Date

2019-11-28

Author

Yousufi, Kaveh Patakchi
Yılmaz, Mustafa Tuğrul
Ozturk, Kurtulus
Yücel, İsmail
Yılmaz, Koray Kamil

Metadata

Show full item record

Item Usage Stats

235
views

0
downloads

Cite This

Radar-based precipitation estimates rely on algorithms that utilize observations of backscattered waves sent from the radar. By utilizing these backscattered observations and different reflectivity-rainfall (Z-R) equations, radar-based precipitation estimates are obtained. However, this process contains uncertainties due to used Z-R equations or other error sources such as beam blockage, attenuation, etc. on the radar beams. The majority of the studies in the literature focused on reducing errors caused by Z-R relation and radar beam correction. This study investigates relative performances of different statistical algorithms to improve radar-based quantitative precipitation estimation (QPE) using ground observation data, rather than focusing on the evaluation of different Z-R relationships and reflectivity corrections commonly implemented in the literature. Initial investigations are implemented using datasets obtained over Antalya radar, while this study will be extended using all 18 radar-based observations acquired over different regions of Turkey. Implemented statistical methodologies include Mean Field Bias (MFB), Local Multiplicative Bias (LMB), Local Additive Bias (LAB), Local Mixed Bias (LMIB), and Multiple Linear Regression (MLR) adjustment techniques. To test the performance of these algorithms, cross-validation methods have been used. In cross-validation, 50%, 25%, 12.5% of the station-based observations are excluded for validation while the remaining are used for the calibration in different experiments. Both the calibration and validation results obtained from all rainfall events of 2015 for Antalya radar suggests LMB, LMIB, and LAB adjustment methods perform better in terms of agreement with the rain-gauge observations. However, LMB doesn’t contribute significantly to the improvement of the correlation coefficient. The advantage of the proposed time-independent MLR method is that does not require any gauge observation for real-time radar estimation correction, and it is quite successful in addressing the radar underestimation problem on a monthly scale. However, the spatial distribution of the RMSE and Correlation Coefficient plots suggest multiplicative bias correction methods (MFB, LMB, and MLR) fail to contribute to improving the accuracy of the radar QPE in the areas with a partial or complete radar beam blockage.

URI

http://uzalmet.mgm.gov.tr/Default.aspx
https://hdl.handle.net/11511/84518

Collections

Department of Civil Engineering, Conference / Seminar