Evaluating Cloud Contamination in Clear-Sky MODIS Terra Daytime Land Surface Temperatures Using Ground-Based Meteorology Station Observations

Abstract Environment Canada meteorological station hourly sampled air temperatures Tair at four stations in the southwest Yukon were used to identify cloud contamination in the Moderate Resolution Imaging Spectroradiometer (MODIS) Terra clear-sky daytime land surface temperature (LST) and emissivity daily level-3 global 1-km grid product (MOD11A1, Collection 5) that is not flagged by the MODIS quality algorithm as contaminated. The additional cloud masking used qualitative ground-based sky condition observations, collected at two of the four stations, and coincident MODIS quality flag information. The results indicate that air temperature observed at a variety of discrete spatial locations having different land cover is highly correlated with MODIS LST collected at 1-km grid spacing. Quadratic relationships between LST and air temperature, constrained by ground observations of “clear” sky conditions, show less variability than relationships found under “mainly clear” and “mostly cloudy” sky conditions, and the more clouds observed in the sky coincides with a decreasing y intercept. Analysis of MODIS LST and its associated quality flags show a cold bias (<0°C) in the assignment of the ≤3-K-average LST error, indicating MODIS LST has a maximum average error of ≤2 K over a warm surface (>0°C). Analysis of two observation stations shows that unidentified clouds in MODIS LST are between 13% and 17%, a result that agrees well with previous studies. Analysis of daytime values is important because many processes are dependent on daylight and maximum temperature. The daytime clear-sky LST–Tair relationship observed for the good-quality confirmed cloud-free-sky MODIS LST quality flag can be used to discriminate cloud-contaminated grid cells beyond the standard MODIS cloud mask.