Evaluation of MODIS land surface temperature products for daily air surface temperature estimation in northwest Vietnam

Abstract

ABSTRACTRecently, the MODerate resolution Imaging Spectroradiometer (MODIS) land surface temperature (LST) is considered one of the most suitable ways to retrieve air surface temperature (Ta) – one of the most important and widely used climate variables for a wide range of applications. In fact, many successful studies have been reported in many regions of the world. Each day, four MODIS LST data are available; from two sensors (Terra and Aqua) at two local overpass times (daytime and night-time). However, due to their different overpass times, most studies have used LST daytime and night-time for daily maximum (Tmax) and minimum (Tmin) air surface temperature estimation, respectively. Therefore, the performance of each individual LST data, the effect of the dynamic combination of these four LST data, the effect of land surface characteristics, and the effect of LST quality on the same estimation in the same region on the accuracy of estimated Ta remains unclear. In this study, we evaluated and tested all individual LST data as well as all possible combinations of the four MODIS LST data from two separate stations with distinct land surface characteristics in northwest Vietnam for 10 years (from 2004 to 2013) under two sky conditions (all clear sky conditions and only good data – i.e. Quality Control (QC) value of 0) for daily Ta (Tmax, Tmin, and Tmean) estimations. In addition, the mixed data of the two stations were also evaluated. Our results showed that Terra LST data have a higher correlation with Ta than Aqua LST; which is consistent for both stations and both quality conditions (all clear sky and only good data). A closer overpass time with Tmax or Tmin occurrence time did not guarantee a higher accuracy of Ta estimation. Using only good LST data produced a higher accuracy of Ta estimation than using all clear sky data. However, if the percentage of good data is low (i.e. less than 30%), then the all clear sky data will provide better results for Tmax estimation. Comparing the performance of the different combinations when using the single station and mixed station data, combinations including at least one night-time LST produced stable and high accuracy Tmin and Tmean estimates, while the combinations with only daytime LST produced very low accuracy results. For Tmax estimation, the results were less impacted by LST quality; however, they were strongly impacted by different combinations and land surface characteristics.