Remote Sensing Detection Methods And Driving Factors For Spatiotemporal Dynamic Change Of Urban Impervious Surface And Fractional Vegetation Cover

The rapid urbanization process has promoted continuously expansion of urban land,resulting in an increase of Urban Impervious Surface(UIS),a loss of vegetation cover,a change of Fractional Vegetation Cover(FVC),bringing about a series of urban environmental and ecological problems.A timely and accurate grasp of the spatio-temporal dynamics in UIS and FVC is of great practical significance to the sustainable development of cities and the construction of an ecological,intelligent and harmonious urban community of destiny.This paper developed an accurate urban index model and a vegetation index model,respectively.The influence of seasonal transition and spatial location shifts on surface reflectance of Landsat remote sensing images(LRSI)was taken into account in the two models.An improved particle swarm optimization(PSO)algorithm was used to select the index threshold automatically.By examining the time series values of these indices over time,determination of land cover change types for each pixel was made to map UIS and vegetation dynamics from the long time series LRSI,to give a clear view of the spatio-temporal patterns of UIS and vegetation during the period 1999 to 2018 over the study area.A prototype application was illustrated by Hefei,the core city of Wanjiang city belt.The main contents and conclusions of this paper are as follows:(1)A more accurate UIS index model(UISI)for urban impervious surface extraction was proposed.Automatic selection of the index image threshold to overcome the influence of spatiotemporal changes on the accurate extraction of UIS from Landsat long-time series observations was made,at the same time,the time series index curves of a pixel was exmined to determine the change type of the pixel,thus automated mapping of the spatiotemporal patterns of UIS was realized.Further,a comprehensive validation and performance evaluation framework was constructed,followed by an application of the above-mentioned methodologies to the downtown areas of Beijing,Hefei and Guangzhou.The relative accuracy of UIS extraction from the proposed model in Beijing,Hefei and Guangzhou was 92.43%,93.58%and92.61%,respectively.The Interclass Correlation Coefficient(ICC)of the segmented UIS pixel clusters was 0.845,0.876 and 0.865 in the three cities,respectively.And,the relative average accuracy of UISI index was higher than that of the commonly used UIS indices by 5.76%(Beijing),6.87%(Hefei),and 6.01%(Guangzhou)respectively.In the prototype case of Hefei,from 1999 to 2018,UIS continued to increase,with an increased total area of 1802.43 km~2,at an average annual growth rate of 15.95%.The early growth rate was generally high,giving way to a decelerated and fluctuated pace in the late stage,and the highest growth rate of 33.13%was observed in 2002 and the lowest of 4.75%in 2017.The growth rate of UIS development intensity was 15.75%.The dominant direction of UIS expansion was towards the southwest and northeast,with the downtown of Hefei as the center,while the other six directions had a similar lower developing pace simultaneously.The UIS spatial change in the downtown of Hefei in the past 20 years was divided into four main stages according to speed incresase:(i)the period of fast growth from 1999 to 2003,the average annual growth rate was at 23.38%.(ii)The period of fluctuated fast-growth from 2004 to 2008.The annual growth rate was unstable,with an average annual UIS growth rate of 18.58%.(iii)The period of fluctuated and decelerated growth from 2009 to 2013,the average annual UIS growth rate was 6.95%and the highest growth rate was 12.1%,dropped by 11.64%in contrast to that of 2004-2008,but the average annual UIS growth area was 24.19 km~2.(iv)During the period between 2013 and 2018,there was a stable and small magnitude of growth,with an average annual growth rate of 5.71%and an annual standard deviation of 1.5%.Over the past20 years,the total UIS in the downtown of Hefei increased by 504.88 km~2.(2)An Optimized Dynamic Range Vegetation Index(ODRVI)was proposed to improve the estimation of FVC.Similarly,the Landsat images with spatiotemporal differences were used to validate the ODRVI.Automated mapping and spatiotemporal dynamic detection of FVC were realized.The relative accuracy of the ODRVI in Beijing,Hefei and Guangzhou was93.35%,96.79%and 95.88%,respectively.Compared to the typical commonly used vegetation indices,in the three cites,the accuracy of ODRVI was higher than the typical vegetation indices by 7.24%,7.40%and 8.55%.The standard deviation of accuracy the ODRVI was lower than those of other commonly used indices by 3.72%.The fitting accuracy between the ODRVI index and FVC was higher than those of other commonly used indices,with R~2 increasing by6.14%.The ODRVI and corresponding improved FVC estimations were used for the spatiotemporal dynamic analysis of FVC over the past 20 years in Hefei.Over the past 20 years,the proportion of vegetation coverage decreased by 14.76%in Hefei.The vegetation coverage area and FVC levels generally showed a fluctuated decreasing trend,and the spatiotemporal variations were divided into four stages according to FVC area:(i)continuous rapid decline period(1999-2004),the annual decline rate was 1.76%;(ii)A rapid fluctuation decline period(2005-2008),the average annual decrease was 776.65 km~2;(iii)A low level of decline period(2009-2013),the average annual decrease was 382.90 km~2,and(iv)A fluctuated increasing period(2014-2018),The annual increase was 258.92 km~2.The changes in FVC grades mainly focused on very low-level FVC and medium-level FVC.During the 20 years,the area covered by high-level FVC increased by 17.65 km~2.The change rates of FVC are as follows:medium-level FVC>very-low-level FVC>low-level FVC>medium-high-level FVC>high-level FVC.(3)The driving factors analysis of UIS and FVC were as follows:the most important economic,social and ecological driving factors on UIS were the gross national product,built-up area and built-up greening coverage,respectively.The goodness-of-fit of the three types of driving factors was 96.50%,95.79%and 88.97%,and the Mann-Kendall statistics were 6.132,5.613 and 4.250,respectively.The three types of driving factors were in an upgoing trend over time(1999-2018).The mean temperature in April-October and annual mean FVC over the past 20 years tended to be well fitted.The fitting of average rainfall and sunshine duration with FVC variation showed great fluctuations.The rainfall was generally higher in the west than in the east,and higher in the south than in the north.Extreme meteorological conditions had great influence on very-low-level FVC and low-level FVC.In the southern mountains,more rainfall,FVC was higher than that in the north,west and east of Hefei downtown.Hills and mountains in the east,west and south had high-level FVC.FVC was negatively correlated with UIS.The correlation between UIS and FVC was higher than that between meteorological factors in the downtown of Hefei.The influential importance of UIS on the total area of FVC and FVC of each grade was ranked as follows:FVC total area>low-level FVC>very-low-level FVC>medium-high-level FVC>high-level FVC.