Monitoring The Riparian Vegetation Cover After5.12Wenchuan Earthquake Along Minjiang River Valley Based On Unmixing Of Remote Sensing Imagery Data: A Case Study Of Yingxiu-Wenchuan Section

Disturbances, especially large and infrequent disturbances are usually understoodas events that disrupt the equilibrium state of an ecosystem by producing abruptchanges in its structural and functional characteristics that may persist for decades tocenturies. Earthquake, as a typical catastrophic disturbance event, plays an importantrole in several ecological processes by partially or completely removing thevegetation layer. The Wenchuan Earthquake (Richter scale8) on May12,2008insouthwestern China caused widespread ecosystem damages in Longmenshan area. Itis important to evaluate the natural vegetation recovery process and provide basicinformation on ecological aspects of the recovering environment after the earthquake,especially in the riparian zone which is a hot area in biodiversity conservation and ineco-hydrological function.In this paper, a case study of the vegetation at riparian zone of theYingxiu-Wenchuan section along Mingjiang river was selected. To circumventweather limits of remote sensing in the Wenchuan earthquake-hit areas and to meetthe need for regional observation analyses, three Landsat TM images pre-andpost-earthquake in the Minjiang arid valley were used for analysis. To improve theprecision, the LSMM（Linear spectral mixing model）is used to unmix the pixels in thestudy area. Then, the post-earthquake vegetation cover values were compared to thepre-earthquake value to determine the extent to which the vegetation was damaged inrelation to the pre-earthquake pattern, and the rate of recovery was evaluated. Spatialcharacteristics of vegetation loss and natural recovery pattern were analyzed in relation to elevation, slope and aspect. The main results are as follow:(1)The cumulative percent of the pixels DN<0.05in RMSE images of2007,2008and2011is99%,99%and95%, respectively, and the average of the RMSE images isless than0.025, which shows that the TM image and LSMM are feasible to unmix themixing pixels of the study’s images.(2)There is good relationships between NDVI and the vegetation fractionextracted by LSMM, with the decisive coefficients greater than0.8(P<0.001). Theresult indicated that the vegetation fraction is available to estimate vegetation cover inthe study area.(3) The average vegetation cover decreased13.95%in2007-2008, andrecovered7.92%in2008-2011under the natural condition. The severely damagedsites occurred mainly within the range of1100-1700m elevation and on slopes of25-55o. After three year of natural regeneration,56.20%of the destroyed areas wererecovered.