Analysis Of Agricultural Landscape Heterogeneity At Multi-scale In Gongyi City

Landscape patterns are formed by various interacting biological and non-biological process,landscape heterogeneity is the important property of the landscape pattern. This paper focused on analyzingthe different spatial scales of landscape heterogeneity. Firstly, choosing four different human disturbancedegree of urban area, farmland, hills and mountain landscape a4km×4km sample area respectively, thenusing moving window algorithm and counting the contagion index and the evenness index at landscapelevel from the moving window radius of100m,250m,500m,1000m,1250m and1500m. Relying on thecontagion index indicates the characteristics of landscape composition and the evenness index means thefeature of landscape configuration. Secondly, according to the different of elevation divides into the riverplains, hills, and mountains, then selecting10sampling points in landscape partition respectively. Set up100m,250m,500m,750m,1000m,1250m and1500m as the radius of the circular buffer. Through thevariation characteristics of the21landscape index in different buffer radius, judging1250m as thecharacteristic scale. At the characteristic scale, we sift all landscape index and remain nine landscape indexat P<0.01correlation level, which indicates PD, AREA_AM, FRAC_AM, CONT, AI, IJI, PR, SHDI,COHESION. Finally, base on the change of NDVI value as the characteristic of landscape heterogeneity,then analysis of the spatial resolution influencing landscape heterogeneity by using semivariance functioncurve.The results prove that：1) As moving window radius increases gradually, the maximum of contagion index and evennessindex decline in the four landscapes, however, the minimum of contagion index and evenness index presentthe opposite trend expect for urban area. The mean value of contagion index in the urban area and farmland is bigger than it in the hills and mountain landscape, but the mean value of evenness index has the oppositecharacteristics. The value of contagion index and evenness index has significant correlation in the urbanarea, farmland, hills and mountain landscape, the value of R2is greater than0.9, expect for urban area at the1000m and above moving window radius. The slope of the fitted curve presents increasing trend with theincrease of moving window radius gradually. When the radius of the moving window increase gradually，there are large numbers of various patches in the landscapes bringing about the whole change of landscapecomposition and configuration, that would have an effect on the biodiversity, ecosystem stability andecological system service. From the landscape heterogeneity level, the more complex landscapecomposition and configuration, the higher level of landscape heterogeneity, that means the numbers and thetypes of patches are different, therefore they have a lot of composition form resulting from the formation oflandscape heterogeneity. From the disturbance intensity, the heterogeneity level of hills and mountain ishigher than urban area and farmland; it proves that the human disturbance factors play an important role onthe formation of landscape heterogeneity. Form the landscape of urban area, farmland, hills and mountainsin different moving window radius contagion index and evenness index of scatter plot; it reflects that thecontagion index and evenness index have strong correlation each other. The fitted curve reflects thelandscape composition and configuration along the change of gradient affected by the disturbance intensity.The four landscapes bear the influence of different disturbance intensity, and this kind of influence hasscale dependence.2）From PD, AREA_AM and FRAC_AM which represent the patches complex of the landscape,patches of hills are more richness than the other two regions, and the dominant area proportion of patchesare remarkable in the mountains especially. In terms of the shape of the patches, all patches’ shape isnormal in the region. From AI, CONT and IJI which mean the landscape configuration, there are a lot ofIV different patches aggregating in the hills, but different patches distribute evenly in the mountain and theyhave more chance to adjacent to each other. From SHDI and PR which stand for landscape diversity, therehave the largest numbers and types of patches in the river plains landscape. From COHESION whichindicates landscape adhesiveness, the core patches have the absolute advantage in the mountain.3）On the basis of the bilinear resampling, extracting all kinds of landscape types of NDVI valueat30m,60m and90m spatial resolution respectively, from the box-plot of NDVI at different spatialresolution along with the change of spatial resolution, it shows that the same kind of landscape of NDVIhas a little change along with the change of spatial resolution, and the influence of spatial resolution isnot of significance. According to the characteristics of semivariance function curve, the semivariancefunction curve of farmlands, canals and pits, buildings show a little level volatility, that means the structureand elements of landscape are simple. There are many kinds of patches at different spatial scales in theforest landscape and these patches arrange complicatedly, as a result, the forest landscape has the strongheterogeneity. The heterogeneity of unused landscape along with the spatial autocorrelation increasesgradually.