Construction And Application Of Remote Sensing Estimation Model For Phragmites Australis Wetland NPP

Net Primary Productivity(NPP)is one of the important parameters of wetland function,which directly reflects the production capacity of wetland vegetation communities in the natural environment and the quality status of wetland ecosystem.Phragmites australis wetland is a unique herbaceous wetland type.Accurate estimation of wetland NPP has important scientific significance for understanding regional carbon cycle and assessing ecosystem health.Considering of climate gradient and distribution pattern comprehensively,three typical wetlands—Qixinghe Wetland in Sanjiang Plain,Chagan Lake Wetland in Songnen Plain and Shuangtaihekou Wetland in Liaohe Plain of China are chosen as the study areas.Combining of field observation,remote sensing and GIS(Geographic Information System)analyses,and statistical analysis,we carry out the model building and evaluation,estimation of NPP based on this model,and spatial pattern and influencing factors analyses of three sample areas.In this paper,the medium resolution remote sensing satellite images of Landsat 8 OLI(Operational Land Imager)were selected as data sources.Object-oriented classification method was used to extract and mapping the Phragmites australis wetland distribution information.Linear Spectral Mixture Model were adopted to decompose the mixed pixels of Landsat 8 OLI image and accurately extract the spectral reflectance of Phragmites australis,then,all spectral vegetation indexes of three Phragmites australis wetlands were recalculated.Aboveground Biomass(AGB),Leaf Area Index(LAI),leaf chlorophyll content,Photosynthetically Active Radiation(PAR)data of Phragmites australis in three wetland areas were sampled in summer of 2014 when AGB of vegetation reaches the maximum.Sensitivity of different vegetation indexes on chlorophyll content,Light Use Efficiency(LUE),Absorbed Photosynthetic Active Radiation(APAR)and NPP were analyzed to determine the optimal vegetation indexes for LUE and APAP inversion.Based on the basic structure of light use efficiency model,remote sensing-based NPP models for Phragmites australis were constructed and evaluated,finally,optimal NPP model was determined.Through this optimal model,NPP estimation and analyses on spatial pattern and affecting factors were implemented for selected three wetlands.The aim of this study is to strengthen the understanding of carbon budget of Phragmites australis wetland ecosystem and to provide a scientific basis for the assessment of wetland environment and the management and protection of wetland ecosystem.Main conclusions we got are as follows.(1)Object-oriented Phragmites australis wetland extraction and spatial pattern analysisArea of the Qixinghe Wetland,Chagan Lake Wetland and Shuangtaihekou Wetland were 122.19,75.29 and 439.61 km2.The Qixinghe wetland has the highest proportion of Phragmites australis,with relatively dense patch and good landscape connectivity.Patches in the Chagan Lake wetland is relatively small.Area of Phragmites australis is the largest in the Shuangtaihekou wetland,mainly distributed in the northern reserve along the Shuangtaizi River.The overall classification accuracy of three wetlands is higher than 82% which indicate that fine classification of wetland at the community scale could be done ideally using the object-oriented classification method based on the multi-temporal remote sensing images to investigate the spectral features,shape features and phenological differences of different communities.(2)Vegetation indices accurate calculation and sensitivity analysis to the key parameters of light use efficiency modelUsing Linear Spectral Mixture Model to accurately extract wetland spectral reflectance information,a variety of remote sensing spectral vegetation indexes(VI)were calculated.In addition to EVI(Enhanced Vegetation Index)performed slightly worse in reflecting the differences of Phragmites australis wetland characteristics,other spectral vegetation indexes all showed a good description.NDVI(Normalized Difference Vegetation Index),RVI(Ratio Vegetation Index),WDVI(Weighted Difference Vegetati on Index),CIgreen(Green Chlorophyll Index)and MSAVI(Modified Soil Adjusted Vegetation Index)have the consistent spatial pattern characteristics.In general,the vegetation index values rank in the descending order: cultivated land > reed vegetation > other wetland vegetation > water body.The analysis found that there is a close relationship between LUE and chlorophyll content and vegetation index,which indicate that based on the ecology principle,LUE regional inversion can be achieved based on vegetation indexes.Among seven selected vegetation indexes,NDVI is the most sensitive to LUE(P < 0.01;y=8.4548x1.2025,R2=0.62),so it is choosen as the best VI to estimate teh LUE of Phragmites australis in this study.MSAVI has the most sensitive relationship to in APAR,with the correlation coefficient of 0.705.So we choose MSAVI as the optimal VI for APAR estimation.Above analysis also verified the feasibility of using vegetation index to estimate APAR.(3)NPP model based on basic structure of light use efficiency model: construction,evaluation and applicationBased on the sensitivity analysis of spectral vegetation indexes related to the chlorophyll content to LUE and APAR,this paper constructs a remote sensing model taking NDVI and MSAVI as parameters.Through comparison,the model of NPP =(–8E–06(937.36 × NDVI1.8918)2 + 0.0113(937.36 × NDVI1.8918)+ 0.9407)×(–8069.8 × MSAVI2 + 10292 × MSAVI – 1542.9)is determined as the best one to estimate wetland NPP with the accuracy 72%.Based on the estimated and actual average NPP value of Phragmites australis for the study area of this paper,the accuracy of BIO-BGC and CASA models are 13% and 24%,while that for the model constructed in this paper is 89%,Which clarify that this model is better than the commonly used NPP products as to the precision,which contributes to a better understanding of the role of Phragmites australis wetland in the regional carbon balance,and to provide the scientific basis for ecosystem management.Based on the above optimal model,Phragmites australis NPP for three wetlands were estimated.Mean NPP in descending order is Qixinghe Wetland of 3001 gC·m–2·yr–1,Chagan Lake Wetland of 3050 gC·m–2·yr–1,and Shuangtaihekou Wetland of 3621 gC·m–2·yr–1.Spatially,Phragmites australis NPP distribution is similar to the vegeatation indexes.Comparing the patterns of climatic factors with that of NPP,we found that the annual average temperature,annual precipitation,annual sunshine duration are important factors affecting the spatial pattern of NPP in three wetlands.To sum up,based on the plant physiology and ecology principle and advantages of remote sensing data,we can conveniently and efficiently achieve LUE accurate estimation based on vegetation index,which solve the key problem of vegetation NPP estimation and analyses and provide a new way for the study of regional vegetation NPP and carbon cycle.Meanwhile,we found that constructing NPP estimation model based on remote sensing data in the form of function combination taking vegetation indexes as independent variables is Practical and reliable based on basic structure of international mainstream model for vegetation NPP estimation.Estimated NPP using constructed model has high reliability.This study provides a new perspective for regional vegetation productivity estimation and analysis.For promotion of quantitative remote sensing application in the field of Wetland Science and promotion of global change and terrestrial ecosystems research in China is of great significanceLight use efficiency model is still the major means for NPP related researche s at home and abroad.In order to further understand and optimize light use efficiency model for wetland vegetation NPP estimation,it needs to strengthen the observation and investigation of ecological structure parameters for reed community,to shorten t he span of observation which could obtain more detailed and accurate field data,and help to explore wetland carbon cycle mechanism at different scales under the background of global change.At the same time,as to the remote sensing identification for plants community structure,we can try to combine a variety of methods and multi sources data to meet the increasingly strong demand of fine management of wetland ecological system on community structure identification.