Abstract
Electrical resistivity (ER) can be used to assess soil water in the field. This study investigated the possibility of extending the use of ER to measure plant available soil water variables, i.e. available soil water (ASW), total transpirable SW (TTSW), and fraction of transpirable SW (FTSW) using a pedotransfer approach. In a vineyard, 224 electrical resistivity tomography (ERT) transects and 672 time domain reflectometry (TDR) soil water profiles were acquired over 2 years. Soil physical–chemical properties were measured on 73 soil samples from eight different sites. To estimate the amount of soil water available to plants, grapevine (Vitis vinifera L.) water status was monitored by means of leaf water potentials. A benchmark experiment was carried out to compare four machine-learning techniques: multivariate adaptive regression splines (MARS), k-nearest neighbours (KNN), random forest (RF), and gradient boosting machine (GBM). Model interpretation led to a deeper understanding of the relationships between electrical resistivity and soil properties when predicting soil water availability for the plant. The models assessed had good predictive performance and were therefore used to map ASW, TTSW and FTSW in the vineyard. ER coupled to machine-learning algorithms was shown to be a good proxy for quantification and visualisation of plant available soil water with low disturbance.
Similar content being viewed by others
Abbreviations
- Ψpd :
-
Pre-dawn leaf water potential
- ASW:
-
Available soil water
- ERV:
-
Electrical resistivity variation
- FERV:
-
Fraction of electrical resistivity variation
- FTSW:
-
Fraction of transpirable soil water
- GBM:
-
Gradient boosting machine
- KNN:
-
K-Nearest neighbours
- MARS:
-
Multiple adaptive regression splines
- RF:
-
Random forest
- SWHC:
-
Soil water holding capacity
- TTSW:
-
Total transpirable soil water
References
André, F., van Leeuwen, C., Saussez, S., Van Durmen, R., Bogaert, P., Moghadas, D., et al. (2012). High-resolution imaging of a vineyard in south of France using ground-penetrating radar, electromagnetic induction and electrical resistivity tomography. Journal of Applied Geophysics, 78, 113–122.
Andrenelli, M. C., Magini, S., Pellegrini, S., Perria, R., Vignozzi, N., & Costantini, E. A. C. (2013). The use of the ARP system to reduce the costs of soil survey for precision viticulture. Journal of Applied Geophysics, 99(SI), 24–34.
Attia Al Hagrey, S. (2007). Geophysical imaging of root-zone, trunk, and moisture heterogeneity. Journal of Experimental Botany, 58(4), 839.
Bates, D., Maechler, M., Bolker, B., Walker, S. (2014), lme4: Linear mixed-effects models using Eigen and S4. R package version 1.1-7.
Bickel, P., Diggle, P., Fienberg, S., Gather, U., Olkin, I., & Zeger, S. (2008). Statistical learning from a regression perspective. New York: Springer.
Breiman, L. (2001). Random Forests. Machine Learning, 45(1), 5–32.
Brillante, L., Bois, B., Mathieu, O., Bichet, V., Michot, D., & Lévêque, J. (2014a). Monitoring soil volume wetness in heterogeneous soils by electrical resistivity. A field based pedotransfer function. Journal of Hydrology, 516, 55–66.
Brillante, L., Bois, B., Mathieu, O., Lévêque, J. (2014b). Spatio-temporal analysis of grapevine water behaviour in hillslope vineyards. The example of Corton Hill, Burgundy. In: Proceedings of the X th International Terroir Congress – Tokaj-Eger - Hungary, ENCITV (pp. 124–129). Available online at: http://congresdesterroirs.org/articles/telecharger/203
Brillante, L., Gaiotti, F., Lovat, L., Vincenzi, S., Giacosa, S., Torchio, F., et al. (2015a). Investigating the use of gradient boosting machine, random forest and their ensemble to predict skin flavonoid content from berry physical-mechanical characteristics in wine grapes. Computers and Electronics in Agriculture, 117, 186–193.
