Smart Water Magazine
Connecting Waterpeople

You are here

AI pilot to address transboundary water challenges in Southern Africa

53
  • AI pilot to address transboundary water challenges in Southern Africa
  • Dr. Clara Bocchino of SWP and Tapiwa Chiwewe of IBM Research-Africa detail how tools based on artificial intelligence techniques such as machine learning are being put to use to improve the management of transboundary water.

About the entity

USAID
USAID transforms. It transforms families, communities, and countries – so they can thrive and prosper. Whether by preventing the next global epidemic, responding to a devastating earthquake, or helping a farmer access tools to grow her business.
Global Omnium
· 53

Themes

Worldwide, more than 263 watersheds and 300 aquifers are known as transboundary, meaning they cross the political boundaries of two or more countries. The people who rely on these water sources are intrinsically linked locally across borders and regionally with intensive urban and industrial water uses, thus creating a system of hydrological, economic and social interdependence.

For example, over-abstraction in a country upstream can lead to water shortages for both downstream people and their ecosystems. Additionally, pollutants can be carried from one country to another, which is particularly troubling in areas where urban and rural systems coexist with variable distances. Climate change further complicates such challenges because of increasingly erratic rainfall volumes, patterns and variability, which leads to droughts and floods. Countries may also differ in socio-economic development, water-use regulations and management capacity, further complicating the effective management of transboundary waters.

Despite these challenges, a range of useful tools based on artificial intelligence techniques (AI), such as machine learning and deep learning, are increasingly available to improve the management of transboundary water. Up-to-date information systems on water availability and demand can help increase the accuracy of models for determining the impacts of potential changes in the management and use of shared watersheds and aquifers. Starting this month, experts at IBM Research, Wits University, University of the Western Cape, Umvoto Africa and Delta-H, who know how to deploy these technologies in South Africa, are starting a new pilot to develop new techniques that are more user-friendly in the regional context.


Graphic By IBM

Pilot Launching

The Ramotswa Aquifer between South Africa and Botswana has been identified for a new pilot project which kicked off in January thanks to grants from USAID, the Department of Science and Technology of South Africa, and the Southern African Development Community (SADC) Groundwater Management Institute.  The location is an ideal transboundary water resource where AI techniques can be applied to improved databases to reveal the patterns for water sustainability. Applying these techniques can lead to improved management of the resource and contribute towards harmonized policy. The region is in a semi-arid and water-stressed area and supports industries, as well as a significant population in both countries.

AI techniques are useful for the analysis of big data that can be obtained from transboundary groundwater resources such as the Ramotswa Aquifer. Underlying patterns and trends within the data can be revealed to assist with decision-making and improved management.

Changes in water chemistry and quality in such water resources can also reveal spatial and temporal activities that span large areas and time scales. Samples collected at various boreholes across the area would usually reflect local chemistry and/or transitions that have occurred in the water over space and time.

More specifically, the team of researchers from IBM and the four research awardees will focus on:

  • Creation of an AI-ready databased for integrated decision-making of the Ramotswa Aquifer;
  • Time series forecasting from sampling points with deep learning;
  • Classification and clustering of large water datasets using semi-supervised and unsupervised machine learning techniques that may provide interesting patterns that are important in shedding more light into the behavior and relationships of the water through space and time;
  • An evidence base for decision-making based on transboundary water resource models at local and regional scale;
  • Creating an applicable transboundary water sustainability strategy applicable to other regional shared water basins.

At the end of 2020, experts will use the findings to provide a platform for the science-policy-practice nexus. This platform will inform policy formulation, decision making, risk assessment and potential design of response strategies in the event of any contamination of the transboundary water resource. It will be presented at the 2021 Symposium organized by the Water Research Council of South Africa for the Big Data Analytics and Transboundary Water Collaboration for Southern Africa.