Electronic Thesis and Dissertation Repository

Thesis Format



Master of Science


Civil and Environmental Engineering


Najafi, Mohammad Reza


Reliable estimation of precipitation, as the most important variable in hydrological modelling, is crucial for water resources management. Rain gauges that provide precipitation measurements at point scale have inherent limitations and difficulties in remote regions and complex terrains due to accessibility, gauge undercatch, among others. Alternatively, satellite and radar precipitation data can estimate precipitation at high spatial and temporal resolution by utilizing several types of space and ground-borne sensors. However, due to the indirect estimates of precipitation by remotely sensed products, their measurements are subject to systematic biases and are required to be evaluated and bias adjusted before using at a specific area.

This study investigates the performance of multiple high-resolution remotely sensed precipitation estimates at hourly and daily time scales over Canada for 2014-2018. Four products of the recently released Integrated Multi-satEllite Retrievals for Global precipitation measurement (IMERG-V06) and the Multi-Radar Multi-Sensor (MRMS) Precipitation Rate data for different seasons are analyzed. Evaluations are based on a suite of metrics to assess different characteristics of precipitation estimates using quality-controlled hourly gauge data considered as the truth. The results suggest that Calibrated precipitation (PrCal) outperforms the other IMERG products and estimates the amount of precipitation relatively well particularly over the Prairies and during fall and summer. Over the western and eastern coastal regions, IMERG tends to overestimate precipitation intensities by around 25%. The discrepancy between satellite and ground-based data is higher for more intense precipitation events. Further analyses indicate that while MRMS tends to overestimate the amount of precipitation, it outperforms the IMERG products based on several metrics, especially in detecting the occurrence of precipitation over the eastern coastal regions. Overall, the study of IMERG V06 and MRMS precipitation estimates at a relatively high temporal resolution indicates that both products have the potential to complement ground-based observations over Canada.

Further, a regression quantile mapping method is developed to adjust the systematic spatial and temporal biases of IMERG PrCal across Canada. For this purpose, several climatic and topographic explanatory variables are resampled and applied in the regression-based model to extend satellite bias correction over the ungauged pixels. The proposed method shows promising results by reducing RBias (by ~32%) and increasing correlation values (by ~ 0.31). The bias-corrected precipitation product (for 2014-2018)can be applied by researchers and various stakeholders, across Canada, for the analysis of extreme precipitation events, water resources management, design of infrastructure, among others.

Finally, the application of daily IMERG data in streamflow simulation is demonstrated by using the original data to drive the calibrated Raven rainfall-runoff model over the Bathewana watershed in southern Ontario for 2001-2015. By comparing with the observed flow, the obtained results indicate that IMERG tends to underestimate the streamflow, however, it is able to preserve its temporal variation reasonably well. Overall, results suggest that further improvements of IMERG data should be considered by its algorithm developers to enhance the quality of this product in cold weather conditions.

Summary for Lay Audience

Precipitation is the most important component in hydrological applications. Therefore, reliable measurement of precipitation is crucial for having more accurate monitoring of water resources supplies and forecasting extreme weather events such as floods. However, due to the high spatiotemporal variability of precipitation, its accurate estimation is a challenging task especially over complex terrain where the ground-based rain gauges are either sparse or nonexistent. Recently for dealing with the limitations of ground-based stations availability, remotely sensed algorithms that use satellite and radar data have been developed to estimate precipitation. Nevertheless, the remote sensing-based data need to be evaluated before using due to the indirect nature of their estimates. The most well-known and recently released satellite-based precipitation products named Integrated Multi-satEllite Retrievals for Global precipitation measurement (IMERG-V06) and the Multi-Radar Multi-Sensor (MRMS) data are evaluated in this study to investigate the performance of such a high spatiotemporal resolution precipitation data over Canada. Although the findings of this study indicate the promising value of satellite and radar precipitation over most parts of the country, it still shows bias in some regions. Therefore, a Regression-based Quantile Mapping (RQM) method is developed to correct the biases associated with the IMERG data spatially and temporally over the entire country. The proposed framework can significantly improve the IMERG data in different regions during the study period (2014-2018) and provide a high quality of precipitation data over Canada. In addition to statistical evaluations and bias correction, the ability of IMERG precipitation in daily streamflow simulation is assessed by forcing it in a calibrated hydrological model. For this purpose, the Raven model calibrated by using the ground-based rainfall data over the Batchawana as a small watershed (1280 km2) located in the southern part of Ontario, Canada is selected. Due to the error of input IMERG precipitation as well as the uncertainty of the calibrated Raven model, the output simulated streamflow is not promising. However, simulated streamflow by forcing IMERG data can capture the trend of observed discharge reasonably. Overall this study provides insights into remotely-sensed data over Canada and helps to have a high spatiotemporal resolution of precipitations.

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.