Site-specific early season potato yield forecast by neural network in Eastern Canada

Abstract

Deterministic potato (Solanum tuberosum L.) growth models hardly rely on driving seasonal field variables that directly characterize spatial variation of plant growth. For example, the SUBSTOR model computes the leaf area index (LAI) as an auxiliary variable from meteorological conditions and soil properties. Empirical models may account for seasonal LAI functions and accurately predict potato yield. The objective was to evaluate multiple linear regression (MLR) and neural networks (NN) as predictive models of potato yield. Using data from several replicated on-farm experiments conducted over 3 years, model performance was evaluated for their capacity to forecast tuber yields 9, 10 and 11 weeks before harvest compared to SUBSTOR. A 3-input NN using LAI functions and cumulative rainfall yielded the most accurate estimations and forecasts of tuber yields. This NN showed that tuber yield of contrasting zones was mostly a function of meteorological conditions prevailing during the first 5-8 weeks after planting. Subsequent development of tubers was essentially controlled by biomass allocation to tubers. The NN models were more coherent than MLR and SUBSTOR for two reasons: (1) the use of seasonal LAI directly as input rather than computed as an auxiliary variable and (2) the non-linearity of the modeling process resulting in more accurate estimation of the temporal discontinuities of potato tuber growth. This model showed potential for application in precision agriculture by accounting for temporal and spatial real-time climatic and crop data.

Predictive ability of machine learning methods for massive crop yield prediction

“…This characteristic is very useful to model complex non-linear behaviors, such as a function for crop yield prediction. ML techniques most successfully applied to CYP have been M5-Prime regression trees (Wang & Witten, 1997;Frausto-Solís et al, 2009;Marinković et al, 2009;Ruß & Kruse, 2010), artificial neural networks (Liu et al, 2001;Drummond et al, 2003;Safa et al, 2004;Fortin et al, 2011), support vector regression (Ruß, 2009) and k-nearest neighbor (Zhang et al, 2010). However, no comparisons covering all the aforementioned techniques have been made for a large amount of crops.…”

“…Multiple linear regression and other classical linear methods have been compared to CYP problem (Sudduth et al, 1996;Drummond et al, 2003;Fortin et al, 2011). In contrast to previous works, this paper builds the MLR models using the best attribute subset, which improves the models' predictive accuracy.…”

“…Machine Learning (ML) techniques are based on non-parametric and semi-parametric structures, with validation relying on prediction accuracy (Breiman, 2001). Regression trees (Roel & Plant, 2004), artificial neural networks (Liu et al, 2001;Drummond et al, 2003;Irmak et al, 2006;Fortin et al, 2011) and support vector regression (Jaikla et al, 2008;Ruß, 2009) are common ML techniques applied for CYP purposes. Previous works suggest that data-driven models have better adaptability for cropping planning than crop growth models due to their friendly implementation and performance (Safa et al, 2004;Irmak et al, 2006).…”

“…Some comparisons among regression models for CYP have been made, looking for the most accurate technique. Drummond et al (2003) and Fortin et al (2011) have compared classical statistical models against Artif icial Neural Networks (ANNs). Ruß (2009) compared ANNs, regression trees and support vector regression.…”

“…Ruß (2009) compared ANNs, regression trees and support vector regression. Despite the high site-dependency, neural networks have been widely recognized as robust models, and they have obtained good results for CYP (Liu et al, 2001;Fortin et al, 2011). On the other hand, support vector regression models have proven to be more accurate than ANN and regression trees for some crop datasets (Ruß, 2009).…”