Land-surface temperature dynamics in the Upper Mekong Basin derived from MODIS time series

Kurzfassung

Land surface temperature (LST) is an important indicator for climate variability and can be sensed remotely by satellites with a high temporal resolution on a broad spatial scale. In this research, Moderate Resolution Imaging Spectroradiometer (MODIS) LST is used to derive a 13 year time series on the Upper Mekong Basin (UMB), belonging to the People’s Republic of China and the Republic of the Union of Myanmar, to analyse the spatial pattern and temporal development of LST. The data set shows the regular annual curve of surface temperatures with maximum values in summer and minimum values in winter. Average temperatures in the southern parts of the basin are higher than in the northern part. Spatial gradients between maximum and minimum LST as well as gradients between daytime and night-time LST are much lower in the southern parts than in the northern parts, which are characterized by a strong topography. The pixel-wise variability of monthly means was found to be in the range of ±4°C for most pixels in the daytime scenes, whereas the night-time scenes show a lower variability with most pixels in the range of ±1°C. The variability of LST in the northern areas clearly exceeds that in the southern areas. Some inter-annual variations occur, mainly during summer: in some years a two-peak distribution is found, which is explained by the generally low number of observations in the respective months. A primary challenge of optical satellite data in the UMB is cloud contamination in the summer months, where peak rainfall occurs. In the Mekong Highlands for instance, the average number of available daytime observations of MODIS LST in July is one observation per month only. It can be assumed that climate statistics calculated from such data is biased. In this context, two gap-filling algorithms were applied to two test areas for the year 2002 and results are discussed in the article. Another issue with MODIS LST data are day-to-day differences in the acquisition time. A temporal homogenization was applied to selected LST data, converting them to one fixed acquisition time. The converted data were compared to the original data set. No significant influence could be found.