طبقه‌بندي و پيش‌بيني تغييرات مكاني-زماني سطوح نفوذ ناپذير شهري و اثرات آن بر شدت جزيره حرارتي

Classification and prediction of spatio-temporal Change of impervious urban surfaces and its impacts on urban heat intensity

پديده جزاير حرارتي به ‌عنوان يكي از مخاطرات، فعاليت‌ها و زندگي انسان در محيط ‌هاي شهري را تحت تأثير قرار مي‌دهد. سطوح نفوذناپذير شهري يكي از عوامل مهم در تغييرات جزيره حرارتي است. تصاوير سنجش ‌از دور روشي ارزان، كارآمد و سريع در بررسي شدت جزاير حرارتي و تغييرات سطوح نفوذناپذير در محيط‌ هاي شهري محسوب مي­ شود. لذا هدف از اين تحقيق بررسي و ارتباط بين سطوح نفوذناپذير و تغييرات شدت جزاير حرارتي است. منطقه مورد مطالعه در اين پژوهش شهر رشت است و از سري زماني تصاوير لندست مربوط به سال 1989 تا سال 2018 استفاده شده است. روش پژوهش بدين صورت است كه ابتدا پيش‌پردازش اوليه بر روي تصاوير انجام‌گرفته و سپس با استفاده از شاخص NDISI به طبقه‌بندي سطوح نفوذناپذير شهري پرداخته‌ شده است. براي تعيين حد آستانه تفكيك سطوح نفوذناپذير (اراضي ساخته‌شده) از سطوح نفوذپذير (اراضي ساخته نشده)، از روش آستانه گذاري Otsu استفاده‌ شده است. دقت طبقه‌بندي با استفاده از نقاطي كه به‌صورت تصادفي انتخاب‌ شده بود، مورد ارزيابي قرار گرفت. در اين تحقيق از مدل CA- Markov براي پيش‌بيني تغييرات آتي سطوح نفوذناپذير شهري استفاده‌شده است و درنهايت ارتباط بين سطوح نفوذناپذير شهري و تغييرات شدت جزيره حرارتي موردبررسي قرارگرفته است. نتايج اين پژوهش حاكي از دقت كلي 5/84 تا 90 درصد براي روش NDISI بوده است. اختلاف نقشه پيش‌بيني CA- Markov با نقشه واقعيت كمتر از 8 درصد بوده و نشان از قابل‌اعتماد بودن اين مدل است. ارتباط بين سطوح نفوذناپذير و جزاير حرارتي حاكي از همبستگي مثبت و قوي بين 69/0 تا 89/0 براي سال‌هاي مختلف بوده است. جهت تغييرات سطوح نفوذناپذير شهري و تغييرات شدت جزيره حرارتي با يكديگر منطبق بوده است.

Introduction The urban heat islands are hazardous to the health of urban residence, their activities, lifestyle and the quality of their life. This phenomenon occurs, in particular, as a result of the urbanization process, land use/cover changes and the rate of impervious surface coverage. Since early 1970s, the urban heat islands have been studied using remotely sensed data; because this approach is cheaper, more efficient and faster than traditional techniques to detect the heat islands as well as to examine the severity of them. However, less attention has been paid on the relationship between urban heat island (UHI) and impervious surface patterns. Therefore, this work aims to study UHI based on the analysis of land-surface temperature (LST) and impervious surface patterns (ISP) retrieved from remote sensing data covering a 29-year period. Materials and methods In this research, the city of Rasht as the center of Gilan province, Iran, is taken as the study area. Rasht is the largest city in the South Cost of Caspian Sea. In order to study the relationship between UHI and both LST and ISP, the time series of Landsat-5 / Thematic Mapper (TM) sensor, Landsat-7 / Enhanced Thematic Mapper Plus (ETM+) sensor and Landsat-8 / Operational Land Imager (OLI) sensor as well as Thermal Infrared Sensors (TIRS) of Landsat-8 from 1989 to 2018 have been utilized. Then preprocessing of satellite images including geometric correction and image referencing, radiometric corrections, and atmospheric corrections were applied on the images prior to other image processing steps. Then, by using the Normalized Difference Impervious Surface Index (NDISI), the impervious urban surfaces classified. The Otsu thresholding method was employed to determine a threshold value for the separation between impenetrable surfaces (constructed) and permeable surfaces (not constructed) in each utilized image. The classification accuracy was evaluated considering 300 randomly selected points. After mapping land use change over the years from 1989 to 2018, the future land use changes in the impenetrable urban areas were simulated to the year 2036 using CA-Markov model. Finally, the relationship between urban impermeable coverage and thermal island intensity changes were studied. Results and discussion The results of this study showed an overall accuracy of 84.5 percent to 90 percent for impervious surface classification using the NDISI method and the Otsu threshold. The results of the CA-Markov's model also indicate overall accuracy of 83.6 percent for impervious surface prediction. The difference in CA-Markov's prediction map with a reality map was less than 8 percent; hence, CA-Markov can be considered as reliable method in predicting land-use change in Rasht. Moreover, the obvious peaks and valleys values can be seen in the histogram of NDISI index; therefore, the determined threshold has well been able to classify the impervious surface. The spatio-temporal change of impervious surface showed an increasing trend, more than double over the city, from 1989 to 2018. Moreover, the prediction results of CA-Markov model indicates that the impervious area would double again within the next 18 years. The highest levels of urban impervious are located at an average distance of 5,000 meters from the city center, which has had an important impact on the thermal island's severity. The relationship between impervious surface and thermal islands showed a positive and strong correlation coefficient between 0.69 and 0.89 for various years. Furthermore, the pattern of urban impervious surface growth and thermal island intensity changes coincided with each other. The spatio-temporal change of UHI showed that the spatial extent of heat islands in Rasht was increasing with time and the temporal trend of UHI was also increased. Moreover, the trend of heat island changes illustrated that area of regions with very low and low temperature were decreased. On the other hand, the coverage of regions with medium, high and very high temperature were increased. Conclusion The time series of Landsat images along with spectral indices are the convenient dataset to classify the impervious surface of the city with proper accuracy. The produced land cover map can also be employed as a proper input data for prediction models. The spatio-temporal analysis of urban heat island in Rasht illustrated that the urban heat intensity was increased. This trend was because of increasing rate of impervious urban surface. Ultimately, in order to control the heat island, it is required to prevent the unplanned urban construction and developments.