Brillante, L., Mathieu, O., Bois, B., van Leeuwen, C., & Lévêque, J. (2015b). The use of soil electrical resistivity to monitor plant and soil water relationships in vineyards. SOIL, 1, 273–286. doi:10.5194/soil-1-273-2015.
Campbell, R. B., Bower, C. A., & Richards, L. A. (1948). Change of electrical conductivity with temperature and the relation of osmotic pressure to electrical conductivity and ion concentration for soil extracts. Soil Science Society of America Proceedings, 13, 66–69.
Celette, F., Ripoche, A., & Gary, C. (2010). WaLIS. A simple model to simulate water partitioning in a crop association: The example of an intercropped vineyard, Agricultural Water Management, 97(11), 1749–1759.
Costantini, E. A. C., Pellegrini, S., Bucelli, P., Storchi, P., Vignozzi, N., Barbetti, R., et al. (2009). Relevance of the Lin’s and Host hydropedological models to predict grape yield and wine quality. Hydrology and Earth System Sciences, 13(9), 1635–1648.
Dry, P., Loveys, B. R., Botting, D., & During, H. (1996). Effects of partial root-zone drying on grapevine vigour, yield, composition of fruit and use of water. In C. S. Stockley, A. N. Sas, R. S. Johnstone, & T. H. Lee (Eds.), Proceedings of the 9th Australian Wine Industry Technical Conference (pp. 126–131). Adelaide: Winetitles.
Eugster, M. J. A. (2011). Benchmark experiments. A tool for analyzing statistical learning algorithms. Ph.D. thesis, Ludwig-Maximilians-Universitat Munchen, Munich, Germany
Friedman, J. H. (1991). Multivariate adaptive regression splines. The Annals of Statistics, 19(1), 1–141.
Friedman, J. H. (2001). Greedy function approximation: A gradient boosting machine. The Annals of Statistics, 29(5), 1189–1232.
Goulet, E., & Barbeau, G. (2006). Contribution of soil electrical resistivity measurements to the studies on soil/grapevine water relations. Journal International des Sciences de la Vigne et du Vin, 40(2), 57–69.
Goutouly, J. P., Rousset, D., Perroud, H. (2006). Characterization of spatial and temporal soil water status in vineyard by DC resistivity measurements. In Proceedings of the VIth International Terroir Congress – Bordeaux-Montpellier - France (ENCITV) (pp. 292–297). Available online at: http://congresdesterroirs.org/articles/telecharger/605
Guilioni, L., & Lhomme, J. P. (2006). Modelling the daily course of capitulum temperature in a sunflower canopy. Agricultural and Forest Meteorology, 138(1–4), 258–272.
Hadzick, Z. Z., Guber, A. K., Pachepsky, Y. A., & Hill, R. L. (2011). Pedotransfer functions in soil electrical resistivity estimation. Geoderma, 164, 195–202.
Hastie, T., Tibshirani, R., & Friedman, J. (2009). The elements of statistical learning: data mining, inference, and prediction (2nd ed.). The Netherlands: Springer.
Hillel, D. (1998). Environmental soil physics. San Diego: Academic Press.
Hothorn, T., Bretz, F., & Westfall, P. (2008). Simultaneous inference in general parametric models. Biometrical Journal, 50(3), 346–363.
IUSS Working group, WRB. (2014). World reference base for soil resource 2014. International soil classification system for naming soils and creating legends for soil maps. Rome: FAO.
Kramer, P. J. (1944). Soil moisture in relation to plant growth. The Botanical Review, 10(9), 525–559.
Lacape, M. J., Wery, J., & Annerose, D. J. M. (1998). Relationships between plant and soil water status in five field-grown cotton (Gossypium hirsutum L.) cultivars. Field Crops Research, 57(1), 29–43.
Lebon, E., Dumas, V., Pieri, P., & Schultz, H. R. (2003). Modelling the seasonal dynamics of the soil water balance of vineyards. Functional Plant Biology, 30, 699–710. doi:10.1071/FP02222.
Lecoeur, J., & Guilioni, L. (1998). Rate of leaf production in response to soil water deficits in field pea. Field Crops Research, 57(3), 319–328.
