Forest Disturbance Detection By Remote Sensing Based On MODIS Data

Forest disturbance play an important impact on terrestrial ecosystems,which is also the main uncertain factor to assess the carbon balance of terrestrial ecosystems.Forest disturbance detection in a long sequence mainly needs to construct vegetation indices.But the response of spectral in vegetation change is complex,and there is no anyone index could completely capture the change information of multi-dimensional spectral.Existing forest disturbance indices have own-advantage and disadvantage.In order to improve the accuracy of monitoring,the development of algorithms should be studied.Using composed MODIS ground reflection products and land cover type dataset MCD12Q1 with the spatial resolution of 500m from 2001 to 2013,a new Integrated Forest Z-score is constructed.Forest disturbance in Fujian are monitored by vegetation change tracker.The forest disturbance area was extracted from 2001 to 2013 in Fujian province.The conclusions are as follow:(1)The comparison of forest disturbance indices shows that extraction capacity of Disturbance Index(DI)for forest cutting,plant diseases and insect pests,and afforestation is strong.The SI of DI and Integrated forest z-score(IFZ)are higher than 1 for different disturbance,which indicate that these two indices can be used to monitor multiple disturbance types.But the standard deviation of DI is higher and DI is unstable in the time series.IFZ is of the stability feature in time series.(2)The advantage of brightness,greenness and humidity and the stability of IFZ in time series which ean distinguish different land cover types better are concerned.The study development a new Integrated Forest Z-score(DIFZ).It is of the stability feature in time series and be more responsive in different forest disturbance types.(3)Verification points were selected by stratified sampling,the accuracy of DIFZ,Dl,Normalized Burn Ratio(NBR),Normalized Difference Moisture Index(NDMI)and Normalized Difference Vegetation Index(NDVI)were compared by using the mixed matrix statistics.The total accuracy of new index is 94%and its kappa coefficient reaches to 0.93.It shows that the monitoring accuracy of new index is higher than other indices,and the monitoring accuracy of forest disturbance is improved.(4)The result shows that a necessary congruence between the monitoring result by the new Integrated Forest Z-score and yearly forest loss dataset of Global Forest Watch in high resolution.Compared by forest pine wilt disease occurred in subcompartment survey data in Fujian Sanming city in 2009,the proposed index can better detect plant diseases and insect pests.It indicates that the new Integrated Forest Z-score can accurately monitor forest disturbance.