Daytime Land Surface Temperature Extraction from MODIS Thermal Infrared Data under Cirrus Clouds.

Xiwei Fan,Zhao-Liang Li,Hua Wu,Guangjian Yan,Bo-Hui Tang

doi:10.3390/s150509942

Abstract

Simulated data showed that cirrus clouds could lead to a maximum land surface temperature (LST) retrieval error of 11.0 K when using the generalized split-window (GSW) algorithm with a cirrus optical depth (COD) at 0.55 μm of 0.4 and in nadir view. A correction term in the COD linear function was added to the GSW algorithm to extend the GSW algorithm to cirrus cloudy conditions. The COD was acquired by a look up table of the isolated cirrus bidirectional reflectance at 0.55 μm. Additionally, the slope k of the linear function was expressed as a multiple linear model of the top of the atmospheric brightness temperatures of MODIS channels 31–34 and as the difference between split-window channel emissivities. The simulated data showed that the LST error could be reduced from 11.0 to 2.2 K. The sensitivity analysis indicated that the total errors from all the uncertainties of input parameters, extension algorithm accuracy, and GSW algorithm accuracy were less than 2.5 K in nadir view. Finally, the Great Lakes surface water temperatures measured by buoys showed that the retrieval accuracy of the GSW algorithm was improved by at least 1.5 K using the proposed extension algorithm for cirrus skies.

Highlights

Land surface temperature (LST) is an important parameter because of its control on the upward terrestrial radiation and the energy exchange between the Earth’s surface and the atmosphere [1,2].Satellite remote sensing offers the only possibility to measure LST over extended regions with high temporal and spatial resolutions [3]
Using the cirrus optical depth (COD) that can be interpolated from the isolated cirrus bidirectional reflectance (ICBR) look up table (LUT) in actual applications, and with the slope k determined from TOA brightness temperatures in MODIS channels 31–34 and split-window channel emissivities, the LST retrieve accuracy of the generalized split-window (GSW) algorithm under cirrus skies is improved using the correction term LSTCOD − k*COD = LSTAC
Conventional LST retrieval algorithms from TIR data are developed and applied for clear-sky conditions, and the influences of cirrus clouds are not considered in their development

Summary

Introduction

Land surface temperature (LST) is an important parameter because of its control on the upward terrestrial radiation and the energy exchange between the Earth’s surface and the atmosphere [1,2].Satellite remote sensing offers the only possibility to measure LST over extended regions with high temporal and spatial resolutions [3]. Because of the complex influences of cirrus clouds on atmospheric radiative transfer and relatively low atmospheric transmittances, the current LST retrieval algorithms using satellite thermal-infrared (TIR) data typically do not consider the influences of cirrus clouds and are only applied for clear-sky conditions. Cirrus clouds, which are relatively optically thin, have a global coverage of approximately 20%, with over 60%–70% in the tropics [4], and a thin cirrus layer may be present as much as 80% of the time in tropical regions [5]. It has been noted that globally distributed high and thin cirrus clouds introduce serious retrieval difficulties of atmospheric temperature and humidity profiles and surface geophysical parameters from space-based platforms, owing to the semitransparency of these clouds at visible and infrared wavelengths [6,7,8]. Regarding the effects of cirrus clouds on estimates of sea surface temperature (SST), Xu and Sun [8] indicated an error of 1.5–2.0 K on SST retrieval occurs if clear-sky

Objectives

Methods

Results

Conclusion