Model ensemble techniques of machine learning algorithms for soil moisture constants in the semi-arid climate conditions

被引:0
|
作者
Alaboz, Pelin [1 ]
机构
[1] Isparta Univ Appl Sci, Dept Soil Sci & Plant Nutr, Isparta, Turkiye
关键词
model average; pedotransfer functions; soil water; eau du sol; moyenne du mod & egrave; le; fonctions de p & eacute; dotransfert; POINT;
D O I
10.1002/ird.3037
中图分类号
S3 [农学(农艺学)];
学科分类号
0901 ;
摘要
In recent years, the use of prediction models based on intelligent algorithms has become widespread in soil science. However, each algorithm has advantages and disadvantages, and variable results can occur on different datasets. The evaluation of ensemble techniques for solving these problems is the current approach. Water problems will arise due to global warming, and soil water will become more important. This study aims to evaluate the predictive accuracy of different machine learning algorithms (support vector machine regression (SVR), random forest (RF), artificial neural network (ANN), and multivariate linear regression (MLR)) and ensemble techniques (equal weight [EQ], Bates-Granger-BG), Granger-Ramanathan (GR), Akaike information criterion (AIC), and Bayesian information criterion (BIC)) on the field capacity (FC), wilting point (WP) and available water content (AWC) of soils. As a result, higher prediction accuracy was obtained with the RF algorithm than with the value machine learning algorithm in the estimation of moisture constants. The coefficients of determination (R2) obtained for the prediction of FC, WP, and AWC via the RF algorithms were 0.624, 0.759 and 0.641, respectively. MLR had the highest error rate. Among the ensemble techniques, GR was the most successful. Lin's concordance correlation coefficient (LCCC) values obtained from the estimation of FC, WP, and AWC with the GR model were 0.801, 0.894, and 0.801, respectively. The root mean squared error (RMSE) and mean absolute error (MAE) values obtained in the estimation of the available water content with the MLR algorithm were 1.905 and 1.435, respectively, whereas these values were 1.173 and 0.767, respectively, when the GR model was used. As a result of the present study, better predictive results were obtained with ensemble techniques instead of evaluating the algorithms individually. Ces derni & egrave;res ann & eacute;es, l'utilization de mod & egrave;les de pr & eacute;diction bas & eacute;s sur des algorithmes intelligents s'est r & eacute;pandue dans la science des sols. Cependant, chaque algorithme pr & eacute;sente des avantages et des inconv & eacute;nients, et des r & eacute;sultats variables peuvent se produire sur diff & eacute;rents ensembles de donn & eacute;es. L'approche actuelle est d'& eacute;valuer des techniques d'ensemble pour r & eacute;soudre ces probl & egrave;mes. Les probl & egrave;mes d'eau vont appara & icirc;tre en raison du r & eacute;chauffement climatique, et l'eau du sol va devenir plus importante. Cette & eacute;tude vise & agrave; & eacute;valuer la pr & eacute;cision pr & eacute;dictive de diff & eacute;rents algorithmes d'apprentissage automatique (r & eacute;gression & agrave; vecteurs de support (SVR), for & ecirc;t al & eacute;atoire (RF), r & eacute;seau de neurones artificiels (ANN) et r & eacute;gression lin & eacute;aire multivari & eacute;e (MLR)) et techniques d'ensemble (poids & eacute;gal [EQ], Bates-Granger-BG), Granger-Ramanathan (GR), crit & egrave;re d'information d'Akaike (AIC) et crit & egrave;re d'information bay & eacute;sien (BIC)) sur la capacit & eacute; au champ (FC), le point de fl & eacute;trissement (WP) et la teneur en eau disponible (AWC) des sols. En cons & eacute;quence, une pr & eacute;cision de pr & eacute;diction plus & eacute;lev & eacute;e a & eacute;t & eacute; obtenue avec l'algorithme RF qu'avec l'algorithme d'apprentissage automatique des valeurs pour l'estimation des constantes d'humidit & eacute;. Les coefficients de d & eacute;termination (R2) obtenus pour la pr & eacute;diction de FC, WP et AWC au moyen des algorithmes RF & eacute;taient de 0,624, 0,759 et 0,641 respectivement. La MLR affichait le taux d'erreur le plus & eacute;lev & eacute;. Parmi les techniques d'ensemble, le GR a eu le plus de succ & egrave;s. Les valeurs du coefficient de corr & eacute;lation de concordance de Lin (LCCC) obtenues & agrave; partir de l'estimation de FC, WP et AWC avec le mod & egrave;le GR & eacute;taient de 0,801, 0,894 et 0,801, respectivement. Les valeurs de l'erreur quadratique moyenne (RMSE) et de l'erreur absolue moyenne (MAE) obtenues pour l'estimation de la teneur en eau disponible avec l'algorithme MLR & eacute;taient de 1,905 et 1,435 respectivement, alors que ces valeurs & eacute;taient de 1,173 et 0,767 respectivement, lorsque le mod & egrave;le GR a & eacute;t & eacute; utilis & eacute;. & Agrave; la suite de la pr & eacute;sente & eacute;tude, de meilleurs r & eacute;sultats pr & eacute;dictifs ont & eacute;t & eacute; obtenus avec des techniques d'ensemble au lieu d'& eacute;valuer les algorithmes s & eacute;par & eacute;ment.
