Sensitivity of Landsat 8 Surface Temperature Estimates to Atmospheric Profile Data: A Study Using MODTRAN in Dryland Irrigated Systems

Abstract

Abstract:The land surface temperature (LST) represents a critical element in efforts to characterize global surface energy and water fluxes, as well as being an essential climate variable in its own right. Current satellite platforms provide a range of spatial and temporal resolution radiance data from which LST can be determined. One of the most complete records of data comes via the Landsat series of satellites, which provide a continuous sequence that extends back to 1982. However, for much of this time, Landsat thermal data were provided through a single broadband thermal channel, making surface temperature retrieval challenging. To fully exploit the valuable time-series of thermal information that is available from these satellites requires efforts to better describe and understand the accuracy of temperature retrievals. Here, we contribute to these efforts by examining the impact of atmospheric correction on the estimation of LST, using atmospheric profiles derived from a range of in-situ, reanalysis, and satellite data. Radiance data from the thermal infrared (TIR) sensor onboard Landsat 8 was converted to LST by using the MODTRAN version 5.2 radiative transfer model, allowing the production of an LST time series based upon 28 Landsat overpasses. LST retrievals were then evaluated against in-situ thermal measurements collected over an arid zone farmland comprising both bare soil and vegetated surface types. Atmospheric profiles derived from AIRS, MOD07, ECMWF, NCEP, and balloon-based radiosonde data were used to drive the MODTRAN simulations. In addition to examining the direct impact of using various profile data on LST retrievals, randomly distributed errors were introduced into a range of forcing variables to better understand retrieval uncertainty. Results indicated differences in LST of up to 1 K for perturbations in emissivity and profile measurements, with the analysis also highlighting the challenges in modeling aerosol optical depth (AOD) over arid lands and its impact on the TIR bands. Days with high AOD content (AOD > 0.5) in the evaluation study seem to consistently underestimate in-situ LSTs by 1-2 K, suggesting that MODTRAN is unable to accurately simulate the aerosol conditions for the TIR bands. Comparisons between available in-situ and Landsat 8 derived LST illustrate a range of seasonal and land surface dynamics and provide an assessment of retrieval accuracy throughout the nine-month long study period. In terms of the choice of atmospheric profile, when excluding the in-situ data, results show a mean absolute range of between 1.2 K to 1.8 K over bare soil and 3.3 K to 3.8 K over alfalfa for the different meteorological forcing, with the AIRS profile providing the best reproduction over the studied arid land irrigation region.

Evaluating Eight Global Reanalysis Products for Atmospheric Correction of Thermal Infrared Sensor—Application to Landsat 8 TIRS10 Data

“…When the water vapor content was less than 2 g/cm 2 , the atmospheric upward radiance and downward radiance differences under the four boundary layer aerosol models increased with an increase of water vapor content, and the trend was the opposite when the water vapor content was greater than 2 g/cm 2 . The four boundary layer aerosol models have different effects on three atmospheric parameters, which is consistent with the research of Rosas et al [20], showing that aerosol optical depth content had an impact on LST estimations over arid land irrigation regions. If the visibility (e.g., Rural 5) is lower than the settings (e.g., Rural 23) when running MODTRAN, the atmospheric transmittance will be overestimated, and this may lead to an underestimation of LST.…”

“…According to the MOD05 product statistics, most of the water vapor content in this area was less than 2 g/cm 2 ; thus, the small LST difference in the real data is reasonable. In addition, the validation results are also consistent with the study by Rosas et al [20]; the LST RMSEs were 1.58 K and 1.56 K for ERA-Interim and NCEP/NCAR Reanalysis 1 over the arid land irrigation region. Although the overall biases (RMSEs) at validation sites between different reanalysis products were less than 0.25 K (0.1 K), the difference across the entire scene is more noticeable.…”

Evaluation of Seven Atmospheric Profiles from Reanalysis and Satellite-Derived Products: Implication for Single-Channel Land Surface Temperature Retrieval

“…The reanalysis profiles described in Section 2.1 were provided at different spatial and temporal resolutions. To obtain atmospheric parameters pixel-by-pixel in a satellite TIR image, three main procedures were used to perform the interpolation of the reanalysis profiles in elevation, space, and time, as follows [23,46]. Figure 4 shows the schematic diagram of the interpolation of the reanalysis profiles in space and time.…”

Capturing the Diurnal Cycle of Land Surface Temperature Using an Unmanned Aerial Vehicle

“…where LST ( • C) is the land surface temperature, T sky (K) is the downwelling sky irradiance temperature, and ε is the surface emissivity [55]. Although there are several ways to estimate T sky [56], we employ the relatively common approach of using the temperature from the aluminum GCPs located in the field [57][58][59][60].…”