解雪峰(1991—),男,博士,讲师,主要从事土地利用及其环境效应研究。E-mail:
试验设置了对照处理(CK)、有机肥(OM)、聚丙烯酰胺+有机肥(PAM+OM)、秸秆覆盖+有机肥(SM+OM)、秸秆深埋+有机肥(BS+OM)和生物菌肥+有机肥(BM+OM)6个处理方式来探讨滨海盐碱地不同改良方式对土壤含盐量、pH、钠吸附比(SAR)和碱化度(ESP)的影响,进而识别影响土壤盐渍化程度的主要因子,并构建多元线性回归模型(Multi-linear Regression,MLR)、BP神经网络模型(BP Artificial Neural Network,BP-ANN)和随机森林模型(Random Forest,RF)对滨海重度盐碱地改良背景下的土壤盐渍化参数进行模拟预测。研究结果表明:各改良措施均能有效地降低表层土壤盐渍化水平,其中SM+OM处理对于土壤含盐量的抑制效果最好,而BM+OM处理则对于土壤碱分的抑制效果最好。改良过程中气象条件和土壤性质均对表层土壤盐渍化水平产生了显著影响。在模型预测中,随机森林模型对土壤含盐量、pH、SAR和ESP的综合预测精度明显优于BP神经网络模型和多元线性回归模型,体现在随机森林模型具有较高决定系数(Coefficient of determination,
Soil salinization is one of the main types of land degradation, which seriously inhibits the improvement of soil quality and the growth and grain yield of crops. Reclamation of coastal land is increasingly being used as a means of raising agricultural productivity and improving food security in China. Determining the importance of potential influencing factors of soil salinization parameters and thus predicting their concentrations are important for formulating targeted control measures to improve soil quality and crop yield in tidal flat reclamation areas.
In this study, six treatments including control (CK), organic manure (OM), polyacrylamide plus organic manure (PAM+OM), straw mulching plus organic manure (SM+OM), buried straw plus organic manure (BS+OM), and bio-organic manure plus organic manure (BM+OM) were applied to explore the effect of different reclamation treatments on different soil parameters. The effect of all treatments on soil salt content (SSC), pH, sodium adsorption ratio (SAR), and exchange sodium percentage (ESP) was analyzed and the main factors affecting the degree of soil salinization were identified. Thereafter, the multi-linear regression model (MLR), BP artificial neural network model (BP-ANN), and random forest model (RF) were conducted to predict the soil salinization parameters (SSC, pH, SAR, and ESP)using covariates, such as air temperature, precipitation, evaporation, wind speed, soil water content, soil temperature, and soil bulk density.
The results indicated that the concentration of SSC, SAR, and ESP gradually increased, while the pH gradually decreased during the oat growing stage. All reclamation treatments effectively reduced the level of surface soil salinization. Among them, SM+OM treatment had the best inhibition effect on SSC, whereas BM+OM treatment had the best inhibition effect on soil pH, SAR and ESP. Besides, both meteorological parameters and soil properties had a significant impact on the level of surface soil salinization during the amelioration of coastal saline-alkali land. Additionally, the RF model performed much better than BP-ANN and MLR as it revealed a much higher coefficient of determination (R2) and Nash-Sutcliffe efficiency (NSE), and lower root mean square error (RMSE) than BP-ANN and MLR model.
The above results indicate that the reclamation treatments can effectively inhibit soil evaporation, improve soil structure, increase soil water holding capacity, and thus reduce the salinization level of surface soil. Our results also suggest that the RF model is a more powerful modeling approach in predicting soil salinization dynamics of coastal saline-alkali land due to its advantages in handling the nonlinear and hierarchical relationships between soil salinization parameters and covariates, and insensitivity to overfitting and the presence of noise in the data. Thus, our findings could provide a reference for predicting the soil salinization parameters in areas with similar environmental conditions.