引用
收藏
页数:12
相关论文
共 50 条
  • [41] COMPARISON OF STATISTICAL AND SOIL-MOISTURE MODELING TECHNIQUES IN A LONG-TERM STUDY OF WHEAT YIELD PERFORMANCE UNDER SEMI-ARID CONDITIONS
    LEWIN, J
    LOMAS, J
    JOURNAL OF APPLIED ECOLOGY, 1974, 11 (03) : 1081 - 1090
  • [42] Soil carbon prediction under different land uses by integration of remote sensing and machine learning algorithms in a semi-arid watershed, Iran
    Mohammad Tahmoures
    Samira Mesri
    Banafsheh Afrasiabi
    Afshin Honarbakhsh
    Ben Ingram
    Arabian Journal of Geosciences, 2023, 16 (1)
  • [43] Ensemble of optimised machine learning algorithms for predicting surface soil moisture content at a global scale
    Han, Qianqian
    Zeng, Yijian
    Zhang, Lijie
    Cira, Calimanut-Ionut
    Prikaziuk, Egor
    Duan, Ting
    Wang, Chao
    Szabo, Brigitta
    Manfreda, Salvatore
    Zhuang, Ruodan
    Su, Bob
    GEOSCIENTIFIC MODEL DEVELOPMENT, 2023, 16 (20) : 5825 - 5845
  • [44] Positive climate feedbacks of soil microbial communities in a semi-arid grassland
    Nie, Ming
    Pendall, Elise
    Bell, Colin
    Gasch, Caley K.
    Raut, Swastika
    Tamang, Shanker
    Wallenstein, Matthew D.
    ECOLOGY LETTERS, 2013, 16 (02) : 234 - 241
  • [45] Nitrous oxide emissions from a cropped soil in a semi-arid climate
    Barton, Louise
    Kiese, Ralf
    Gatter, David
    Butterbach-Bahl, Klaus
    Buck, Renee
    Hinz, Christoph
    Murphy, Daniel V.
    GLOBAL CHANGE BIOLOGY, 2008, 14 (01) : 177 - 192
  • [46] Comparison of field performance of multiple soil moisture sensors in a semi-arid rangeland
    Paige, Ginger B.
    Keefer, Timothy O.
    JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION, 2008, 44 (01): : 121 - 135
  • [47] A model coupling the effects of soil moisture and potential evaporation on the tree transpiration of a semi-arid larch plantation
    Li, Zhenhua
    Yu, Pengtao
    Wang, Yanhui
    Webb, Ashley A.
    He, Cong
    Wang, Yanbing
    Yang, Lili
    ECOHYDROLOGY, 2017, 10 (01)
  • [48] Precipitation and landscape related effects on soil moisture in semi-arid rangelands of Patagonia
    Coronato, FR
    Bertiller, MB
    JOURNAL OF ARID ENVIRONMENTS, 1996, 34 (01) : 1 - 9
  • [49] Spatial Variability of Surface Soil Moisture in Semi-Arid Region of Northern China
    Chen Jiongfeng
    Zhang Wanchang
    2008 INTERNATIONAL WORKSHOP ON EDUCATION TECHNOLOGY AND TRAINING AND 2008 INTERNATIONAL WORKSHOP ON GEOSCIENCE AND REMOTE SENSING, VOL 2, PROCEEDINGS,, 2009, : 356 - +
  • [50] Nitrogen and Soil Moisture Optimization for Tomato Crops in Semi-Arid Areas of Ethiopia
    Tesfaye, Abera
    Wondimu, Tatek
    Tezera, Ketema
    Ashemi, Gebeyehu
    Worku, Tigist
    Gurms, Aynalem
    AIR SOIL AND WATER RESEARCH, 2024, 17