Land surface temperature over global deserts: Means, variability, and trends

Abstract

Land surface air temperature (LSAT) has been a widely used metric to study climate change. Weather observations of LSAT are the fundamental data for climate change studies and provide key evidence of global warming. However, there are very few meteorological observations over deserts due to their uninhabitable environment. This study fills this gap and provides independent evidence using satellite‐derived land surface temperatures (LSTs), benefiting from their global coverage. The frequency of clear sky from MODerate Resolution Imaging Spectroradiometer (MODIS) LST data over global deserts was found to be greater than 94% for the 2002–2015 period. Our results show that MODIS LST has a bias of 1.36°C compared to ground‐based observations collected at 31 U.S. Climate Reference Network (USCRN) stations, with a standard deviation of 1.83°C. After bias correction, MODIS LST was used to evaluate existing reanalyses, including ERA‐Interim, Japanese 55‐year Reanalysis (JRA‐55), Modern‐Era Retrospective Analysis for Research and Applications (MERRA), MERRA‐land, National Centers for Environmental Prediction (NCEP)‐R1, and NCEP‐R2. The reanalyses accurately reproduce the seasonal cycle and interannual variability of the LSTs, but their multiyear means and trends of LSTs exhibit large uncertainties. The multiyear averaged LST over global deserts is 23.5°C from MODIS and varies from 20.8°C to 24.5°C in different reanalyses. The MODIS LST over global deserts increased by 0.25°C/decade from 2002 to 2015, whereas the reanalyses estimated a trend varying from −0.14 to 0.10°C/decade. The underestimation of the LST trend by the reanalyses occurs for approximately 70% of the global deserts, likely due to the imperfect performance of the reanalyses in reproducing natural climate variability.

A surface temperature and moisture intercomparison study of the Weather Research and Forecasting model, in‐situ measurements and satellite observations over the Atacama Desert

“…Remote areas, such as semi‐arid and desert regions, often lack ground‐based observational networks (Zhou and Wang, ), meaning that the vast majority of estimates of surface and near‐surface meteorological parameters are obtained from satellite‐derived data (Li et al, , ; Li and Duan, ). However, these datasets are known to have significant biases (e.g., Yu et al, ; Ermida et al, ), which compromise any attempt to assess and calibrate the performance of a numerical model.…”

“…Northern Chile also exhibits significant variability on decadal time-scales (Schulz et al, 2012). As opposed to heavily populated regions, arid and semi-arid areas generally lack ground-based observational networks (e.g., Li et al, 2016;Zhou and Wang, 2016). Here, instead, surface and near-surface variables are generally estimated from satellite-derived products (Tomlinson et al, 2011;Li and Duan, 2018) that can have significant discrepancies with respect to in-situ observations (e.g., Trigo et al, 2008;Yu et al, 2014;Ermida et al, 2017).…”

New Methods for Deriving Clear‐Sky Surface Longwave Downward Radiation Based on Remotely Sensed Data and Ground Measurements

“…As pointed above, the retrieval uncertainty may exist in the remote sensing based PWC and LST products (Guillevic et al, 2012;Liu et al, 2015;W. Yu, Ma, et al, 2014;Zhou & Wang, 2016). To analysis the influence of PWC and LST's uncertainty on the proposed methods, the nonbias uncertainties were added to the PWC (with 1st, 3rd, 5th, 7th, or 10th percentile maximum uncertainties ranges) and LST (with 0.5, 1, 1.5, 2, 2.5, and 3 K maximum uncertainties ranges), respectively.…”

Drought risk in the Bolivian Altiplano associated with El Niño Southern Oscillation using satellite imagery data

“…The mean temperature of the gauged data was calculated using the arithmetic mean between the maximum and minimum temperature. The relationship enables to correct the LST with a linear regression equation (Zhou and Wang, 2016). The regression performance was evaluated using the relative ME or bias and the E coefficient.…”