土壤盐渍化是土地退化的主要形式之一,严重抑制了土壤性质的改善和作物的生长发育[
试验田块于2016年9月在江苏省南通市通州湾滨海园区建立(32°11′ N,121°22′ E),属北亚热带季风性湿润气候,多年平均降水量和多年平均气温分别为1 029 mm和15.0℃;地下水埋深为1.2~1.8 m。试验区于2008年围垦,主要用于海水养殖;土壤来源于现代的海相和河相沉积,属于砂质壤土。试验区土壤含盐量、pH、钠吸附比和碱化度分别在6.5~15.0 g·kg–1、8.06~8.68、13.29~38.76和15.44%~35.85%之间,属重度盐碱地[
试验设计和具体措施
Experimental design and specific measures
处理措施Treatment | 具体措施Specific measures |
1)Control:对照,2)Organic manure:有机肥,3)Polyacrylamide plus organic manure:聚丙烯酰胺+有机肥,4)Straw mulching plus organic manure:秸秆覆盖+有机肥,5)Buried straw plus organic manure:秸秆深埋+有机肥,6)Bio-organic manure plus organic manure:生物菌肥+有机肥。 | |
CK1) | 对照处理,无任何措施 |
OM2) | 在0~20 cm深度施用鸡粪有机肥,用量为15 t·hm–2 |
PAM+OM3) | 综合施用PAM结构改良剂和鸡粪有机肥,其中PAM结构改良剂为非离子800万分子量,用量为2 t·hm–2;均匀施用于20 cm深度,用量为15 t·hm–2 |
SM+OM4) | 在0~20 cm深度施用鸡粪有机肥,用量为15 t·hm–2;然后将小麦秸秆剪为10 cm每段,以15 t·hm–2的用量覆于地表 |
BS+OM5) | 将小麦秸秆剪为10 cm每段,以15 t·hm–2的用量埋于地下20 cm深度,然后在0~20 cm深度施用鸡粪有机肥,用量为15 t·hm–2 |
BM+OM6) | 复合施用嘉华生物菌肥和鸡粪有机肥,其中嘉华生物菌肥含生物细菌0.2亿个·克–1,有机质≥45%,有效养分(N+P+K)≥12%;二者均匀施用于20 cm深度,用量为15 t·hm–2 |
研究区大气降雨量、大气温度、风速、蒸发量等参数由小型气象站监测,试验期内气象参数变化特征如
试验期内气象参数变化特征
Characteristics of meteorological parameters during the period of the experiment
多元线性回归模型(Mutiple Linear Regression,MLR)因其结构简单、易于计算和解释,已成为土壤科学领域较为常用的预测模型之一[
式中,
BP神经网络(BP-Artificial Neural Network,BP-ANN)是一种通过类似于人类神经系统的信息处理技术,在生物神经网络的启示下建立的数据处理模型[
式中,
在隐含层的选择中,选择四层神经网络模型,网络中间隐含层神经元的传递函数采用双曲正切函数(tanh),即tanh(x)= [exp(x)– exp(–x)]/[exp(x)+ exp(–x)],输出层神经元传递函数采用恒等函数,优化算法采用贝叶斯规则法(TRAINBR),训练目标误差为0.001[
式中,
随机森林(Random Forest,RF)结合了分类与回归树、随机属性选择和装袋的算法思想,使得每棵分类回归树均完全生长,从而得到了低偏移的树;同时,随机属性选择和装袋的方法使得随机森林中的每个个体其相关性均较低[
所有数据的统计分析检验均在SPSS 19.0 for Windows软件中完成,并在Sigma Plot 13.0软件中进行图形绘制。BP神经网络模型、随机森林模型和多元线性回归模型均在SPSS Modeler 18.0软件中构建。模型构建过程,随机选取70%为训练数据集,剩下30%为验证数据集。土壤含盐量、土壤pH、土壤SAR和土壤ESP的预测精度通过决定系数(Coefficient of Determination,
燕麦生育期内,各处理方式下表层土壤含盐量均随时间推移呈逐渐上升的趋势,且不同改良方式均能有效地抑制土壤盐分(
不同改良方式a)土壤含盐量、b)pH、c)钠吸附比和d)碱化度动态特征
Dynamic of soil salt content (a), pH (b), sodium adsorption ratio (c), and exchange sodium percentage (d) under different treatment
燕麦生育期内,所有处理方式下表层土壤pH均随时间推移呈逐渐下降的趋势,且不同改良方式下表层土壤pH同样差异显著(
燕麦生育期内,各处理方式下表层土壤SAR均随时间推移逐渐上升,且不同改良方式在整个生育期内均能有效的降低土壤SAR水平,以BM+OM处理的效果最好(
燕麦生育期内,各处理方式下表层土壤ESP的变化特征与SAR相似,均随时间推移逐渐上升,且不同改良方式下土壤ESP也存在显著差异,以BM+OM处理最低(
Pearson相关分析表明燕麦生育期内表层土壤盐渍化程度显著受土壤环境因子影响(
表层土壤盐渍化参数与土壤环境因子相关关系
