Integration of satellite-measured LST data into cokriging for temperature estimation on tropical and temperate islands


  • Sunyurp Park

    Corresponding author
    1. Department of Satellite Geoinformatics Engineering, Kyungil University, Gyeongsan, Republic of Korea
    • Department of Satellite Geoinformatics Engineering, Kyungil University, Gyeongsan, Republic of Korea.
    Search for more papers by this author


Cokriging, a multivariate geostatistical interpolation method, uses elevation information as a secondary variable to improve the prediction accuracy of air temperature in mountainous areas. Although the secondary variable plays an important role in cokriging, the performance of interpolation largely depends on the amount of input data. The purpose of this study was to improve air temperature estimation by merging hypertemporal Moderate Resolution Imaging Spectroradiometer land surface temperature (LST) data and ground observations as interpolation data. Two significantly different island environments, tropical Hawaii Island (USA) and temperate Jeju Island (Korea), were selected for interpolation experiments. Spatiotemporal characteristics of air temperature prediction were compared between three cokriging methods and the conventional inverse-distance weighted interpolation. Due to the year-round trade winds, there was significant difference in prediction errors between windward and leeward slopes of Hawaii Island. LST-derived temperature was not regularly sampled from cloud-prone, windward slopes. As a result, prediction reliability was lower on windward slopes than leeward slopes, and overall prediction accuracy decreased in the wet season. Jeju Island is a mid-latitude volcanic island heavily influenced by the Asian monsoons. This climatic setting creates the seasonal variations of air temperature that is far greater than its spatial variations. The environmental lapse rate (ELR) of Jeju Island became much steeper in winter, and prediction accuracy and reliability were reduced due to an increase in the spatial variations of air temperature. With the addition of satellite-derived air temperature data, the root mean square errors of cokriging decreased by 27.3–52.9% for Hawaii Island and 34.6–37.6% for Jeju Island depending on cokriging models. Copyright © 2010 Royal Meteorological Society