Statistical analysis and predicting spatio-temporal variations of urban heat islands using remote sensing data

The urban heat island (UHI) as a climatic effect of urbanization can negatively impact the flora and fauna involved in urban and suburban ecosystems, the presence of pollutants, air quality, energy and water consumption, as well as human health and economy. Therefore, spatiotemporal analysis of the urban heat island changes has been considered an effective approach to understanding the impact of urbanization on the urban and suburban ecosystem, which also can support sustainable urban development and planning. Accordingly, this study contributes a novel approach to identifying the trend and predicting the pattern of UHI changes using statistical analysis, Shannon's entropy and chi-score statistics.

The study area of this research is the city of Rasht and its surroundings, a region located in the north of Iran. This research was implemented using remote sensing images from 1991 to 2021 that were collected by LANDSAT 5 and 8 with a fixed time interval of 10 years. All images were captured in the summer. In order to conduct this research in the pre-foresight stage, first, the required preprocessing, including atmospheric and radiometric corrections applied to the satellite images. Then, the surface biophysical characteristics of the study area were extracted from the satellite images. In the third step, the land surface temperature was computed using satellite images in 2021. In the fourth step, Multivariate linear regression between surface biophysical characteristics and the land surface temperature in 2021 was applied and then the cellular automata-Markov chain model was utilized to predict the land surface temperature for 2031. Finally, the pattern of changes in the urban heat island of Rasht city was investigated using statistical analysis in different geographic directions and different time periods.

The results of this study indicate that the highest positive correlation (R=0.89) was between NDBI and LST. Moreover, the highest negative correlation (R=-0.81) was between the greenness and LST. Our results also showed that the lowest correlation (R=0.42) was between the brightness and LST. The predicted LST corresponding to surface biophysical characteristics using a multivariate linear regression model illustrates the low error of this approach (RMSE=1.33K) in 2021. This means that the predicted values in 2021 are close to the real values, and therefore, this model can be trusted to predict LST in 2031. Statistical analysis of the patterns of observed and expected changes in UHI clearly illustrated that Rasht urban expansion and the UHI expansion will consistently continue to increase from 1991 to 2031. However, the expansion rate changes over time and space. Moreover, these analyses also showed that the UHI of Rasht city has a high degree of freedom and a high degree of sprawl. Thus, and as a result, its degree of goodness is negative.

The pattern of UHI changes is highly dependent on the pattern of built-up land changes: as a result, sustainable development, resilience and environmental protection of Rasht requires direct monitoring and control of the pattern of urban growth, such as preventing changes in built-up areas and agricultural lands in suburban areas by incorporating a vertical form of development as well as constructing green roofs and walls and using high-reflectance building materials.