Estimating High Spatial Resolution Clear-Sky Land Surface Longwave Radiation Budget from MODIS and GOES Data

Abstract

The surface radiation budget (SRB) is important in addressing a variety of scientific and application issues related to climate trends, hydrological and biogeophysical modeling, and agriculture. The three longwave components of SRB are surface downwelling, upwelling, and net longwave radiation (LWDN, LWUP, and LWNT). Existing surface longwave radiation budget (SLRB) datasets have coarse spatial resolution and their accuracy needs to be greatly improved.

This study develops new hybrid methods for estimating instantaneous clear-sky high spatial resolution land LWDN and LWUP from the Moderate Resolution Imaging Spectroradiometer (MODIS, 1km) and the Geostationary Operational Environmental Satellites (GOES, 2-10 km) data. The hybrid methods combine extensive radiation transfer (physical) and statistical analysis (statistical) and share the same general framework. LWNT is derived from LWDN and LWUP.

This study is the first effort to estimate SLRB using MODIS 1 km data. The new hybrid methods are unique in at least two other aspects. First, the radiation transfer simulation accounted for land surface emissivity effect. Second, the surface pressure effect in LWDN was considered explicitly by incorporating surface elevation in the statistical models.

Nonlinear models were developed using the simulated databases to estimate LWDN from MODIS TOA radiance and surface elevation. Artificial Neural Network (ANN) models were developed to estimate LWUP from MODIS TOA radiance. The LWDN and LWUP models can explain more than 93.6% and 99.6% of variations in the simulated databases, respectively. Preliminary study indicates that similar hybrid methods can be developed to estimate LWDN and LWUP from the current GOES-12 Sounder data and the future GOES-R data.

The new hybrid methods and alternative methods were evaluated using two years of ground measurements at six validation sites from the Surface Radiation Budget Network (SURFRAD). Validation results indicate the hybrid methods outperform alternative methods. The mean RMSEs of MODIS-derived LWDN, LWUP, and LWNT using the hybrid methods are 16.88, 15.23, and 17.30 W/m2. The RMSEs of GOES-12 Sounder-derived LWDN and LWUP are smaller than 23.70 W/m2. The high spatial resolution MODIS and GOES SLRB derived in this study is more accurate than existing datasets and can be used to support high resolution numerical models.