Applying climate reparative finance toward water security.

Water insecurity is increasing globally as one of the first perceivable effects of climate change. While water management is typically a local challenge, climate finance mechanisms offer the potential to switch climate-damaging capital toward climate reparative water infrastructure and create a sustainable, performance-based funding stream to incentivize safe water services globally.

Estimating groundwater use and demand in arid Kenya through assimilation of satellite data and in-situ sensors with machine learning toward drought early action.

Groundwater is an important source of water for people, livestock, and agriculture during drought in the Horn of Africa. In this work, areas of high groundwater use and demand in drought-prone Kenya were identified and forecasted prior to the dry season. Estimates of groundwater use were extended from a sentinel network of 69 in-situ sensored mechanical boreholes to the region with satellite data and a machine learning model. The sensors contributed 756 site-month observations from June 2017 to September 2021 for model building and validation at a density of approximately one sensor per 3700 km2. An ensemble of 19 parameterized algorithms was informed by features including satellite-derived precipitation, surface water availability, vegetation indices, hydrologic land surface modeling, and site characteristics to dichotomize high groundwater pump utilization. Three operational definitions of high demand on groundwater infrastructure were considered: 1) mechanical runtime of pumps greater than a quarter of a day (6+ hr) and daily per capita volume extractions indicative of 2) domestic water needs (35+ L), and 3) intermediate needs including livestock (75+ L). Gridded interpolation of localized groundwater use and demand was provided from 2017 to 2020 and forecasted for the 2021 dry season, June-September 2021. Cross-validated skill for contemporary estimates of daily pump runtime and daily volume extraction to meet domestic and intermediate water needs was 68%, 69%, and 75%, respectively. Forecasts were externally validated with an accuracy of at least 56%, 70%, or 72% for each groundwater use definition. The groundwater maps are accessible to stakeholders including the Kenya National Drought Management Authority (NDMA) and the Famine Early Warning Systems Network (FEWS NET). These maps represent the first operational spatially-explicit sub-seasonal to seasonal (S2S) estimates of groundwater use and demand in the literature. Knowledge of historical and forecasted groundwater use is anticipated to improve decision-making and resource allocation for a range of early warning early action applications.

A contribution to drought resilience in East Africa through groundwater pump monitoring informed by in-situ instrumentation, remote sensing and ensemble machine learning.

The prevalence of drought in the Horn of Africa has continued to threaten access to safe and affordable water for millions of people. In order to improve monitoring of water pump functionality, telemetry-connected sensors have been installed on 480 electrical groundwater pumps in arid regions of Kenya and Ethiopia, designed to improve monitoring and support operation and maintenance of these water supplies. In this paper, we describe the development and validation of two classification systems designed to identify the functionality and non-functionality of these electrical pumps, one an expert-informed conditional classifier and the other leveraging machine learning. Given a known relationship between surface water availability and groundwater pump use, the classifiers combine in-situ sensor data with remote sensing indicators for rainfall and surface water. Our validation indicates a overall pump status sensitivity (true positive rate) of 82% for the expert classifier and 84% for the machine learner. When the pump is being used, both classifiers have a 100% true positive rate performance. When a pump is not being used, the specificity (true negative rate) is about 50% for the expert classifier and over 65% for the machine learner. If these detection capabilities were integrated into a repair service, the typical uptime of pumps during drought periods in this region could potentially, if budget resources and institutional incentives for pump repairs were provided, result in a drought-period uptime improvement from 60% to nearly of 85% - a 40% reduction in the relative risk of pump downtime.