|Institution:||Indiana State University|
|Keywords:||Multi-timescale; LST; trend; seasonality; decomposition; urbanization|
|Full text PDF:||http://hdl.handle.net/10484/5569|
Spatial and temporal patterns of land surface temperature (LST) have been used in studies of surface energy balance, landscape thermal patterns and water management. An effective way to investigate the landscape thermal dynamics is to utilize the Landsat legacy and consistent records of the thermal state of earth’s surface since 1982. However, only a small proportion of studies emphasize the importance of historical Landsat TIR data for investigating the relationship between the urbanization process and surface thermal properties. This occurred due to the lack of standardized LST product from Landsat and the unevenly distributed remote sensing datasets caused by poor atmospheric effects and/or clouds. Despite the characterization of annual temperature cycles using remote sensing data in previous studies, yet the statistical evidence to confirm the existence of the annual temperature cycle is still lacking. The objectives of the research are to provide statistical evidence for the existence of the annual temperature cycle and to develop decomposition technique to explore the impact of urbanization on surface thermal property changes. The study area is located in Los Angeles County, the corresponding remotely sensed TIR data from Landsat TM over a decadal year (2000-2010) was selected, and eventually a series of 82 cloud-free images were acquired for the computation of LST. The hypothesis technique, Lomb-Scargle periodogram analysis was proposed to confirm whether decadal years’s LSTs showed the annual temperature cycle. Furthermore, the simulated LSTs comprised of seasonality, trend, and noise components are generated to test the robustness of the decomposition scheme. The periodogram analysis revealed that the annual temperature cycle was confirmed statistically with p-value less than 0.01 and the identified periodic time at 362 days. The sensitivity analysis based on the simulated LSTs suggested that the decomposition technique was very robustness and able to retrieve the seasonality and trend components with errors up to 0.6 K. The application of the decomposition technique into the real 82 remote sensing data decomposed the original LSTs into seasonality, trend, and noise components. Estimated seasonality component by land cover showed an agreement with previous studies in Weng & Fu (2014). The derived trend component revealed that the impact of urbanization on land surface temperature ranged from 0.2 K to 0.8 K based on the comparison between urban and non-urban land covers. Further applications of the proposed Lomb-Scargle technique and the developed decomposition technique can also be directed to data from other satellite sensors.