The role of citizen science in hydrology: combating lack of data and joining science and local policy

As I mentioned last week, one of the most important aspects of citizen science is its validity and whether the data can pass the stringent scientific quality control. In this study, the researchers explored this particular aspect and gave some valuable advice on the wider application of community-based hydrological and meteorological monitoring programme.

This study is initiated by the AMGRAF (Adaptive Management of shallow Groundwater for small-scale irrigation and poverty alleviation in sub-Saharan AFrica) project where its main goal is to investigate the potential of utilising shallow groundwater resources for small-scale irrigation and poverty alleviation. This therefore links nicely with one of my goals of the blog, which is to explore how the assessment of water resources come to inform policy actions.

The location of the study is at Danila woreda, Ethiopia (figure 1) and the installed monitoring points are shown in figure 2.

Figure 1. Study area.

Figure 2. Locations of monitoring points (close to arrowhead in Figure 1). MW= monitoring well, DSC = Dangesheta Service Cooperative, DAO = Dangesheta Agricultural Office.

Why was the community involved?

There is very limited data available for rainfall, river discharge and groundwater level. The researchers therefore seek help from the local communities for data collection. Specific issues with formally recorded data are outline below:

There is only one source precipitation data within the study area, Dangila woreda. Although there are 8 other rain gauges outside the study area, they all situate at different altitudes, limiting the ability to make useful comparison. Gridded datasets could be an alternative source of data for precipitation. However, as their resolution which ranges from 0.25◦ to 1.25◦ is too coarse and comprise of a highly variable topography, it is likely that the results may be compromised too.

The past river discharge data faced significant inconsistency with gaps up to years. Additionally, two monitored river gauges within Dangila woreda are no longer functioning and their existing records were only inconsistently digitised. Furthermore, due to the lack of good quality data, there is often no information on the level of peak flood and its timing.

As for groundwater level, there are only very limited data on boreholes and groundwater available from formal sources. With community monitored data, there is now information on water table depth and recession.

How the community was involved?

The research team consulted the local community to identify suitable place for installation of rain and river gauges and well monitors. Consequently, the areas chosen were convenient for regular measurement by the community and allowed good quality data to be collected (Figure 3). (e.g. avoiding areas where trees intercepting rainfall from rain gauges and steeper channels where obvious river stage fluctuations could be clearly captured by stage board)

There were also regular workshops throughout the project to ensure the locals understand the results and the wider implication of the research. The level of engagement is maintained through feedbacks and many even actively noted down extra information (e.g. conditions and timings of peak flood) which is extremely helpful for improving the hydrological understanding of this data poor environment.

Figure 3. Photographs of monitoring in action. Left: Kilti river guage, middle: rain guage, and right: monitoring well.

How are the data judged?

The “Guide to climatological practices” (WMO, 2011) through the following tests checks the representativeness and accuracy of measured data:

• Format tests: error in dates and numbers
• Completeness tests: whether the data (or lack thereof) is significant to the study
• Consistency tests
• internal consistency checks: e.g. max rainfall > min rainfall
• temporal consistency checks: e.g. regular time interval
• spatial consistency checks: e.g. similar rainfall in neighbouring areas
• summarisation consistency checks: e.g. monthly values adds up to correct annual values
• double mass check: e.g. relationship between stage and discharge do not change if the data is consistent
• comparison with formally recorded values available

How successful was it? 

Figure 4 below clearly shows the community data performs much better than the gridded datasets (TRMM, ERA-Interim, NASA-MERRA) across all time periods in terms of correlation with daily rainfall from Dangila rain gauge and has higher average of correlation, 0.73, than the expected value of 0.68. Also, the bias (best if closer to 1) is the lowest for the community data (shown in Figure 5). 

Figure 4. Pearson's Product Correlation Coefficient (PPC) between formally recorded daily rainfall and different sources.

Figure 5. Bias between  formally recorded daily rainfall and different sources.

Although both river discharge and groundwater data lack formally recorded data for validation, they are deemed ‘highly satisfactory’ after several statistical analysis. It is also realised that community data could benefit from more frequent measurement to better capture the nature of flash flood.

Summary and my reflection

Through this exercise, the shallow groundwater system of this data poor area is much better understood. For example, areas with slow drops of groundwater level after rainfall events signifies significant baseflow, therefore a suitable site for potential groundwater extraction. The collected data also proved to be much more applicable than the global gridded data set for hydrological modelling on a local scale. Furthermore, the high correlation, 0.73, between collected and formally measured data also means that the historical data could be used to extend the length of the model.

While the idea of stakeholder participation has gained significant grounds in environmental management, to date there has been very few researches on the role of citizen science in quantitative environmental monitoring. This example showed that through early engagement with the local communities from the inception of the project to data collection and analysis, valuable data could be obtained easily by the everyone, especially in areas where no alternative data exists! As it is usually areas like this that could benefit most significantly through an inclusive and participatory hydrological research, this has huge implications for the data poor area worldwide!  Indeed, through the words of the Danila woreda people, they feel longer the ‘subject of study’ but rather important partners of the project. They take pride in their work from which other communities that face similar issues can learn and benefit. This is an example of empowerment of the local community, but also one that highlights the role scientists could play in connecting the disparity between science and local policy (Ridder and Pahl-Wostl, 2005).