High‐resolution ensemble surface insolation estimates through assimilation of coarse‐scale retrievals into a simple physical model: 1. Physical model development and accuracy tests

Abstract

A simple physical radiative transfer insolation model (mG‐model) is developed for use in both prediction (part 1) and data assimilation applications (part 2). The primary novel aspect of the model is the use of the Visible Infrared Solar‐infrared Split window Technique (VISST) pixel‐level cloud product. The parsimony of the model and high‐resolution VISST cloud product allows for the computationally efficient prediction of high‐resolution solar radiation fields. The mG‐model, which has 0.02° spatial and half‐hourly temporal resolutions, was tested over the Southern Great Plains during the summer of 1997 and compared to measurements from the Atmospheric Radiation Measurement (ARM) project and the Global Energy Water Exchange (GEWEX) Continental Scale International Project and GEWEX Americas Prediction Project (GCIP/GAPP) Shortwave Radiation Budget (SRB) product. When investigating the insolation accuracy conditioned on cloud fraction, the mG‐model and the SRB data showed similar accuracy under clear and partly cloudy‐sky conditions. However, there are better results in the coarser SRB product under cloudy‐sky conditions, which makes the general performance of the SRB product better than that of the mG‐model. This result is the primary motivation for employing a data assimilation scheme to assimilate the SRB data into the mG‐model, which would combine the benefits of the mG‐model (higher resolution) and the SRB insolation product (resulting from a more complicated physical retrieval model) and provide dynamic error bound estimates that are reflective of input uncertainty. This data assimilation approach is described in part 2.