not as accurate model. Such approximations could yield a compromised but still quality solution, and then research will focus on academic publishing for the sake of seeking to understand the first principles problems so as to facilitate broader academic engagement.
Vyacheslav Kungurtsev in currently PI for CTU for National Contract Research for Hydryopower Cascade Modeling and Optimization. The same software base as in development for his work here can be used to model the Kenyan river networks. He is also Senior Research Staff Member for Czech National Project on Scientific Acceleration with Machine Learning.
EU and National Proposals for Agroforestry Innovation, involving the same partners, pending