Interplay And Forecasting Of The Urbanization Effect On Vegetation Loss And Heat Island Effect In Nine Large Cities In Pakistan

Rapid urbanization has excessive stress on natural resources and is one of the main and most powerful anthropogenic activities of humans.Protection of the environment and vegetation is an important global issue.Urbanization is a process to convert the rural lands into buildings,factories,and impervious surfaces,which lead into decrease in vegetation.Urbanization directly impacts the vegetation cover of its surroundings,and urbanization cause the rise in temperature which is the primary base of urban heat island effect.The urban heat island can have a large impact on building and urban design.Urban heat island in winter can reduce the energy consumption in heating system,while the heat island in summer would cause increasing energy consumption in air-conditioning.Therefore,it is important to investigate and quantify the urban heat island effect and impact of urbanization on vegetation.It is essential to develop a comprehensive understanding of plausible change in urbanization in different climatic zones of the world.It can lead us to strengthening of management of urbanization processes.Increased urbanization has impacted on vegetation cover（VC）in various regions.However,little reports on the impact of urbanization on vegetation and heat island impact has achieved in Pakistan,which is facing very rapid urbanization.In this study,the impact of urbanization on vegetation and urban heat island for nine large cities of Pakistan were investigated and forecasting.The Moderate Resolution Imaging Spectroradiometer（MODIS）enhanced vegetation index（EVI）data with 1000 m spatial resolution（MOD13A3）and MODIS land cover data with 500 m spatial resolution（MCD12Q1）were used to examine the spatial and temporal trends for urbanization effects on VC at 9 major cities Pakistan.By using land surface temperature（LST）data（MOD11A2）from MODIS,daytime,and nighttime surface Urban heat island intensity（SUHII）was calculated by the difference between urban temperature to rural temperature,and the temporal trends of daytime,nighttime SUHII and impact of urbanization on vegetation were tested by using Mann Kendall（MK）trend test and linear regression analyses.For future forecasting of the impact of urbanization on vegetation,the Recurrent Neural Network（RNN）algorithm with long short-term memory（LSTM）architecture was adopted,and autoregressive integrated moving average（ARIMA）model was applied for forecasting of SUHII effect.Pearson correlation was applied to check the relation of SUHII and impact of urbanization on vegetation.The main conclusions can be given:（1）The urbanization in Pakistan has significant impact on vegetation cover and SUHII.TheΔEVI（urban EVI-rural EVI）was utilized to represent the urbanization impact on VC in Pakistan.It was found that averaged urban area of all nine cities was increased 4.14%from 2005（306.675 km²）to 2019（319.375 km²）.The average annualΔEVI of nine cities is-0.04992/decade from 2005-2019,among which theΔEVI of all 9 cities are negative.Significant（p<0.05）decrease was observed inΔEVI in seven cities out of nine for the period of 2005 to2019.For all nine cities the annual urbanization’s area effect on VC for 2005 to 2019 was 447.06km².Insignificant trends were observed in seven cities of nine for annual average of urbanization’s area effects on VC.Significant decreasing（p<0.05）trends for springΔEVI were observed in all 9 cities,among which 7 cities showed significant decreasing trend for summer and autumn season and in winter season 8 cities exhibited significant decreasing trends.While the largest decreasing trend for averagedΔEVI is observed in autumn(-0.00224 yr^-1,p<0.01),the minimum trend is observed in summer(-0.00330 yr^-1,p<0.01).These results indicate that the impacts of urbanization on VC should be taken attention in sustainable development in Pakistan,which is also pronounced in other developing countries.（2）The temporal variations in diurnal,seasonal,annual scale of land surface temperature（LST）and surface urban heat island intensity（SUHII）of 6 cities of Punjab,Pakistan during 2006 to2020 was investigated by using MODIS data,and the future SUHII was forecasted using ARIMA model.Results suggested that the average mean SUHII for daytime was 2.221℃ and average mean nighttime SUHII was 2.82℃ during 2006 to 2020.The seasonally and annual average SUHII for daytime and nighttime exhibited increasing trends.Among which the spring average daytime SUHII showed maximum upward trend by 0.486℃ yr^-1（p<0.05）,and annual average nighttime SUHII showed maximum increasing rate of 0.485℃ yr^-1（p<0.05）.Additionally,a significant annual and seasonal（spring and summer）average SUHII increase was noted（p<0.01）.While the highest increasing trend in SUHII is found in spring(0.043℃ yr^-1,p<0.01),while the minimal increasing(0.005 yr^-1,p<0.05)was observed in winter.Furthermore,ΔEVI and SUHII are significantly negatively correlated with each other for all the seasons except winter,which has strongest correlation for spring（-0.894,p<0.01）.While the increasing SUHII and decreasing vegetation in all the cities call for appropriate actions,geospatial monitoring,and effective policies to assure urban comfortability and livability.（3）The RMSE of ARIMA model for daytime SUHII were 0.74（Lahore）,0.77（Faisalabad）,0.75（ISB/RWP）,0.82（Gujranwala）,1（Multan）and 0.91（Sialkot）,for nighttime SUHII RMSE values were 0.66（Lahore）,0.53（Faisalabad）,0.38（ISB/RWP）,0.66（Gujranwala）,0.64（Multan）and 0.68（Sialkot）for the validation data,which indicates that ARIMA can be utilized as reliable method for forecasting of SUHII.The seasonal ARIMA（SARIMA（0,0,1）（1,1,1）12）was applied to SUHII data for forecasting.The projection by ARIMA model suggested an increase of.04℃ in average daytime SUHII and 0.1℃ increase in average nighttime SUHII till 2030.（4）The long short-term memory recurrent neural network（LSTM-RNN）and artificial intelligence（AI）method was applied to forecast futureΔEVI.The forecasted results ofΔEVI which revealed thatΔEVI is decreasing in all cities.ΔEVI calculated from EVI data of MODIS for 6 different cities of Punjab,Pakistan.The root-mean-square error（RMSE）were recorded as 0.03,0.07,0.02,0.03,0.05,and 0.06 for city of Faisalabad,Gujranwala,ISB/RWP,Lahore,Multan,and Sialkot,respectively.The mean absolute percentage error（MAPE）were calculated as 0.12,0.55,0.24,0.18,0.28,and 0.47 for city Faisalabad,Gujranwala,ISB/RWP,Lahore,Multan,and Sialkot,respectively,which indicates that LSTM-RNN can be used as new reliable AI technique for forecasting.Results suggested that the average of projectedΔEVI for next ten years are-0.23,-0.21,-0.09,-0.13,-0.22 and-0.11 for Faisalabad,Gujranwala,ISB/RWP,Lahore,Multan,and Sialkot,respectively.