Correlation between soil salinization parameters and soil environmental factors in surface soil
大气温度 |
降雨量 |
蒸发量 |
风速 |
土壤含水量 |
土壤温度 |
土壤容重 |
|
① Soil salt content,② Soil sodium adsorption ratio,③ Soil exchange sodium percentage。注:*和**分别表示在0.05和0.01水平上显著相关。Note:* and ** represents a significance at 5% and 1% level,respectively. | |||||||
土壤含盐量① | 0.215* | –0.336** | 0.586** | 0.289** | –0.525** | 0.307** | 0.644** |
pH | –0.187* | 0.245** | –0.156 | –0.205* | 0.077 | –0.244** | 0.222* |
土壤钠吸附比② | 0.126 | –0.192* | 0.462** | 0.202* | –0.521** | 0.184 | 0.723** |
土壤碱化度③ | 0.116 | –0.184 | 0.450** | 0.211* | –0.539** | 0.178 | 0.733** |
土壤含盐量全部数据、训练数据和验证数据的预测效果均表现为随机森林模型 > BP神经网络模型 > 多元线性回归模型(
土壤含盐量动态a)多元线性回归模型、b)BP神经网络模型、c)随机森林模型预测与验证
Dynamic prediction and verification of soil salt content based on multiple linear regression model (a), BP neural network model (b) and random forest model (c)
如
土壤pH动态a)多元线性回归模型、b)BP神经网络模型、c)随机森林模型预测与验证
Dynamic prediction and verification of soil pH based on multiple linear regression model (a), BP neural network model (b) and random forest model (c)
土壤SAR全部数据、训练数据和验证数据的预测效果同样表现为随机森林模型 > BP神经网络模型 > 多元线性回归模型(
土壤钠吸附比动态a)多元线性回归模型、b)BP神经网络模型、c)随机森林模型预测与验证
Dynamic prediction and verification of soil sodium adsorption ratio based on multiple linear regression model (a), BP neural network model (b) and random forest model (c)
土壤ESP全部数据、训练数据和验证数据的预测效果同样表现为随机森林模型 > BP神经网络模型 > 多元线性回归模型(
土壤碱化度动态a)多元线性回归模型、b)BP神经网络模型、c)随机森林模型预测与验证
Dynamic prediction and verification of soil exchange sodium percentage based on multiple linear regression model (a), BP neural network model (b) and random forest model (c)
本研究中,SM+OM、BS+OM、PAM+OM、BM+OM和OM处理下的土壤含盐量、SAR和ESP在生育期内均较CK处理明显下降,表明各改良措施均能显著降低表层土壤盐渍化水平。本研究中,SM+OM处理在整个生育期内均能够非常显著降低土壤盐渍化程度,基本在轻度盐碱土和中度盐碱土的范围内波动。诸多研究表明秸秆覆盖能够增加地表覆盖度,提高土壤持水能力、调控土壤温度,抑制表层土壤水分蒸发和潜水上行,降低了根区土壤盐渍化程度[
土壤盐渍化动态同样也受到气象条件和土壤性质的影响[
本研究表明,随机森林模型和BP神经网络模型均能很好地应用于改良背景下的土壤盐分、pH、土壤SAR和土壤ESP的预测,而多元线性回归模型的预测效果则较差。这主要是因为在改良背景下,土壤盐渍化参数同时受到多种环境因素的影响(如气候因素、土壤结构因素、土壤肥力等),其空间变异特征较大[
(1)不同改良措施均能显著地降低表层土壤盐渍化水平,其中SM+OM措施对土壤含盐量的抑制效果最好;而BM+OM措施对pH、钠吸附比和碱化度的调控效果最佳。(2)土壤盐渍化程度显著受气象条件和土壤性质的影响,其中土壤含盐量与大气温度、土壤温度、降水量、蒸发量、土壤含水量、风速和土壤容重均存在显著相关关系;pH与大气温度、降水量、风速、土壤温度和土壤容重存在显著相关关系;SAR和ESP则均与蒸发量、风速、土壤含水量和土壤容重显著相关。(3)在预测模型中,RF模型对土壤含盐量、pH、SAR和ESP的综合预测精度明显优于BP-ANN和MLR,主要体现在随机森林模型具有较高的
