WATER RESOURCE MONITORING 

Background

Often surrounded by high ground, a catchment is an area of land where water collects when it rains. Catchments can greatly vary in size from small sub-catchments (200 square kilometers or less) to large catchments (1000 square kilometers or more). The virgin land usually has a thick vegetation cover including forests with thick undergrowth. In most countries, the land is highly protected to prevent encroachment by humans. Many catchments have one or a couple of major rivers originating from them, these rivers are kept alive by the water available in the catchment hence by monitoring the river water parameter such as water level, inferences about the catchment can be made. In recent times, catchments have been heavily affected by climate change and catchment degradation.To detect, monitor and quantify the effects of these threats to the well-being of the catchment, monitoring river water parameters such as water level and water turbidity is necessary. The result of the monitoring is data that can be used in diagnosing the catchment and in improving conservation efforts. The data can also be combined with other relevant datasets such as rainfall data in the catchment and satellite maps to predict future events in the catchments and catastrophes such as floods. At DSAIL we have developed, tested and deployed water level monitoring systems along the Muringato River in the Muringato watershed in Nyeri county, Kenya and also along River Ewaso-Nyiro North in the Ewaso Nyiro basin in Laikipia county, Kenya. We have also developed and deployed a water turbidity, water temperature, water electro- conductivity and water pH monitoring system along the Muringato River.

Muringato Catchment

The map below shows the extent of the Muringato catchment from the Aberdare Ranges in Kenya (Upper catchment) to the Lower catchment areas near Dedan Kimathi University of Technology.

System Development

The water level monitoring systems developed are meant to replicate and improve on the effectiveness of the often bulky, complex and expensive to build and maintain infrastructures for stage monitoring that have been developed over the years. The data collection is done in real time and the data is visualized on a web app. The sensor network utilized in sending the data to the network server is the LoRa network. The figure below shows the data collection architecture.

Currently we have water level data from February 2021 to date. And we intend to continue monitoring for a longer time stretch. The devices run on batteries which are charged using solar panels. The battery voltage is monitored to make sure the solar panels are effective and also to track the amount of power available.

Accomplishments

Over the last four years the project has fully taken shape and now it is in the expansion stage where deployments and analysis can happen at a larger scale. Below are some notable achievements over the years:

• Data acquisition along River Muringato from 2021 to date
• Data acquisition along River Ewaso Nyiro North in the Upper Ewaso Nyiro basin.
• Development of energy efficient sensor nodes and reliable sensor networks
• Utilization of the datasets in interesting machine learning tasks such as Anomaly detection in time series data and Multistep time series forecasting.
• By incorporating other datasets such as TAHMO climatic data in the study, the project has opened other avenues of research.

In the last one year, we have managed to put out another publication on how anomaly detection was tackled in water level data cleaning. Soon we'll be publishing DSAIL-Maji. DSAIL-Maji is a dataset containing water level time series sets for the years the project has been active.
Another notable project area that we have been working on is power analysis. We have done a lot of sensor node tests to determine the power requirements for the sensor node versions we developed. The tests and analysis have assisted us in providing the adequate power requirements during deployment. Also included under power analysis, is instrumentation. Most sensor nodes deployed in the wild are deployed to monitor one or two parameters. To monitor the well being of the sensor, onboard sensors such as battery voltage, solar voltage, solar current can be added to the sensor node design to help in monitoring the sensor node “health”.

Next Steps

• Publication: DSAIL-Maji datasets
• Publication: Multi-step time series forecasting
• Project renovation
• More data!!!!

Links

• Jason Kabi website https://kabi23.github.io
• DSAIL Github repo https://github.com/DeKUT-DSAIL/ewaso
  
• Hardware files https://data.mendeley.com/datasets/mszg723r9b/2
• Data visualisation app https://water-monitoring-258811.wl.r.appspot.com
• Project Technical report  https://dekut-dsail.github.io/assets/documents/technical
  reports/Water%20Resource%20Monitoring/DSAIL-2021-001.pdf