Forest Phenology Dynamics Derived From MOIDS Time-series Data In The Source Region Of The South-to-North Water Diversion Project

Global change is becoming a hot topic recently. Land is the most closely intermediate between the processes of natural change and human activities, and the change of land use/cover is the main reason of global climate change. Therefore, obtain landuse/cover and its change information rapidly is of great significance to reveal the human-land relationship and relieve the trend of global warming. Remote sensing, a quantitative target detection technology, can effectively reveal the characteristics and change information of objects by using the electromagnetic spectrum information, which has the advantages of high speed, high efficiency, low cost and so on.Phenology is one of the most sensitive ecological indicators of vegetation response to climate change. The combination of remote sensing technology and vegetation phenology can effectively improve the accuracy of the ground observation, thus deepening the understanding of the law of land use/cover chang. This paper use time-series vegetation index based on MODIS, extracted phenology index and use it as an effective basis for classification and studying on the phenology of vegetation. Specificly time-sequence data of enhanced vegetation index(EVI) of 342 images were used to extract phenometrics of startseason, endseason, lengthseason, max of EVI, basevalue of EVI and amplitude. Combined these phenometrics with some auxiliary data to have a classification of the water source region and have a validation for the the result accuracy. Forestland phenology spatial pattern and the changes of 13 years was studied based on the classification result, as well the distrbution and change rule with the altitude gradient.The result show that:1. It is an effective way to extract phenometrics based on EVI time-series for land use classification, the method is reliable. In this paper, we obtained a good classification result based on the 6 phenometrics.2. The overall accuracy of each classification method based on MODIS in the water source area indicate that decision tree classification of cart with 9 bands keeps stable with highest accuracy.3. The results of land use/cover classification based on time series EVI showed that water source region is mainly forestland. The proportion of forestland area 52.5%, shrubland and grassland 6.4%,38.7% of agriculturalland, buildingland 0.7% and water and wetlands 1.7%, the other 0.1%. The distribution of different features in different altitude gradient is that:low altitude with agricultureland and forest dominnated, high altitude forest and grassland dominated. With the slope incresing agricultureland, buildingland and waterregion reduced, woodland and grassland distribution in different slope have no obvious difference. The overall accuracy is about 75%.4. Startseason of forestland within water source region is concentrated in 100th days-150th days with distribution that the east is earlier than west, while startseason is later with increase of altitude; Endseason of forestland is concentrated in 270th days-310th days, and endseason is earlier with increase of altitude; The lengthseason is about 120days-200days, and the distribution pattern is the same as startseason, with the increase of the altitude the lengthseason is shorter.5. the area of start season advanced、no change and delayed of phenology change with the linear regression accounted for 28.91%,49.82% and 21.27%, and the area of endseason advanced-. no change and delayed accounted for 20.49%,48.94%,30.57%, while the area of lengthseason shorten, basically unchange, extended is 27.02%、33% and 39.98%, which show that startseason in some areas are delayed distributed in Danjiangkou Reservoir Area, Danjaing River Basin and so on, endseason of advanced, delayed and keep unchanged all exist with a not obvious distrbution pattern. The pattern of lengthseason is the same as startseason with the corresponding of a advanced region to shorten region. The distrbution of phenology change obtained from subtraction method is consistent with the linear regression method, but the degree of change is more severe in advance and dalay. All in all, startseason is slightly delayed and lengthseason is lightly shorten at an altitude of 800m below, and above 800m startseason has a extremely weak advance,lengthseason has a very weak extension, the variation is not significant, the chang pattern of endseason is not obvious at all lattitude gradients.