Observed differences between near-surface air and skin temperatures using satellite and ground-based data

Abstract

Accurate estimates of long-term land surface temperature (T s) and near-surface air temperature (T a) at finer spatio-temporal resolutions are crucial for surface energy budget studies, for environmental applications, for land surface model data assimilation, and for climate change assessment and its associated impacts. The Atmospheric Infrared Sounder (AIRS) and Moderate Resolution Imaging Spectroradiometer (MODIS) sensors onboard the Aqua satellite provide a unique opportunity to estimate both temperatures twice daily at the global scale. In this study, differences between T a and T s were assessed locally over regions of North America from 2009 to 2013 using ground-based observations covering a wide range of geographical, topographical, and land cover types. The differences between T a and T s during non-precipitating conditions are generally 2-3 times larger than precipitating conditions. However, these differences show noticeable diurnal and seasonal variations. The differences between T a and T s were also investigated at the global scale using the AIRS estimates under clear-sky conditions for the period 2003-2015. The tropical regions showed about 5-20 C warmer T s than T a during the daytime , whereas opposite characteristics (about 2-5 C cooler T s than T a) are found over most parts of the globe during the night-time. Additionally, T s estimates from the AIRS and the MODIS sensors were inter-compared. Although large-scale features of T s were essentially similar for both sensors, considerable differences in magnitudes were observed (>6 C over mountainous regions). Finally, T a and T s estimates from the AIRS and MODIS sensors were validated against ground-based observations for the period of 2009-2013. The error characteristics notably varied with ground stations and no clear evidence of their dependency on land cover types or elevation was detected. However, the MODIS-derived T s estimates generally showed larger biases and higher errors compared to the AIRS-derived estimates. The biases and errors increased steadily when the spatial resolution of the MODIS estimates changed from finer 3 to coarser. These results suggest that representativeness error should be properly accounted for when validating satellite-based temperature estimates with point observations.

Intercomparison of In Situ Sensors for Ground-Based Land Surface Temperature Measurements

“…Whereas during the daytime, the LST was higher than T a and it varied from 0.86 to ~4.98 °C with highest values at Audubon for the entire dataset. This difference is within the range of values reported by [ 27 ] and [ 74 ] over many USCRN sites. Nighttime AT was a more reliable proxy for LST than daytime LST [ 12 ].…”

Modelagem da Temperatura do Ar na Região do Distrito Federal – Brasil, por meio de Dados Atmosféricos do Sensor AIRS

“…Além disso, a localização da estação INMET_1 foi o segundo critério adotado para a sua seleção, uma vez que ela se situa em área de uso do solo semelhante à estação FAL/UnB, a qual é composta por um mosaico de vegetação nativa de Cerrado e áreas de cultivo. Prakash et al (2018) verificaram menor amplitude térmica ao longo dos meses do ano em dados de temperatura em estações situadas em áreas de mosaico de vegetação nativa e áreas de cultivo agrícola, quando comparadas com áreas compostas principalmente de agricultura.…”

Seasonal Analysis of the 2011–2017 North American Monsoon near its Northwest Boundary

“…We also calculated average diurnal (or diel) cycles by season for weather and soil variables using hourly data aggregated from NevCAN measurements. These graphs normally do not include error bars (see [29] for a recent example) because their purpose is to visualize and summarize observations in an effective way, rather than to formally test for statistical significance Daily meteorological data collected at the site were obtained from the Western Regional Climate Center (https://wrcc.dri.edu/cgi-bin/rawMAIN.pl?idnes4). Weather variables included barometric pressure, precipitation, air temperature, and relative humidity at 2-m height.…”

“…We also calculated average diurnal (or diel) cycles by season for weather and soil variables using hourly data aggregated from NevCAN measurements. These graphs normally do not include error bars (see [29] for a recent example) because their purpose is to visualize and summarize observations in an effective way, rather than to formally test for statistical significance [30]. However, in order to highlight the variability of each hourly average, and also to indicate significant differences among seasons, we added two-standard error bars above and below each mean.…”