Zhang J B, Yang J S, Yao R J, et al. Dynamics of soil water, salt and crop growth under farmyard manure and mulching in coastal tidal flat soil of northern Jiangsu Province[J]. Transactions of the Chinese Society of Agricultural Engineering, 2013, 29(15): 116-125.
张建兵, 杨劲松, 姚荣江, 等. 有机肥与覆盖方式对滩涂围垦农田水盐与作物产量的影响[J]. 农业工程学报, 2013, 29(15): 116-125.
Gao S, Yang J S, Yao R J, et al. Effects of soil amelioration measures mitigating soil salinity and improving crop P uptake in coastal area of north Jiangsu[J]. Acta Pedologica Sinica, 2020, 57(5): 1219-1229.
高珊, 杨劲松, 姚荣江, 等. 改良措施对苏北盐渍土盐碱障碍和作物磷素吸收的调控[J]. 土壤学报, 2020, 57(5): 1219-1229.
Li J G, Pu L J, Zhu M, et al. The present situation and hot issues in the salt-affected soil research[J]. Acta Geographica Sinica, 2012, 67(9): 1233-1245.
李建国, 濮励杰, 朱明, 等. 土壤盐渍化研究现状及未来研究热点[J]. 地理学报, 2012, 67(9): 1233-1245.
Shen H Y, Xie X F, Pu L J, et al. Dynamics of soil water-salt as affected by different improvement measures in coastal reclamation: A case study of Rudong in Jiangsu Province[J]. Chinese Journal of Soil Science, 2019, 50(2): 316-322.
沈洪运, 解雪峰, 濮励杰, 等. 不同改良方式下滩涂围垦区土壤水盐动态--以江苏如东为例[J]. 土壤通报, 2019, 50(2): 316-322.
Xie X F, Pu L J, Zhu M, et al. Spatio-temporal variability of soil salinity and sodicity in agricultural reclaimed coastal wetlands, Eastern China[J]. Archives of Agronomy and Soil Science, 2020, 66(12): 1639-1650.
Zhang J B, Yang J S, Yao R J, et al. The effects of farmyard manure and mulch on soil physical properties in a reclaimed coastal tidal flat salt-affected soil[J]. Journal of Integrative Agriculture, 2014, 13(8): 1782-1790.
Yin A J, Zhang M, Gao C, et al. Salinity evolution of coastal soils following reclamation and intensive usage, Eastern China[J]. Environmental Earth Sciences, 2016, 75(18): 1-11.
Xie X F, Pu L J, Zhu M, et al. Linkage between soil salinization indicators and physicochemical properties in a long-term intensive agricultural coastal reclamation area, Eastern China[J]. Journal of Soils and Sediments, 2019, 19(11): 3699-3707.
Xie X F, Pu L J, Zhu M, et al. Differential effects of various reclamation treatments on soil characteristics: an experimental study of newly reclaimed tidal mudflats on the east China coast[J]. Science of the Total Environment, 2021, 768: 144996.