Liaw, A., & Wiener, M. (2002). Classification and regression by randomForest. R News, 2(3), 18–22.
Loke, M. H. (2014). Tutorial: 2-D and 3-D electrical imaging surveys. Gelugor: Geotomosoft Solutions.
Loveys, B. R., Dry, P. R., Stoll, M., & Mc Carthy, M. G. (2000). Using plant physiology to improve the water use efficiency of horticultural crops. Acta Horticulturae, 537, 187–197.
Lu, Y., Equiza, M. A., Deng, X., & Tyree, M. T. (2010). Recovery of Populus tremuloides seedlings following severe drought causing total leaf mortality and extreme stem embolism. Physiologia Plantarum, 140(3), 246–257.
Michot, D., Benderitter, Y., Dorigny, A., Nicollaud, B., King, D., & Tabbagh, A. (2003). Spatial and temporal monitoring of soil water content with an irrigated corn crop cover using surface electrical resistivity tomography. Water Resources Research, 39(5), 1–20.
Milborrow, S. (2014), Earth: Multivariate adaptive regression spline models, Available online at: https://cran.r-project.org/web/packages/gbm. R package version: 2-4.
Pellegrino, A., Gozé, E., Lebon, E., & Wery, J. (2006). A model-based diagnosis tool to evaluate the water stress experienced by grapevine in field sites. European Journal of Agronomy, 25(1), 49–59.
Pellegrino, A., Lebon, A., Voltz, M., & Wery, J. (2005). Relationships between plant and soil water status in vine (Vitis vinifera L.). Plant and Soil, 266(1–2), 129–142.
Pinheiro, J. C., & Bates, M. D. (2000). Mixed-effects models in S and S-PLUS. New York: Springer.
R Core Team. (2013). R: a language and environment for statistical computing, R Development Core Team, R Foundation for Statistical Computing, Vienna, Austria, ISBN: 3-900051-07-0. Available online at http://www.R-project.org
Richards, L. A., & Wadleigh, C. H. (1952). Soil water and plant growth. In T. S. Byron (Ed.), Soil physical conditions and plant growth (Vol. 2, p. 491). Cambridge: Academic Press.
Ridgeway, G. (2013). gbm: Generalized boosted regression models. Available online at: https://cran.r-project.org/web/packages/earth, R package version 2.1.
Ritchie, J. T. (1981). Water dynamics in the soil-plant-atmosphere system. Plant and Soil, 58(1–3), 81–96.
Rossi, R., Amato, M., Bitella, G., Bochicchio, R., Ferreira Gomes, J. J., Lovelli, S., et al. (2011). Electrical resistivity tomography as a non-destructive method for mapping root biomass in an orchard. European Journal of Soil Science, 62(2), 206–215.
Sadras, V. O., & Schultz, H. R. (2012). Grapevine. In P. Steduto, T. Hsiao, E. Fereres, & D. Raes (Eds.), Crop yield response to water (p. 501). Rome: FAO.
Samouelian, A., Cousin, I., Tabbagh, A., Bruand, A., & Richard, G. (2005). Electrical resistivity survey in soil science: A review. Soil and Tillage Research, 83, 173–193.
Scholander, P. F., Bradstreet, E. D., Hemmingsen, E. A., & Hammel, H. T. (1965). Sap pressure in vascular plants. Negative hydrostatic pressure can be measured in plants. Science, 148(3668), 339–346.
Sinclair, T. R., Holbrook, N. M., & Zwieniecki, M. A. (2005). Daily transpiration rates of woody species on drying soil. Tree Physiology, 25(11), 1469–1472.
Srayeddin, I., & Doussan, C. (2009). Estimation of the spatial variability of root water uptake of maize and sorghum at the field scale by electrical resistivity tomography. Plant and Soil, 319(1–2), 185–207.
Stoll, M., Loveys, B., & Dry, P. (2000). Hormonal changes induced by partial rootzone drying of irrigated grapevine. Journal of Experimental Botany, 51(350), 1627–1634.
