Effect of land cover change and elevation on decadal trend of land surface temperature: a linear model with sum contrast analysis

Land surface temperature (LST) is a significant factor in surface energy balance and global climatology studies. Land cover (LC) and elevation are two factors that affect the change of LST, and their effects depend on different geography. This study aims to demonstrate an alternative approach to examine the change of LST during 20 years (2001 to 2020) on Taiwan Island and to investigate the effect of LC change and elevation on a decadal trend of LST using a linear model that adjusting for each determinate factor. MODIS LST and LC data, as well as GMTED2010 elevation product, were downloaded available website. The natural cubic spline function was used to model annual seasonal patterns in LST. Linear regression model was used to estimate decadal change of long-term LST time series. Weighted sum contrasts linear regression was used to assess the effect of LC transformation and elevation on the decadal LST change by comparing adjusting mean of all factors. The adopted analysis method was an appropriate approach to assess categorical factors than those based on treatment contrasts, requiring specifying a control group to compare means and confidence intervals. Results showed that there was an increase in LST for most of the island. The average daytime and nighttime LST trends were 0.12 and 0.31°C/decade, respectively. However, areas in the southern part of the north-south direction mountain range show a statistically significant increase in LST in both daytime and nighttime. The major landslides caused this noticeable change of surface temperature due to the catastrophic damage of typhoon Morakot in 2009. The results also revealed that the different pattern of LC change has a significant effect on daytime LST, but not on nighttime LST trends. The elevation above 600 m had affected both daytime and nighttime LSTs.