Developing an optimization approach for estimating incident solar radiation at earth surface from multiple satellite data

• Main Author: Zhang, Yi
• Other Authors: Liang, Shunlin, Digital Repository at the University of Maryland, University of Maryland (College Park, Md.), Geography
• Format: Doctoral or Postdoctoral Thesis
• Language: English
• Published: 2019
• Subjects: Geographic information science and geodesy; Greenland
• Online Access: http://hdl.handle.net/1903/24998; https://doi.org/10.13016/hjrf-yvez

Surface incident shortwave radiation (ISR) is a crucial parameter in the land surface radiation budget. Many reanalysis, observation-based, and satellite-derived global radiation products have been developed but often have insufficient accuracy and spatial resolution for many applications. In this dissertation, I propose an optimization-based method (OB-Algorithm) based on a radiative transfer model for estimating surface ISR from Moderate Resolution Imaging Spectroradiometer (MODIS) Top of Atmosphere (TOA) observations by optimizing the surface and atmospheric variables with a cost function. This approach consisted of two steps: retrieving surface bidirectional reflectance distribution function parameters, aerosol optical depth (AOD), and cloud optical depth (COD); and subsequently calculating surface ISR. I also adapted the algorithm to VIIRS data and performed global validation with 34 Baseline Surface Radiation Network (BSRN) sites for both instantaneous and daily mean ISR. Researches on estimating daily and diurnal ISR was also made on Advanced Himawari Imager (AHI) and Advanced Baseline Imager (ABI). Geostationary satellites capture diurnal ISR variation better than polar-orbiting ones, especially in cloudy cases. Validation against measurements at seven Surface Radiation Budget Network (SURFRAD) sites resulted in an R2 of 0.91, a bias of -6.47 W/m2, and a root mean square error (RMSE) of 84.17 W/m2 for the instantaneous results. Validation at eight high-latitude snow-covered Greenland Climate Network (GC-Net) sites resulted in an R2 of 0.86, a bias of -21.40 W/m2, and an RMSE of 84.77 W/m2. These validation results show that the proposed method is much more accurate than the previous studies (usually with RMSEs of 80-150W/m2). The VIIRS ISR results at seven SURFRAD showed RMSEs of 83.76 W/m2 and 27.78 W/m2 for instantaneous and daily ISR, respectively at SURFRAD sites. Results at 34 BSRN sites showed RMSEs of 106.68 W/m2 and 32.76 W/m2 for instantaneous and daily ISR, respectively at BSRN sites. ...