Tracking flood pulses and their impacts on water quality using a low-cost, open-source monitoring network in East Africa
River systems are comprised of complex drainage networks, and understanding patterns of discharge through these systems is critical to understanding many river ecosystem processes. Obtaining detailed discharge measurements throughout a watershed requires a large, widespread network of sensors, which are often prohibitively expensive and difficult to install and maintain, particularly in remote field locations. In the Mara River of Kenya/Tanzania, we have observed repeated hypoxic events that appear to be driven by high discharge events, with flood pulses from different regions of the watershed driving different levels of dissolved oxygen (DO) decline. Our primary research questions were 1) can we design a distributed network to measure rainfall and discharge variability throughout the basin, and 2) can we use that network to determine how flood pulses from different regions of the catchment vary in their ecosystem impacts on the river? We designed a low-cost, open-source monitoring network using the Arduino platform to measure rainfall and discharge at multiple sites throughout the basin. Using a variety of transmission protocols (GPRS, SMS, WIFI), data is transmitted to a free, online data repository (Thingspeak) in near real-time. By comparing the occurrence and timing of high flow events from different regions of the watershed with continuous water quality monitoring at a downstream site, we can then determine the influence of landscape and hydrology on water quality-related ecosystem processes in the river.
We installed approximately 30 stations (18 rainfall, 12 discharge) at remote sites over a 45 day period. The cost of one station is 5% of the cost of a commercial station, and these stations have proven to be approximately 95% reliable. Between Sept. 2014 and Feb. 2015, we documented 20 flood pulse events, ranging from a 29-272% increase in discharge. These events had a significant, positive relationship with DO declines (R2=0.58, p<0.001), which ranged in magnitude from 5-96%. Variability around the occurrence and size of these declines was partially determined by the origin of the flood pulse, with flood pulses from the dryland regions of the catchment most likely to cause a hypoxic event. Discharge is a master variable driving most river ecosystem processes, and variability in rainfall, runoff, and discharge throughout a watershed can have profound effects. Distributed networks of low-cost, open-source sensors in remote regions will advance our understanding of hydrology and ecosystem dynamics such as tracking discharge events and downstream ecosystem consequences in space and time.