| As world urbanization accelerates greatly, land use/cover change (LUCC) has already become one of the major manners for human to change the earth and been the key driving force of environmental change. Urban impervious surfaces (IS), which are the major components of urbanized regions and can be utilized to quantify urban development and land use intensity, have been found to link urban land use and urban ecological environment change from a new perspective. Thus, it has given the best breakthrough point upon the effect of urbanization on eco-environment. Higher the impervious surface percent (ISP), higher heat transfer presented by the surface energy, which affects heavily the spatial heterogeneity of urban thermal environment. Herefore, the investigation of temporal-spatial change of ISP and its thermal environmental effect has drawn a great deal of attention from science community.At present, many scholars at home and abroad have investigated the temporal-spatial change of ISP, but most of them only described, compared and explained to the spatial distribution and difference. On the other hand, the relationship between ISP and land surface temperature (LSI) have been investigated in different regions and many similar contusions have been also made, but most of them were based on traditional statistical analysis method. For the temporal-spatial change of ISP and its thermal environmental effect, the rules in deeper level have not been understood well. In fact, the temporal-spatial change of ISP and its thermal environmental effect is complexity problem, it is necessary to reveal its rules using theory and methods of complexity science. But few reasearches have been done under the theoretical framework of complexity science. Up to date, fractal theory have been used in the study of urban IS, but most of them have emphasized on morphological fractal or one-dimension fractal based on section lines, lacking of researches of two-dimension fractal on urban ISP spatial distribution. On the other hand, there is lack of deep understanding on the quantitive relationship between ISP and LST from long range cross correlations of two variables by fractal method.Based on the analysis of the existing problems, this paper take shanghai as a study region where urbanization have rapidly developed and land use/coverage have changed greatly. Under the theory framework of complexity science, combining with the theory and method of landscape ecology and geography, this paper identifies the spatial distribution law of ISP and its relationship with LST within Outer Ring Road in Shanghai city by fractal method, hopping to advance the complexity study of ISP temporal-spatial change and its thermal environment effect in theory, and to provide new research perspective and ideas to the study of urban LUCC and its environment effect, and also to support the decisions for urban planning, environment protecting in practice.The main findings are as follows:(1) The spatial distribution of ISP from1997-2010was obtained by linear spectral unmixing method based on vegetation-impervious-surface-soil model principle. The IS area and average ISP value for the study area appeared to increase with fluctuations from1997to2010. The high value of ISP distribute mainly in old city and along Huangpu River in1997, and it expended to Outer Ring Road in2002and2007, however, its aggregation degree decreases in2010. The ISP increased for most pixels in1997-2007, yet it showed decreased for56.17%pixels in2007-2010. The spatial distribution of the data of maximum ISP revealed clearly the urbanization process in the study area. Based on the theory and method of landscape ecology, the IS has been divided into seven types, among which, full cover type (ISP greater than80%) accounted for the largest area, and its area was increased form1997-2010, which led to the decreased of landscape diversity index for IS landscape. During the period of14years, the fragmentation degree of full cover type increased, however, the aggregation index change little, and the patch morphology is regular, implying the influence of urban planning on patch of full cover type.(2) For the ISP data along the horizontal and vertical axes across People’s square from1997-2010, the spatial distribution of ISP all showed scale invariance. The one-dimension box-counting dimension indicated the change of data intensity, and revealed how the ISP data distributed along the axes. On the horizontal axis, the ISP series in1997and2002exhibited multifractality property, however, data series only in1997presented multifractality on the vertical axis, suggesting the more and more uniformed for ISP spatial disturbution along the axes from1997-2010. For two-dimensional images of ISP from1997-2010, the two-dimension box-counting plots illustrated the meaningful fractal power-law or non-power-law behavior in the ISP spatial distribution. The two-dimension box-counting dimensions revealed the complexity of ISP spatial distribution, and with the increasing of thresholds, the complexity decresed. According to the same way, the spatial distribution of all IS types also exhibited spatial self-similar characteristics. Two-dimension fractal dimensions revealed the spatial complexity of ISP spatial distribution.(3) Based on Moran’s I index, the spatial autocorrelation was found in the spatial distribution of ISP during the study period, and the autocorrelation presented the aggregated-dispersed-aggregated process from1997-2010. In addition, by the two-dimension detrended fluctuation analysis (DFA), the long-range dependent nature in the ISP spatial distribution have been presented and the DFA scaling exponents corresponded to1/f noise, suggesting the same change characteristic in ISP spatial distribution in different spatial scales, that is to say that with the increase of ISP for some pixel, the ISP of pixels in some spatial scale would increase, vice versa. In an effort to explore the multifractal characteristics in the ISP spatial distribution by multifractal detrended fluctuation analysis (MFDFA), the self-similarity was observed, however, multifractality was not found. The long range autocorrelation indicated that the spatial distribution of ISP had the features of fractal Brownian movemen. Larger the DFA scaling exponents, stronger the longe range autocorrelations, and slower the fluctuation among the ISP data, and more uniformed the spatial distribution of ISP data. Therefore, DFA scaling exponents could reflect the equilibrium of ISP spatial distribution.(4) Urban LST in1997-2010was estimated from the Landsat TM/ETM+thermal band. For the spatial distribution of urban LST, the standard deviations decreased throughout the study period. From1997-2010, the scope of urban thermal distribution extended gradually to the outer ring road, and the intensity enhanced, but urban heat island was less pronounced. In addition, by two-dimension box-counting dimensions, the scale invariance of LST spatial distribution was illustrated. Moreover, the self-similar and fractal long-range dependent nature was observed for LST spatial distribution in1997-2010, where the DFA scaling exponents were close1/f noise, but there was no multifractality in the long-range dependence. These fractal features were all similar with that of ISP spatial distribution, which confirmed further the important influence of IS spatial distribution on LST from a new perspective.(5) Relationship between LST and ISP was investigated. By scatter plot, a striking linear and positive correlation has been found in1997. However, the non-linear relationships were more and more apparent from2002-2010. Based on seven IS types, a major change trajectories to full cover type have been obtained, which dominated the LST change in the whole study area. Moreover, the percent composition and configuration of IS types were measured by a series of landscape metrics, and their effects on LST were calculated. The results suggested that the configuration of IS types significantly affected the magnitude of LST, but the composition of IS types was more important in determining LST than their configuration. Therefore, the urban thermal environment effect could be mitigated not only by balancing the relative amounts of IS types, but also by optimizing their spatial configuration. (6) For the horizontal and vertical axes across People’s square, the long-range cross-correlations between ISP and LST data were quantified and analyzed by the detrended cross-correlation analysis method (DCCA) and multifractal detrended cross-correlation analysis (MFDCCA) method. On the vertical axis, a positive long-range cross correlation was observed in1997, but the cross correlation was not power-law. However, positive and power-law long-range cross correlations were observed in2002-2010. Further, based on the cross-correlate scaling exponents, the long-range cross-correlations exhibited decreased in1997-2010. In addition, multifractal long-range cross correlations were found in2007and2010, and the multifractality changed little in2007-2010. On the horizontal axis, two scale-invariant regions were presented, where the positively long-range cross correlations were showed. In the first scale-invariant region, the long-range cross correlations were power-law in1997and2002, and in the second scale-invariant region, power-law long-range cross correlations were found in2007and2010. The long-range cross-correlations showed increased from1997-2010in the smaller scale-invariant region, however, it is decreased in the larger scale-invariant region. Multifractal long-range cross correlations were also found in2002-2010, and the multifractality was decreasing from2002-2010. The multifractality was stronger on the vertical axis than that on the horizontal axis. |