Shen H, Abuduwaili J. Spatial distribution of soil moisture and salinity and their influence factors in the farmland of Manas River catchment, Northwest China[J]. Chinese Journal of Applied Ecology, 2015, 26(3): 769-776.
沈浩, 吉力力·阿不都外力. 玛纳斯河流域农田土壤水盐空间分布特征及影响因素[J]. 应用生态学报, 2015, 26(3): 769-776.
Yao R J, Yang J S, Zou P, et al. Bp neural network model for spatial distribution of regional soil water and salinity[J]. Acta Pedologica Sinica, 2009, 46(5): 788-794.
姚荣江, 杨劲松, 邹平, 等. 区域土壤水盐空间分布信息的BP神经网络模型研究[J]. 土壤学报, 2009, 46(5): 788-794.
Xie X F, Pu L J, Zhu M, et al. Evolution and prospects in modeling of water and salt transport in soils[J]. Scientia Geographica Sinica, 2016, 36(10): 1565-1572.
解雪峰, 濮励杰, 朱明, 等. 土壤水盐运移模型研究进展及展望[J]. 地理科学, 2016, 36(10): 1565-1572.
Guo X H, Bi Y J, Sun X H, et al. Prediction model of soil water and salt transport on yield of summer squash under mulch drip irrigation with brackish water[J]. Transactions of the Chinese Society of Agricultural Engineering, 2019, 35(8): 167-175.
郭向红, 毕远杰, 孙西欢, 等. 西葫芦微咸水膜下滴灌土壤水盐运移对产量影响的预测模型[J]. 农业工程学报, 2019, 35(8): 167-175.
Guo Y, Yin X W, Li Y, et al. Soil water and salt dynamics and its coupling model at cropland-treebelt-desert compound system[J]. Transactions of the Chinese Society of Agricultural Engineering, 2019, 35(17): 87-101.
郭勇, 尹鑫卫, 李彦, 等. 农田-防护林-荒漠复合系统土壤水盐运移规律及耦合模型建立[J]. 农业工程学报, 2019, 35(17): 87-101.
Zou P, Yang J S, Fu J R, et al. Artificial neural network and time series models for predicting soil salt and water content[J]. Agricultural Water Management, 2010, 97(12): 2009-2019.
Wu W C, Zucca C, Muhaimeed A S, et al. Soil salinity prediction and mapping by machine learning regression in Central Mesopotamia, Iraq[J]. Land Degradation & Development, 2018, 29(11): 4005-4014.
Yang A X, Ding J L. Comparative assessment of two methods for estimation of soil organic carbon content by Vis-NIR spectra in Xinjiang Ebinur Lake Wetland[J]. Transactions of the Chinese Society of Agricultural Engineering, 2015, 31(18): 162-168.
杨爱霞, 丁建丽. 新疆艾比湖湿地土壤有机碳含量的光谱测定方法对比[J]. 农业工程学报, 2015, 31(18): 162-168.
Vermeulen D, van Niekerk A. Machine learning performance for predicting soil salinity using different combinations of geomorphometric covariates[J]. Geoderma, 2017, 299: 1-12.
Wang Y Y, Qi Y B, Chen Y, et al. Prediction of soil organic matter based on multi-resolution remote sensing data and random forest algorithm[J]. Acta Pedologica Sinica, 2016, 53(2): 342-354.
王茵茵, 齐雁冰, 陈洋, 等. 基于多分辨率遥感数据与随机森林算法的土壤有机质预测研究[J]. 土壤学报, 2016, 53(2): 342-354.
Xie X F, Wu T, Zhu M, et al. Comparison of random forest and multiple linear regression models for estimation of soil extracellular enzyme activities in agricultural reclaimed coastal saline land[J]. Ecological Indicators, 2021, 120: 106925.
Xie X F, Pu L J, Shen H Y, et al. Effects of soil reclamation on the oat cultivation in the newly reclaimed coastal land, Eastern China[J]. Ecological Engineering, 2019, 129: 115-122.
Bao S D. Soil and agricultural chemistry analysis[M]. 3rd. ed. Beijing: China Agriculture Press, 2000.
