Disaggregating National Re-Analysis Air Temperature using Remotely-Sensed Land Surface Temperature

Meteorological re-analysis data, such as the National Land Data Assimilation System (NLDAS), are very useful in many operations and applications. For example, we are using NLDAS air temperature and humidity to develop historical daily heat metrics over the coterminous U.S. Although the resolution of NLDAS is nominally 1/8 degree (~12 km), it is in reality coarser because the NLDAS variables are created via spatial interpolation of the 32 km North American Regional Re-analysis (NARR). Thus, most urban-scale features, such as the urban heat island, are not captured by NLDAS. In many applications, such as human health or electrical energy demand, finer-resolution data are required.

In an exploratory effort to create a higher-resolution version of NLDAS air temperature data, we have implemented an algorithm that imposes 1 km spatial variations in MODerate Resolution Imaging Spectroradiometer (MODIS) Land Surface Temperature (LST) onto the coarser-resolution NLDAS air temperatures. The approach is based on the assumption that in the absence of strong horizontal temperature advection, air temperature is driven by sensible heat flux from the surface, thus the spatial patterns of air temperature mimic the patterns of LST. Air temperature variations, however, are much smaller in magnitude than corresponding LST variations. Therefore, our method computes and applies normalized MODIS LST spatial anomalies to disaggregate daily maximum NLDAS air temperature.

In our approach to disaggregate NLDAS air temperatures to a 1 km grid, the spatial pattern of LST is represented by a grid array averaged over all available arrays for a season. As a demonstration, we used a mean summer (June – August) LST gridded data set generated from daytime Aqua MODIS LST data for all summer days during 2003-2008. This data set clearly captures urban-rural temperature variations, which makes it well-suited for the intended purpose of developing daily heat metrics that represent these spatial patterns of heat. This LST product, being generated as a seasonal composite over several years, represents the typical LST pattern for summer; use of this in the down-scaling algorithm requires the assumption that relative spatial patterns of air temperature at the sub-NLDAS scale (< 12 km) are relatively constant from day to day within the respective season.

In this presentation, the disaggregation algorithm will be described in detail. The creation of the summer composite MODIS LST data set will be described and examples shown. Finally, examples of disaggregated (1 km) NLDAS air temperature, and associated validation statistics, will be presented.