Using phenology-based enhanced vegetation index and machine learning for soybean yield estimation in Paraná State, Brazil

Richetti, Jonathan; Judge, Jasmeet; Boote, Kenneth J.; Johann, Jerry Adriani; Uribe-Opazo, Miguel Ángel; Becker, Willyan Ronaldo; Paludo, Alex; Silva, Laíza Cavalcante de Albuquerque

doi:10.1117/1.jrs.12.026029

Cited by 22 publications

(15 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Gradient boosting machines provided stronger predictions in terms of LCCC and RMSE only once, in the flowering model using all variables. The superior results of RF are in agreement with other studies where RF provided higher predictive accuracy compared with other machine‐learning algorithms for predicting yield or other crop parameters (Kayad et al., 2019; Liang et al., 2015; Richetti et al., 2018; Wu et al., 2019). In the current study, the superior performance of RF compared with GBM is attributed to how each machine‐learning algorithm builds the optimal model.…”

Section: Discussionsupporting

confidence: 89%

Predicting within‐field cotton yields using publicly available datasets and machine learning

2021

View full text Add to dashboard Cite

Early detection of within-field yield variability for high-value commodity crops, such as cotton (Gossypium spp.), offers growers potential to improve decision-making, optimize yields, and increase profits. Over recent years, publicly available datasets have become increasingly available and at a resolution where within-field yield prediction is possible. However, the viability of using these datasets with machine learning to predict within-field cotton lint yield at key growth stages are largely unknown. This study was conducted on two cotton fields, located near Mungindi, New South Wales, Australia. Three years of yield data, soil, elevation, rainfall, and Landsat imagery were collected from each field. A total of 12 models were created using: (a) two machine learning algorithms: random forest (RF) and gradient boosting machines (GBM); (b) three growth stages: squaring, flowering, and boll-fill; and(c) two different amounts of variables: all variables and the optimal variables determined by a recursive feature elimination (RFE). Results showed a strong agreement between predicted and observed yields at flowering and boll-fill when more information was available. At flowering and boll-fill, root mean square error (RMSE) ranged between 0.15 and 0.20 t ha −1 and Lin's concordance correlation coefficient (LCCC) ranged between 0.50 and 0.66, with RF providing superior results in most cases. Models created using the optimal variables determined by the RFE provided similar results compared to using all variables, allowing greater model accuracy and resolution with targeted sampling. Overall, these findings indicate significant potential of publicly available datasets to predict within-field cotton yield and guide decisionmaking in-season.

show abstract

Section: Discussionsupporting

confidence: 89%

Predicting within‐field cotton yields using publicly available datasets and machine learning

2021

View full text Add to dashboard Cite

show abstract

“…It can also be found that R 2 and the adjusted R 2 for model validation using total production are smaller than that for model calibration using crop yield per unit area, which is reasonable because errors in the estimated total production include errors in both yield estimation and crop classification. Compared with previous crop yield estimation studies using machine learning methods [7,11,43] with the R 2 ranging from 0.2 to 0.8, the accuracy of the present crop yield estimations was acceptable.…”

Section: Resultscontrasting

confidence: 52%

“…The ANN has been successfully applied to yield estimation of various crops, such as maize [11], wheat [37], potato [38], melon [39] and grassland dry matter yield [40]. The RF method has also been used in crop yield estimation, especially for large areas of maize [7,41,42], soybean [43] and wheat [44]. Therefore, most available studies were on yield estimations of maize and wheat, but few studied yield estimations of sunflower, an important economic crop in arid regions of Northwest China.…”

Section: Introductionmentioning

confidence: 99%

Multi-Year Mapping of Major Crop Yields in an Irrigation District from High Spatial and Temporal Resolution Vegetation Index

Shang

2018

Sensors

View full text Add to dashboard Cite

Crop yield estimation is important for formulating informed regional and national food trade policies. The introduction of remote sensing in agricultural monitoring makes accurate estimation of regional crop yields possible. However, remote sensing images and crop distribution maps with coarse spatial resolution usually cause inaccuracy in yield estimation due to the existence of mixed pixels. This study aimed to estimate the annual yields of maize and sunflower in Hetao Irrigation District in North China using 30 m spatial resolution HJ-1A/1B CCD images and high accuracy multi-year crop distribution maps. The Normalized Difference Vegetation Index (NDVI) time series obtained from HJ-1A/1B CCD images was fitted with an asymmetric logistic curve to calculate daily NDVI and phenological characteristics. Eight random forest (RF) models using different predictors were developed for maize and sunflower yield estimation, respectively, where predictors of each model were a combination of NDVI series and/or phenological characteristics. We calibrated all RF models with measured crop yields at sampling points in two years (2014 and 2015), and validated the RF models with statistical yields of four counties in six years. Results showed that the optimal model for maize yield estimation was the model using NDVI series from the 120th to the 210th day in a year with 10 days’ interval as predictors, while that for sunflower was the model using the combination of three NDVI characteristics, three phenological characteristics, and two curve parameters as predictors. The selected RF models could estimate multi-year regional crop yields accurately, with the average values of root-mean-square error and the relative error of 0.75 t/ha and 6.1% for maize, and 0.40 t/ha and 10.1% for sunflower, respectively. Moreover, the yields of maize and sunflower can be estimated fairly well with NDVI series 50 days before crop harvest, which implicated the possibility of crop yield forecast before harvest.

show abstract

“…Gusso et al (2017) developed a satellite remote sensingbased procedure to estimate soybean production before crop harvest, with coefficients of determination ranging from 0.91 to 0.98. Richetti et al (2018) used machine learning algorithms with Enhanced Vegetation Index (EVI) for soybean in Brazil and obtained a mean error of 3.5 kg ha −1 , RMSD of 373 kg ha −1 , and Willmott's d of 0.85.…”

Section: Introductionmentioning

confidence: 99%

Crop Modeling With Less Data: The Fao Model for Soybean Yield Estimation

2021

Self Cite

View full text Add to dashboard Cite

Using phenology-based enhanced vegetation index and machine learning for soybean yield estimation in Paraná State, Brazil

Cited by 22 publications

References 24 publications

Predicting within‐field cotton yields using publicly available datasets and machine learning

Predicting within‐field cotton yields using publicly available datasets and machine learning

Multi-Year Mapping of Major Crop Yields in an Irrigation District from High Spatial and Temporal Resolution Vegetation Index

Crop Modeling With Less Data: The Fao Model for Soybean Yield Estimation

Contact Info

Product

Resources

About