Developing Testate Amoebae-environmental Variable Transfer Function In Sphagnum-Dominated Peatlands In Changbai Mountains,Northeast China

At present,paleoenvironmental research is a hot topic in geoscience research.Through quantitatively reconstructing paleoenvironment can make a certain contribution to predicting future environmental changes in the study of past global environmental changes.The transfer function model is used as a method for quantitative reconstruction of paleoenvironment.At present,testate amoebae as a biological indicator of hydrological conditions,have made a lot of achievements in the study of peatland paleoenvironment abroad,but more and more attention has been paid to the application of testate amoebae in paleoenvironment in China.This paper selects five Sphagnum peatlands in the Changbai Mountains:Lushuihe（LSH）,Laolike（LLK）,Dongfanghong（DFH）,Hengshan（HS）and Anbei（AB）.According to the hydrological gradient set the profile line and set the sampling points evenly along the profile line.Each sampling point was covered different habitats such as the top,middle and low part of Sphagnum and selected for sampling.The latitude,longitude,altitude,DWT,and pH,of the sampling point are obtained.By establishing a database of testate amoebae species and environmental factors on the peatland of Changbai Mountain,the main environmental factors that affect the composition of testate amoebae species are determined.This paper discussed the ecological indicator significance of testate amoebae.We established the transfer function between testate amoebae and main environmental factors,optimized the transfer function and evaluated its performance,and applied it to the Dongfanghong peatland section in this area.The main conclusions are as follows:1.There were 133 effective samples from the five peatlands in the Changbai Mountains.Sixty-five species of testate amoebae were identified,with a total of 23,696 shells.Among them,the species with relative abundance greater than 2%were Assulina muscorum,Hyalosphenia elegans,Euglypha rotunda type,Cyclopyxis arcelloides type,Nebela tincta,Heleopera sphagni,Hyalosphenia papilio,Cryptodifflugia oviformis,Centropyxis aculeate type),Euglypha tuberculata type,Phryganella acropodia,et al.At the same time,four kinds of testate amoebae never appeared in the Changbai Mountains were discovered:Amphitrema wrightianum,Bullinularia indica,Wailesella eboracencis and Nebala jiuhuensis.2.This paper mainly chooses linear model redundancy analysis（RDA）to study the relationship between testate amoebae community and environmental variables.In the RDA,six environmental variables explained 37.9%of variation in species data,depth to water table（DWT）and pH individually explained significantly 21.8%and 12.4%respectively.Both variables reached p<0.001（Monte Carlo significance tests）.RDA analysis showed that DWT and pH of peatland were the main environmental factors affecting the composition of testate amoebae,which can be used as the target variable to develop the testate amoeba-environmental transfer function.3.First,we used the weighted average model（WA）、weighted average partial least squares（WA-PLS）and mximum likelihood（ML）to develop the transfer function of testate amoebae-water depth and pH.When we used the unfiltered 133 modern training sample sets to develop the transfer function,the prediction performances were poor.After filtering abnormal samples of residual values greater than 20%of the environmental gradient,the prediction performance of the models was significantly improved.WA.inv was better than other models in predicting depth to water table（DWT）(RMSEP_LOO=5.32,R²_LOO=0.82),and ML was the best model for predicting pH((RMSEP_LOO=0.17,R²_LOO=0.74).4.We used the Leave-one-site-out cross validation method to test the impact on the cluster structure sampling.After removing the abnormal samples,the performance of most models was lower than before.The difference between RMSEP_LOOand RMSEP_LOSO was not significant among the models.The RMSEP_seg produced by different models generally follows the trend where a higher frequency of samples equates to a lower RMSEP_seg,which were all lower than the standard deviation,indicating that the transfer function could predict the DWT land pH.A test of spatial autocorrelation was applied to all the models and showed that all models have a similar response that when the fraction of deleted samples increases,the performance deteriorates.Models were affected by spatial autocorrelation to some extent.5.Applying the constructed model to the DFH profile so that it could reconstruct the DWT and pH（1005-2018AD）.The reconstruction results showed that the DFH peatland environment had gradually become arid and weakly acidic to acidic in the past 1000 years.The environmental changing process of peatlands was divided into 3 stages:（1）53cm-28cm（1005-1959AD）The average DWT was 11.35±5.45cm;average pH was 5.60±0.22.（2）28cm-19cm（1959-1995AD）The average DWT was 12.47±5.63cm.Compared with the previous stage,the peatland environment tended to be arid;the average pH value was 5.32±0.16.（3）19cm-0cm（1995-2018AD）The DWT was 16.55±5.38cm,and the average pH were 5.04±0.15.At the lowest pH stage,the peatland environment was more acidic at this stage.