Building Transfer Functions Between Testate amoeba and Environmental Variables with ‘rioja’ Package
LI Hong-kai1, 2, LI Wei-wei1, PU You-bao1, WANG Cong-yang1, WANG Song-mei1, YANG Xiao-lin1
1. State Environmental Protection Key Laboratory of Wetland Ecology and Vegetation Restoration, Northeast Normal University, Changchun,Jilin 130024, China2. Institute for Peat and Mire Research, School of Geographical Science, Northeast Normal University, Changchun, Jilin 130024, China
应用R语言rioja软件包的加权平均(Weighted Averaging,WA)和加权平均偏最小二乘(Weighted Averaging Partial Least Squares,WA-PLS)模型建立了长白山区泥炭藓泥炭地有壳变形虫与水位埋深(depth to water table,DWT)、pH和泥炭湿度的转换函数,为古环境定量重建奠定了基础,也提供了rioja软件包应用的实例和参考。结果表明水位埋深以WA-PLS模型最佳(预测均方根误差RMSEP为7.39 cm,R2=0.74);对于pH和泥炭湿度,WA-PLS第一分量和WA.inv都产生了最小的RMSEP和较高的R2值。pH的RMSEP为0.18,R2为0.72。泥炭湿度的RMSEP为1.95%,R2为0.62。如果泥炭剖面的有壳变形虫种类组成与本研究的训练样本集相同,水位埋深、pH和泥炭湿度可以分别以±7.39 cm、±0.18和±1.95%的平均误差进行重建。
R language, as an open source programming language and software environment, is widely used in statistics for its free availability. The ‘rioja’ package of R specially deals with the analysis of Quaternary science data, containing functions for constrained clustering, transfer functions and plotting stratigraphic data. Testate amoebae are a group of unicellular protists living in terrestrial habitats. Their decayed resistant and morphologically diagnostic shells (tests) allow them to be extensive used as proxy in peat based paleoenvironmental reconstruction. This study aimed to: ① Present an example of application of ‘rioja’ package; ② build Testate amoeba-based transfer functions for quantitatively reconstructing paleoenvironment changes in Changbai Mountains with peat archive. The training set was constituted by 75 samples collected from four peatlands, Hani(42°12′50″N, 126°31′05″E), Jinchuan(42°20′47″N, 126°21′35″E), Chichi(42°03′16″N,128°03′22″E) and Yuanchi(42°01′55″N,128°25′58″E), in Changbai Mountains, northeast China. Three factors, depth to water table (DWT), pH and peat moisture, were selected as the target environmental variables. The models of Weighted Averaging (WA) and Weighted Averaging Partial Least Squares (WA-PLS) were used to build transfer functions. Leave-one-out was chosen as cross validation method. The results showed that the second component of WA-PLS is the best models for DWT producing a RMSEP of 7.39 and R2 of 0.74. For pH and peat moisture, both first component of WA-PLS and WA with inverse deshrinking could be regarded as the best models for they have the lowest RMSEP and relatively higher R2. The RMSEP of pH is 0.18 and R2 is 0.72, while for peat moisture RMSEP is1.95% and R2 is 0.62. The performances of the transfer function were comparable with other studies in the world. DWT, pH and peat moisture could be quantitive reconstructed with the mean errors of ±7.39 cm, ±0.18 and ±1.95%, respectively, if Testate amoebae assemblage of profiles was the same as the training set in this study.
