Regionalization of GR4J model parameters and generation of streamflow data for periods and locations without observations

Regionalization of GR4J model parameters and generation of streamflow data for periods and locations without observations

1.Host Institution:
Jomo Kenyatta University of Agriculture and Technology (JKUAT), P.O. Box 62 000 – 00200 Nairobi, Kenya

2. Supervisors:

(1) John M. Gathenya, PhD (mgathenya@gmail.com)
Professor, Hydrology, Climate Services, Environmental Services, Jomo Kenyatta University of Agriculture & Technology, School of Biosystems and Environmental Engineering
(2) Roger Stern (r.d.stern@reading.ac.uk), Professor of Applied Statistics (Emeritus), University of Reading, UK

3. Project Description:
The airGR is a suite of conceptual rainfall-runoff models and consists of the daily four-parameter lumped model, GR4J and the monthly two-parameter lumped model GR2M (airGR package | R Documentation). These models have been widely applied in many parts of the world but their application in Africa is still very low.  They can be used to fill gaps or extend streamflow time series in small watersheds where reliable climate data is available. The input data for these models is precipitation and potential evapotranspiration expressed in units of depth (mm/day or mm/month). Streamflow data is needed for calibration and validation of model parameters. A small watershed in this case is one where a single precipitation time series (gauge station, areal average, gridded data) is representative of the whole watershed. Prediction of flows in ungauged watersheds is another application of such models. In the absence of observed streamflow data, one has to employ parameter regionalization approaches, where fitted parameters are transferred from watersheds that are hydrologically similar and used together with precipitation and potential evapotranspiration data to generate streamflow data. Hydrological similarity is influenced by factors such as distance between watersheds, size and geometry of watersheds, topography, geology and soils, land use and climate. For each of these factors, a set of relevant metrics are extracted from databases or from analyses of geospatial data. Some skills or technical support in GIS may be required. There are a number of regionalization procedures in literature and none is considered to be universally acceptable, so the best way is to select and evaluate promising ones. The objective of this project is to test selected parameter regionalization approaches in terms of their performance in generating streamflow data in small watersheds for periods or locations without observations. The task can proceed as follows:

1. Calibrate and validate GR4J and GR2M for several small watersheds using available precipitation, potential evapotranspiration and streamflow data and obtain the fitted parameter sets for each watershed.
2. Test performance of the model in filling gaps in streamflow time series.
3. Establish statistical relationships between model parameters and the watershed characteristics of the selected catchments,
4. Test the performance of parameter transfer approaches in simulating streamflow in ungauged locations.

Perhaps tasks (i) and (ii) can be seen as phase I and task (iii) and (iv) as phase II

4. Expected outcome:
Many countries combine their climatic and hydrological services. The aims of this internship project is to develop and exploit resources for the processing of hydrological data that parallel those for climatic variables. These combine the development of standard methods of analysis for streamflow data with the use of complex simulation models.

Applications for this internship should be submitted via the online application system, stating clearly the title of the internship.

Deadline for applications: February 28th, 2021 at 11:59 PM Central Africa Time (CAT)

Any inquiries about these internships should be sent to:
aims-ms4cr.internship@nexteinstein.org.