Repository logo
 

Radar and satellite observations of precipitation: space time variability, cross-validation, and fusion

Date

2017

Authors

Chen, Haonan, author
Chandrasekar, V., advisor
Reising, Steven C., committee member
Cheney, Margaret, committee member
Mielke, Paul W., committee member

Journal Title

Journal ISSN

Volume Title

Abstract

Rainfall estimation based on satellite measurements has proven to be very useful for various applications. A number of precipitation products at multiple time and space scales have been developed based on satellite observations. For example, the National Oceanic and Atmospheric Administration (NOAA) Climate Prediction Center has developed a morphing technique (i.e., CMORPH) to produce global precipitation products by combining existing space-based observations and retrievals. The CMORPH products are derived using infrared (IR) brightness temperature information observed by geostationary satellites and passive microwave-(PMW) based precipitation retrievals from low earth orbit satellites. Although space-based precipitation products provide an excellent tool for regional, local, and global hydrologic and climate studies as well as improved situational awareness for operational forecasts, their accuracy is limited due to restrictions of spatial and temporal sampling and the applied parametric retrieval algorithms, particularly for light precipitation or extreme events such as heavy rain. In contrast, ground-based radar is an excellent tool for quantitative precipitation estimation (QPE) at finer space-time scales compared to satellites. This is especially true after the implementation of dual-polarization upgrades and further enhancement by urban scale X-band radar networks. As a result, ground radars are often critical for local scale rainfall estimation and for enabling forecasters to issue severe weather watches and warnings. Ground-based radars are also used for validation of various space measurements and products. In this study, a new S-band dual-polarization radar rainfall algorithm (DROPS2.0) is developed that can be applied to the National Weather Service (NWS) operational Weather Surveillance Radar-1988 Doppler (WSR-88DP) network. In addition, a real-time high-resolution QPE system is developed for the Engineering Research Center for Collaborative Adaptive Sensing of the Atmosphere (CASA) Dallas-Fort Worth (DFW) dense radar network, which is deployed for urban hydrometeorological applications via high-resolution observations of the lower atmosphere. The CASA/DFW QPE system is based on the combination of a standard WSR-88DP (i.e., KFWS radar) and a high-resolution dual-polarization X-band radar network. The specific radar rainfall methodologies at Sand X-band frequencies, as well as the fusion methodology merging radar observations at different temporal resolutions are investigated. Comparisons between rainfall products from the DFW radar network and rainfall measurements from rain gauges are conducted for a large number of precipitation events over several years of operation, demonstrating the excellent performance of this urban QPE system. The real-time DFW QPE products are extensively used for flood warning operations and hydrological modelling. The high-resolution DFW QPE products also serve as a reliable dataset for validation of Global Precipitation Measurement (GPM) satellite precipitation products. This study also introduces a machine learning-based data fusion system termed deep multi-layer perceptron (DMLP) to improve satellite-based precipitation estimation through incorporating ground radar-derived rainfall products. In particular, the CMORPH technique is applied first to derive combined PMW-based rainfall retrievals and IR data from multiple satellites. The combined PMW and IR data then serve as input to the proposed DMLP model. The high-quality rainfall products from ground radars are used as targets to train the DMLP model. In this dissertation, the prototype architecture of the DMLP model is detailed. The urban scale application over the DFW metroplex is presented. The DMLP-based rainfall products are evaluated using currently operational CMORPH products and surface rainfall measurements from gauge networks.

Description

Rights Access

Subject

machine learning
remote sensing
weather radar network
quantitative precipitation estimation (QPE)
dual-polarization radar
satellite

Citation

Associated Publications