Keywords:‘rioja’ package
;
transfer function
;
Testate amoebae
;
depth to water table
;
pH
;
peat moisture
LIHong-kai, LIWei-wei, PUYou-bao, WANGCong-yang, WANGSong-mei, YANGXiao-lin. Building Transfer Functions Between Testate amoeba and Environmental Variables with ‘rioja’ Package[J]. Scientia Geographica Sinica, 2013, 33(8): 1022-1028 https://doi.org/10.13249/j.cnki.sgs.2013.08.1022
rioja是C2的开发者Steve Juggins博士开发的基于R语言的扩展软件包[44]。其功能和C2相当,主要针对第四纪研究中常见的数据分析,包括转换函数、约束聚类、剖面图绘制等。转换函数模型主要有加权平均(Weighted Averaging,WA)。包括4种变型:典型回归加权平均(WA classical deshrinking,WA.cla)、反向回归加权平均(WA inverse deshrinking,WA.inv)、典型回归耐受值降权加权平均(Tolerance downweighted WA classical deshrinking,WA.cla.tol)和反向回归耐受值降权加权平均(Tolerance downweighted WA inverse deshrinking,WA.inv.tol)),加权平均偏最小二乘(Weighted Averaging Partial Least Squares,WA-PLS),偏最小二乘(Partial Least Squares,PLS),最大似然(Maximum Likelihood,ML),现代模拟技术(包括Modern Analogue Technique,MAT和Weighted Average Modern Analogue Technique,WMAT),Imbrie & Kipp因子分析 (Imbrie and Kipp Factor Analysis,IKFA)。
本文的物种数据为物种的相对丰度。物种数据和环境因子数据以每样一行的方式分别输入,并读取到R工作空间。排序分析表明水位埋深(depth to water table,DWT),pH和泥炭湿度是影响长白山区有壳变虫的组成的主要环境因子,可作为目标变量进行转换函数的构建[41]。转换函数模型采用WA和WA-PLS,rioja中分别对应函数WA()和WAPLS()。结果的交叉验证采用leave-one-out,也称jack-knifing,通过crossval(…, cv.method=“loo”)实现。评价模型预测性能的优劣主要根据预测均方根误差(root mean square error of prediction,RMSEP)和R2(预测值和实测值相关系数平方)。
Rees A B H, Pedersen C R, et al. Depth distribution of chironomids and an evaluation of site-specific and regional lake-depth inference models: A good model gone bad
The ecology of testate amoebae (Protozoa: Rhizopoda) in oceanic peatlands in Newfoundland, Canada: modelling hydrological relationships for palaeoenvironmental reconstruction
Testate amoebae ecology and a local transfer function from a peatland in western Poland
[J]. , 2008,28(1):164-175.
[24]
LamentowiczM,
Mitchell, E A D. The ecology of testate amoebae (protists) in Sphagnum in north-western Poland in relation to peatland ecology
[J]. ,2005,50(1):48-63.
[25]
Mitchell E AD, Buttler AJ, Warner BGet al.
Ecology of testate amoebae (Protozoa: Rhizopoda) in Sphagnum peatlands in the Jura mountains, Switzerland and France
[J]. ,1999,6(4):565-576.
[26]
Payne RJ,
Mitchell E A D. Ecology of testate amoebae from mires in the central Rhodope Mountains, Greece and development of a transfer function for palaeohydrological reconstruction
[J]. ,2007,158:159-171.
[27]
Payne RJ, Charman DJ. and MatthewsS.
Testate amoebae as paleohydrological proxies in Sürmene Ağaçbaşi Yaylasi peatland (Northeast Turkey)
[J]. ,2008,28(2):311-323.
[28]
Wilmshurst JM, Wiser SK, Charman DJ.
Reconstructing Holocene water tables in New Zealand using testate amoebae: differential preservation of tests and implications for the use of transfer functions
Summer water deficit variability controls on peatland water-table changes: Implications for Holocene palaeoclimate reconstructions
[J]. ,2007,17(2):217-227.
[32]
Charman DJ, HohlV, BlundellA, et al.
A 1000-year reconstruction of summer precipitation from Ireland: Calibration of a peat-based palaeoclimate record
[J]. ,2012,268:87-97.
[33]
Elliott SM, Roe HM, Patterson RT.
Testate amoebae as indicators of hydroseral change: An 8500 year record from Mer Bleue Bog, eastern Ontario, Canada
[J]. ,2012,268:128-144.
[34]
HendonD, Charman DJ, KentM.
Palaeohydrological records derived from testate amoebae analysis from peatlands in northern England: within-site variability, between-site comparability and palaeoclimatic implications
[J]. ,2001,11(2):127-148.
[35]
HendonD, Charman DJ.
High-resolution peatland water-table changes for the past 200 years: the influence of climate and implications for management
[J]. ,2004,14(1):125-134.