Veihmeyer, F. J., & Hendrickson, A. H. (1950). Soil moisture in relation to plant growth. Annual Review of Plant Physiology, 1, 285–305.
Venables, W. N., & Ripley, B. D. (2002). Modern applied statistics with S (4th ed.). New York: Springer.
Williams, L. E., Grimes, D. W., & Phene, C. J. (2010). The effects of applied water at various fractions of measured evapotranspiration on reproductive growth and water productivity of Thompson Seedless grapevines. Irrigation Science, 28(3), 233–243.
Acknowledgments
This work was funded by the Conseil Régional de Bourgogne and the Bureau Interprofessionnel des Vins de Bourgogne (BIVB). The authors wish to thank Ing. Boris Champy and the Domaine Latour for access to the vineyard; Dr. Carmela Chateau Smith for assistance in English; students Sarah De Ciantis, Celine Faivre-Primot, Thomas Marchal and Basile Pauthier for help in laboratory analysis and/or field-data acquisition. Authors wish to thank anonymous reviewers for useful comments that improved the quality of the manuscript.
Author information
Authors and Affiliations
Corresponding author
Appendix: Short description of applied algorithms
Appendix: Short description of applied algorithms
KNN
The KNN approach predicts new samples using the mean (or other summary statistics, such as the median) of the k-closest samples from the training test. Different metrics can be used to calculate the distance between the samples. The most common is the Euclidean distance, which was also used here. The predictive performance of this method depend on the dimension of k, which is the parameter to tune, and must be large enough to avoid overfitting but not so large as to underfit data; the model is also sensitive to irrelevant predictors.
MARS
The MARS model predicts the outcome by creating the best subset of piecewise linear models for each predictor; interaction between predictors can also be included (Friedman 1991). A piecewise linear model is a function composed of linear regressions having different slopes, joined together at knots or cutting points, to allow continuity of the function. The algorithm evaluates all the data as a possible knot, to split the overall model into groups over which a linear model is fitted. A knot is then retained only if the addition of a different group and its corresponding linear regression significantly reduces the overall error (classically, the RMSE). The tuning parameters for MARS are the degree of interaction between predictors and the number of terms (i.e. the number of linear models) in the final model. MARS is implemented in R through the earth package.
Random forest
Random forest (Breiman 2001) is an ensemble of decision trees. Trees are a rule modelling technique, where data are split into sub-samples and different relationships are used for each sub-sample to predict the outcome. Full grown trees have low bias but high variance, because small variations in the data can completely change the splitting rules. Variance reduction can be achieved by using an ensemble of unpruned trees, a forest, on bootstrap resamples of the original data set. Resampling is used to perturb the structure of data and then diversify the trees in the ensemble; the final prediction will be the average of all single tree predictions. To diversify even more the trees in the ensemble, random forest (Breiman 2001) limits the search of the best split to a random sample of the possible predictors. The dimensions of this sample (mtry) have to be tuned, together with the number of data in the final leaves of the trees and the number of trees in the forest.
Gradient boosting machine
The gradient boosting machine (Friedman 2001) also uses an ensemble of trees for prediction, but the building of each successive tree depends on the previous ones. Indeed, in order to minimise a loss function (here the squared error), successive trees are fitted to the gradient (the residuals) of the previous trees, and the new ones are added to the ensemble. The process continues for a number of iterations, which is a tuning parameter (the number of trees). Each iteration, however, does not use the total training dataset, but only a random fraction of it. And, unlike random forest, trees are not full grown, and the number of splits has to be tuned. To limit overfitting, the participation of each tree in the ensemble can be regularised through shrinkage: for each new tree, only a fraction of the newly predicted values is added to the ensemble.
Rights and permissions
About this article
Cite this article
Brillante, L., Bois, B., Mathieu, O. et al. Electrical imaging of soil water availability to grapevine: a benchmark experiment of several machine-learning techniques. Precision Agric 17, 637–658 (2016). https://doi.org/10.1007/s11119-016-9441-1
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11119-016-9441-1