Uncertainties of gridded global precipitation products for hydrological modelling at regional scale

Abstract

Rainfall is a fundamental input for many hydrological and water management studies. In-situ observations are considered an essential source of information in hydrometeorology. However, in recent decades, satellite precipitation products are preferred owing to their additional benefits, such as freely available data with a high spatial and temporal resolution, global coverage, cost effectiveness, covers mountains and oceanic regions, and, more importantly, uninterrupted data. Indeed, a variety of satellite precipitation products are available nowadays; however, the applicability of such products is still limited due to their lack of accuracy with ground-based measurements. In addition, hydrologists and researchers often have problems choosing the optimally performing product suited for the particular research problem at the catchment scale. The present thesis evaluated the potential of these precipitation products at a regional scale.

In the first part of the thesis, the ability of three satellite precipitation products (TRMM-3B42v7, PERSIANN-CDR and GPM-IMERGv6) were evaluated with ground-based IMD gridded data and also by driving the Soil Water Assessment Tool (SWAT) hydrological model for a monsoon driven coastal river basin in the southeast of India at different temporal scales. The results indicate that IMERG exhibited the lowest systematic errors among the three, followed by TRMM-3B42 v7 at different temporal scales. All precipitation products showed better performance at the monthly time scale, and all products overestimated precipitation at daily, monthly and yearly time scales. Hydrological evaluation of the three products using the SWAT model over the study area showed unsatisfactory results at a daily scale during the validation period. IMERG is found to have better performance than the other products in the hydrological evaluation as well.

In the second part of the thesis, a quantile based Bayesian model averaging (QBMA) approach was developed to merge these satellite precipitation products to enhance the accuracy of the precipitation products. The QBMA approach was compared with traditional methods, namely, simple model averaging and one outlier removed. Results indicated that the bias-corrected QBMA outperformed the other methods. On monthly evaluation, it is observed that all the products perform better during July and September than that in June and August. The QBMA approaches do not significantly improve the SMA approach in terms of Probability of Detection. However, the bias-corrected QBMA products have a lower False Alarm Ratio. The developed QBMA approach with bias-corrected inputs outperforms the IMERG product in terms of RMSE.

In the third part of the thesis, an alternative method to calibrate the SWAT eco-hydrological model using available satellite-based GLEAM evapotranspiration data for the data-sparse Oueme river basin in Benin, West Africa, was explored. Twelve different calibrations were implemented with different ground-based, satellite and reanalysis precipitation products to obtain a high performing model for the Cotonou City, Benin. The rainfall features, model performances, parameter uncertainty, prediction uncertainty and the potential relationships among the above component were analysed. Overall, the results show that global satellite-based data can be used as an alternative method to calibrate and SWAT model in a sparsely gauged coastal river basin. Specifically, all the Percent Bias values suggested satisfactory model performances for all the precipitation products. Uncertainty analysis indicates that the precipitation inputs are the main source of uncertainty in the model.

Overall, the results from investigations conducted in the thesis strongly highlight the potential for global precipitation products and the associated uncertainties for hydrological modelling at the regional scale.