Global seamless and high-resolution temperature dataset (GSHTD), 2001–2020

Abstract

Land surface temperature (Ts) and near-surface air temperature (Ta) are key parameters in multiple research fields. In this study, a global seamless (without missing values) and high-resolution (30 arcsecond spatial resolution) temperature (for both Ts and Ta) dataset (GSHTD) from 2001 to 2020 was developed. First, a method called the estimation of the temperature difference (ETD) was proposed to reconstruct both clear- and cloudy-sky Ts. A global seamless 8-day and monthly average all- and clear-sky Ts data were then created using the MODIS Ts data and the ETD method. The seamless monthly average of the mean, maximum and minimum Ta data were further developed using the seamless Ts data, in situ Ta data and Cubist machine learning algorithm. GSHTD has four main advantages. First, GSHTD includes seven types of temperature data: clear-sky daytime and nighttime Ts, all-sky daytime and nighttime Ts, and mean, maximum and minimum Ta. Second, it has global coverage and high spatial resolution. Third, using the ETD method proposed in this study, GSHTD has no missing values. Fourth, the accuracy of GSHTD is high; the average mean absolute errors (MAEs) of ETD in reconstructing the 25 × 25 and 150 × 150 pixel clear-sky daytime (nighttime) Ts data were 0.724 (0.552) and 1.024 (0.895) °C, respectively. The MAEs of ETD were on average 23.2% and 23.7% lower than those of Remotely Sensed DAily land Surface Temperature reconstruction (RSDAST) and interpolation of the mean anomalies (IMAs). The MAEs of the estimated monthly average of the mean, maximum and minimum Ta data were 0.797, 0.994 and 1.056 °C, respectively. The developed GSHTD is freely available at Middle Yangtze River Geoscience Date Center (https://cjgeodata.cug.edu.cn/#/pageDetail?id=97), which will be useful in many studies related to climate change, environmental science and ecology, and epidemiology and human health.