[36]
LamarreA, GarneauM, AsnongH, et al.
Holocene paleohydrological reconstruction and carbon accumulation of a permafrost peatland using testate amoeba and macrofossil analyses, Kuujjuarapik, subarctic Québec, Canada
[J]. ,2012,186:131-141.
[37]
LamentowiczM, CedroA, GałkaM, et al.
Last millennium palaeoenvironmental changes from a Baltic bog (Poland) inferred from stable isotopes, pollen, plant macrofossils and testate amoebae
[J]. ,2008, 265(1-2):93-106.
[38]
Mitchell E AD, van der Knaap W O, van Leeuwen J F N, et al.
The palaeoecological history of the Praz-Rodet bog (Swiss Jura) based on pollen, plant macrofossils and testate amoebae (Protozoa)
[J]. ,2001,11(1):65-80.
[39]
SchoningK, Charman DJ, WastegårdS.
Reconstructed water tables from two ombrotrophic mires in eastern central Sweden compared with instrumental meteorological data
Rees A B H, Pedersen C R, et al. Depth distribution of chironomids and an evaluation of site-specific and regional lake-depth inference models: A good model gone bad
The ecology of testate amoebae (Protozoa: Rhizopoda) in oceanic peatlands in Newfoundland, Canada: modelling hydrological relationships for palaeoenvironmental reconstruction
Testate amoebae ecology and a local transfer function from a peatland in western Poland
0
2008
Mitchell, E A D. The ecology of testate amoebae (protists) in Sphagnum in north-western Poland in relation to peatland ecology
0
2005
Ecology of testate amoebae (Protozoa: Rhizopoda) in Sphagnum peatlands in the Jura mountains, Switzerland and France
0
1999
Mitchell E A D. Ecology of testate amoebae from mires in the central Rhodope Mountains, Greece and development of a transfer function for palaeohydrological reconstruction
0
2007
Testate amoebae as paleohydrological proxies in Sürmene Ağaçbaşi Yaylasi peatland (Northeast Turkey)
0
2008
Reconstructing Holocene water tables in New Zealand using testate amoebae: differential preservation of tests and implications for the use of transfer functions
Summer water deficit variability controls on peatland water-table changes: Implications for Holocene palaeoclimate reconstructions
0
2007
A 1000-year reconstruction of summer precipitation from Ireland: Calibration of a peat-based palaeoclimate record
0
2012
Testate amoebae as indicators of hydroseral change: An 8500 year record from Mer Bleue Bog, eastern Ontario, Canada
0
2012
Palaeohydrological records derived from testate amoebae analysis from peatlands in northern England: within-site variability, between-site comparability and palaeoclimatic implications
0
2001
High-resolution peatland water-table changes for the past 200 years: the influence of climate and implications for management
0
2004
Holocene paleohydrological reconstruction and carbon accumulation of a permafrost peatland using testate amoeba and macrofossil analyses, Kuujjuarapik, subarctic Québec, Canada
0
2012
Last millennium palaeoenvironmental changes from a Baltic bog (Poland) inferred from stable isotopes, pollen, plant macrofossils and testate amoebae
0
2008
The palaeoecological history of the Praz-Rodet bog (Swiss Jura) based on pollen, plant macrofossils and testate amoebae (Protozoa)
0
2001
Reconstructed water tables from two ombrotrophic mires in eastern central Sweden compared with instrumental meteorological data
... 本文的物种数据为物种的相对丰度.物种数据和环境因子数据以每样一行的方式分别输入,并读取到R工作空间.排序分析表明水位埋深(depth to water table,DWT),pH和泥炭湿度是影响长白山区有壳变虫的组成的主要环境因子,可作为目标变量进行转换函数的构建[41].转换函数模型采用WA和WA-PLS,rioja中分别对应函数WA()和WAPLS().结果的交叉验证采用leave-one-out,也称jack-knifing,通过crossval(…, cv.method=“loo”)实现.评价模型预测性能的优劣主要根据预测均方根误差(root mean square error of prediction,RMSEP)和R2(预测值和实测值相关系数平方). ...