Evaluation of estimator variables in air temperature estimation in January and June based on land cover

The near-surface temperature, T_s measured by ground stations provides limited information on the spatial distribution of T_a pattern. A correct estimation of T_a distribution pattern is necessary for a wide range of applications such as hydrology, ecology, meteorology (Wenbin et al., 2013) and biology of vector-borne diseases. In this study, near-surface air temperatures (T_a) using environmental parameters including land surface temperature (LST), altitude, slope, vegetation, latitude, albedo, and mean sea level pressure (MSLP), were estimated for January and July in the period 2001-2015 for Iran. In this study, due to the use of different data sources with different spatial resolutions, all maps were converted to the same spatial resolution of Era-Interim (0.125˚). Then spatial distributions of T_a and LST were determined. The spatial distribution patterns of these two components were also determined by applying the Moran spatial autocorrelation index. Finally, according to the land cover, multivariate regression models are presented for estimating T_a based on seven parameters, including LST, altitude, slope, vegetation and latitude, albedo and MSLP. In the following, the characteristics of each of these data are also described. Standardized regression coefficients were used to determine the most important estimator in each land cover. The correlation between the parameters involved in the study with the absolute difference between the air and surface temperature are negative in January, which means that by increasing, slope, altitude, NDVI, latitude, albedo and MSLP, the difference is reduced and vice versa. Nonetheless, this kind of relationship is not valid in the whole study area, and there are some exceptions. In July the relationship between this difference and slope and NDVI is positive, which means that with increasing altitude, latitude, albedo and MSLP, the differences also increase. In January, waters (99%), urban areas (95%), and barren or sparsely vegetated (92%) have the highest R². While, mixed forests had the lowest R² equal to 27% (Figure 4). The least errors are related to urban areas (0.69 ° C), water (0.75 ° C), and then forest areas (0.9 ° C). The highest errors were observed in cropland and open shrubland equal to 1.35˚C and 1.34˚C. The highest R² was calculated for water (95%), urban areas (94%), mixed forest and open shrubland (93%). The least error occurred in mixed forest (0.3˚C). The main objective of the present study was to develop a model of air temperature estimation from surface temperature and other auxiliary variables (elevation, slope, vegetation, latitude, land cover, albedo and mean sea level pressure). Regression models were presented for estimating T_a in monthly scale. The results can be summarized as: Between the air and surface temperature, the most variability is related to the T_a which in the region of Iran has an annual variation coefficient of 92% in January and 41% in July. In January, slope and altitude are the most important variables in the estimation model so that up to 16% and 12% can explain LST-T_a differences, respectively, while latitude and MSLP are the most important variables in July so that each one of them explains up to 9.6% of these differences in July. The role of land cover in estimating T_a is very important. In addition, the number of pixels located on each land cover category can also play a decisive role in estimation model. Category of water, urban and barren area in January, exhibited the highest R² of 99%, 95% and 92%, respectively. The lowest R² (approximately 27%) is related to grassland and mixed forest. In July, the highest R² is related to water and urban areas about 95 and 94%. R² of grassland increases by approaching summer. The lowest error is recorded for urban area, water and mixed forest in January while the lowest error is related to mixed forest, open shrubland and barren areas in July. The accuracy of estimation models varies according to the months and the land cover. Based on standardized regression coefficients, in January altitude (in barren, urban and cropland area) mean sea level pressure (in grassland and shrubland), slope of mixed forest area and latitude in water area were of great importance in air temperature estimation. While, in June, due to presence of low pressure unter in Iran, the role of local climatic factors has been minimized and mean sea level pressures was the most important estimator almost in all landcovers.