International Journal of Applied Earth Observation and Geoinformation
A simple retrieval method of land surface temperature from AMSR-E passive microwave data—A case study over Southern China during the strong snow disaster of 2008
Research highlights
▸ Brightness temperatures linearly correlates with observation temperature. ▸ MPDI is an effective indicator for characterizing the surface vegetation cover. ▸ Smaller MPDI intervals can obtain higher accuracy of AMSR-E LST simulation. ▸ Only 7 polarization bands and 5 surface types are required by the simplified model. ▸ Average LST retrieval error is 0.91–1.30 °C.
Introduction
Since 1990, five strong cold disasters have occurred in Guangdong Province of Southern China, causing serious economic losses (Wang et al., 2004). In 2008, another strong snow disaster attacked the province again, which is one of the worst storms of the past 50 years. As a result, timely acquirement of regional temperature information on a large scale is becoming more and more urgent for emergency management in such situations. This has recently made the remote sensing of land surface temperature (LST) an important research subject in China. Many methodologies have been established to retrieve LST from thermal infrared satellite sensor data (Mao et al., 2007a, Mao et al., 2007b). However, the thermal remote sensing is greatly influenced by cloud, atmospheric water content and rainfall. Therefore, thermal remote sensing from optical sensors cannot be used to retrieve LST during the periods of cold disasters or under other bad weather conditions. Microwave remote sensing can just overcome these disadvantages. Passive microwave emission can penetrate non-precipitating clouds, thereby providing a better representation of LST under nearly all sky conditions. What is more, daily data are available from microwave radiometers as compared to optical sensors like Landsat™, ASTER or MODIS of which only weekly series products are available. The coarse spatial resolution of passive microwave remote sensing is not a problem for large scale studies and therefore providing nearly 20-year time series by now, which are of great interest for recent climate change studies (Fily et al., 2003).
Passive microwave remote sensing has already been used to retrieve LST for almost 20 years. McFarland et al. (1990) made some significant conclusions: LST for crop/range, moist soils, and dry soils surface types can be retrieved with linear regression models from passive microwave SSMI/I brightness temperatures (BT). The BT of SSMI/I 85 GHz vertical polarization is the primary channel for LST correlation. The 19 GHz band can compensate for the influence of surface water. The difference between 37 and 22 GHz can be utilized to correct the influence of atmospheric water vapor content on the emission. Njoku (1993) found that neural network method was more appropriate for developing a useful LST retrieval algorithm. Multi-channel measurements can estimate and correct the surface emissivity and atmospheric effects. They also established a nonlinear retrieval algorithm, with an accuracy that can reach 2–2.5 °C. Njoku and Li (1997) also used the satellite microwave radiometer data at the range of 6–18 GHz frequency to derive LST. A surface temperature accuracy of 2 °C was achievable, except for bare soils where discrimination between moisture and temperature variability is difficult using this algorithm. Aires et al. (2001) developed a new neural network and variant assimilation method, and the theoretical RMSE of LST retrieval over globe is 1.3 K in clear-sky conditions and 1.6 K in cloudy scenes.
However, the passive microwave retrieval algorithm of LST from AMSR-E BT is rarely seen in the present stage of application of passive microwave radiometry in China. Mao et al., 2007a, Mao et al., 2007b established a regression analysis model between the BT of the AMSR-E bands and MODIS LST products. The average retrieval LST error is about 2–3 °C relative to the MODIS LST products. He also found that the 89 GHz vertical polarization is the best single band to retrieve MODIS LST. However, over 60% of the areas in MODIS LST product are influenced by weather, especially cloud. The MODIS LST itself contains certain errors when the air contains much cloud, atmospheric water content or rainfall. So the regression model between AMSR-E BT and MODIS LST products lacks some practical significance. Our objective here is to establish a regression model between AMSR-E BT and observation temperatures (Ts) from meteorological observation stations over Guangdong Province and to describe a new, simple, yet still efficient algorithm to derive LST under bad weather conditions during the snow disaster of Southern China in 2008. Further more, the ground emissivity has a considerable impact on the accuracy of retrieved LST from remote sensing data (Rubio et al., 1997, Yang and Yang, 2006), and it is also influenced by land surface cover conditions, such as the density of vegetation cover and soil moisture levels. This paper aims to develop such a regression model based on different degrees of vegetation cover.
Section snippets
Study data and area
The AMSR-E instrument on the NASA Earth Observing System (EOS) Aqua satellite is a modified version of the AMSR instrument launched on the Japanese Advanced Earth Observing Satellite-II (ADEOS-II) in 1999. AMSR-E is a successor in technology to the Scanning Multi-channel Microwave Radiometer (SMMR) and Special Sensor Microwave Imager (SSM/1) instruments, first launched in 1978 and 1987, respectively. It provides global passive microwave measurements of terrestrial, oceanic, and atmospheric
The three LST models with MPDI intervals at 0.04, 0.02 and 0.01
The scatter diagrams were presented to compare the MDPI-based LST retrieved from AMSR-E BT with Ts observed by 86 meteorological observation stations. Average LST errors and RMSE by leave-one-out cross-validation were used to evaluate the retrieval results of the three models. Detailed descriptions of the three models are as following:
- (1)
Model 1: We classified the land surface into three types at the MPDI interval of 0.04, and built three linear regression algorithms for each land surface type in
Conclusions
Based on analysis of the passive microwave radiance transfer equation, and MPDI-based surface cover classification, we built a simple yet effective LST retrieval model (average error of LST: 0.91–1.30 °C) by combining AMSR-E BT with observation temperatures. This study is one of the rare applications that use the field temperature to make regression analysis with AMSR-E BT products during a strong cold disaster period of 2008 in Southern China. It is also a referential example for using AMSR-E
Acknowledgements
This study was supported by the science and technology plan fund grant of Guangdong Province, China (2007B020500002-7). Authors wish to thank two anonymous reviewers for their valuable comments and Prof. Chunlin Wang in Climate Cenetr of Guangdong Meteorological Bureau of China for provision of meteorological observation data.
References (14)
- et al.
A simple retrieval method for land surface temperature and fraction of water surface determination from satellite microwave brightness temperatures in sub-arctic areas
Remote Sensing of Environment
(2003) - et al.
Changes in land surface temperatures and NDVI values over Europe between 1982 and 1999
Remote Sensing of Environment
(2006) - et al.
Emissivity measurements of several soils and vegetation types in the 8–14 μm wave band: analysis of two field methods
Remote Sensing of Environment
(1997) - et al.
A modified land surface temperature split window retrieval algorithm and its applications over China
Global and Planetary Change
(2006) - et al.
A new neural network approach including first guess for retrieval of atmospheric water vapor, cloud liquid water path, surface temperature and emissivities over land from satellite microwave observations
Journal of Geophysical Research
(2001) - et al.
An evaluation of AMSR-E derived soil moisture over Australia
Remote Sensing of Environment
(2009) The contribution of AMSR-E 18.7 and 10.7 GHz measurements to improved boreal forest snow water equivalent retrievals
Remote Sensing of Environment
(2008)