Study on vegetation-water level transfer function for wetlands in Sanjiang Plain.

Wang Chunling;Guo Jinping;Wang Guoping;Lou Yanjing;College of Urban and Rural Construction,Shanxi Agricultural University;Key Laboratory of Wetland Ecology and Environment,Northeast Institute of Geography and Agroecology,Chinese Academy of Sciences;

Detrended Correspondence Analysis(DCA);;WA model;;WAPLS model;;Predication performance

[Objective]Plant macrofossil is an important biological proxy for palaeoecology study.The transfer function of vegetation-water level can provide an effective way to reconstruct accurate quantitative palaeoenvironment and palaeoclimate.[Methods]In this paper,the WA(Weighted Averaging)and WAPLS(Weighted Averaging Partial Least Squares)models of riojapackage in R were used to establish the vegetation-water level transfer function of Sanjiang Plain.[Results]The results showed that the prediction performance of the WAPLS model was better than that of the WA model.The third component of the WAPLS model had the best prediction performance,with the lowest RMSEP value(3.69m)and the highest R2 value(0.52).After data filtering,the prediction performance of the two models were improved,the prediction performance of the fourth component of the WAPLS model was the best,with the lowest RMSEP 2.83,which reduced by 23.3% compared to the previous value.The accuracy reached to±2.83 cm,this result was relatively ideal.[Conclusion]This study provided a method and guarantee for the accurate quantitative reconstruction of the water level in Sanjiang Plain.