鲍士旦. 土壤农化分析[M]. 3版. 北京: 中国农业出版社, 2000.
Xie X F, Pu L J, Zhu M, et al. Relationship between landscape patterns and soil salinity based on canonical correspondence analysis in coastal reclamation area[J]. Geographical Research, 2017, 36(3): 495-505.
解雪峰, 濮励杰, 朱明, 等. 基于典范对应分析的滨海滩涂围垦区景观格局与土壤盐渍化关系[J]. 地理研究, 2017, 36(3): 495-505.
Thompson J A, Pena-Yewtukhiw E M, Grove J H. Soil-landscape modeling across a physiographic region: Topographic patterns and model transportability[J]. Geoderma, 2006, 133(1/2): 57-70.
Lagos-Avid M P, Bonilla C A. Predicting the particle size distribution of eroded sediment using artificial neural networks[J]. Science of the Total Environment, 2017, 581/582: 833-839.
Yuan Y, Tan D, Yu S J, et al. A prediction model for shale gas organic carbon content based on improved BP neural network using Bayesian regularization[J]. Geology and Exploration, 2019, 55(4): 1082-1091.
袁颖, 谭丁, 于少将, 等. 基于贝叶斯正则化改进BP神经网络的页岩气有机碳含量预测模型[J]. 地质与勘探, 2019, 55(4): 1082-1091.
Zhao Y G, Li Y Y, Wang J, et al. Buried straw layer plus plastic mulching reduces soil salinity and increases sunflower yield in saline soils[J]. Soil and Tillage Research, 2016, 155: 363-370.
Abd El-Mageed T A, Semida W M, Abd El-Wahed M H. Effect of mulching on plant water status, soil salinity and yield of squash under summer-fall deficit irrigation in salt affected soil[J]. Agricultural Water Management, 2016, 173: 1-12.
Zhao Y G, Pang H C, Wang J, et al. Effects of straw mulch and buried straw on soil moisture and salinity in relation to sunflower growth and yield[J]. Field Crops Research, 2014, 161: 16-25.
Lu S Y, Chen F X, Ngo H H, et al. Effect of straw and polyacrylamide on the stability of land/water ecotone soil and the field implementation[J]. Ecological Engineering, 2016, 94: 12-21.
Chowdhury N, Nakatani A S, Setia R, et al. Microbial activity and community composition in saline and non-saline soils exposed to multiple drying and rewetting events[J]. Plant and Soil, 2011, 348(1/2): 103-113.
Wang Z R, Zhao G X, Gao M X, et al. Spatial variability of soil salinity in coastal saline soil at different scales in the Yellow River Delta, China[J]. Environmental Monitoring and Assessment, 2017, 189(2): 1-12.
Wang T T, Zhai J H, He H, et al. Applicability of BP neural network model and SVM model to predicting soil moisture under incorporation of biochar into soils[J]. Research of Soil and Water Conservation, 2017, 24(3): 86-91.
王彤彤, 翟军海, 何欢, 等. BP神经网络和SVM模型对施加生物炭土壤水分预测的适用性[J]. 水土保持研究, 2017, 24(3): 86-91.
Xie E Z, Zhao Y C, Lu F Y, et al. Comparison analysis of methods for prediction of spatial distribution of soil organic matter contents in farmlands south Jiangsu, China[J]. Acta Pedologica Sinica, 2018, 55(5): 1051-1061.
谢恩泽, 赵永存, 陆访仪, 等. 不同方法预测苏南农田土壤有机质空间分布对比研究[J]. 土壤学报, 2018, 55(5): 1051-1061.
Yu G Q, Li Z B, Zhang X, et al. Dynamic simulation of soil water-salt using BP neural network model and grey correlation analysis[J]. Transactions of the Chinese Society of Agricultural Engineering, 2009, 25(11): 74-79.
于国强, 李占斌, 张霞, 等. 土壤水盐动态的BP神经网络模型及灰色关联分析[J]. 农业工程学报, 2009, 25(11): 74-79.