Forecasting carrot yield with optimal timing of Sentinel 2 image acquisition

Abstract

Accurate, non-destructive forecasting of carrot yield is difficult due to its subterranean growing habit. Furthermore, the timing of forecasting usually occurs when the crop is mature, limiting the opportunity to implement alternative management decisions to improve yield (during the growing season). This study aims to improve the accuracy of carrot yield forecasting by exploring time series and multivariate approaches. Using Sentinel-2 satellite imagery in three Australian vegetable regions, we established a time series of carrot phenological stages (PhS) from ‘days after sowing’ (DAS) to enhance prediction timing. Numerous vegetation indices (VIs) were analyzed to derive temporal growth patterns. Correlations with yield at different PhS were established. Although the average root yield (t ha−1) did not significantly differ across the regions, the temporal VI signatures, indicating different regional crop growth trends, did vary as well as the PhS at when the maximum correlation with yield occurred ($$PhS_{{R2_{max} }} )$$ P h S R 2 max ) with two of the regions producing a delayed $$PhS_{{R2_{max} }}$$ P h S R 2 max (i.e. 90–130 DAS). The best multivariate model was identified at 70 DAS, extending the forecasting window before harvest between 20 to 60 days. The performance of this model was validated with new crops producing an average error of 16.9 t ha−1 (27% of total yield). These results demonstrate the potential of the model at such early stage under varying growing conditions offering growers and stakeholders the chance to optimize farming practices, make informed decisions on selling, harvesting, and labor planning, and adopt precision agriculture methods.

AI-Based Prediction of Carrot Yield and Quality on Tropical Agriculture

“…Root sampling consisted of two data collection periods: 82 and 116 DAS, with 50 points at each data collection in both experimental areas (total: 200 sampling points). These periods were chosen according to the best time for crop modeling according to Suarez et al [13], corresponding to the full radial filling (82 DAS) and the date before crop harvesting (116 DAS). The climatic conditions were 1600 degree-days and 2260 degree-days (base temperature of 3 • C) for 82 and 116 DAS, respectively (Figure 2).…”

“…Different predictive approaches to crop parameters leveraging remote sensing (RS) and artificial intelligence (AI) are being explored in the literature [11][12][13]. RS involves collecting data on the Earth's surface using sensors installed on satellites or remotely piloted aircraft to monitor large areas at low costs [14], control in-season weeds, identify the health of the crop, and predict crop yield in a non-destructive way.…”

“…The accuracy of crop yield prediction varies depending on the climatic conditions during the development of the plants. The correlation between VIs and yield changes throughout the crop cycle, so it would not be ideal to only use a fixed period to model crop yield through remote sensing [12][13][14]. Monitoring geocarpic crops, such as carrots, is more complex because the organ of interest is established underground [15,16], and its extraction is necessary for sampling [17].…”

Optimal Timing of Carrot Crop Monitoring and Yield Assessment Using Sentinel-2 Images: A Machine-Learning Approach

“…Schauberger et al [19] reported the extent of studies on the yield forecasting of horticultural crops, while studies by Suarez et al [20,21] described the importance of carrot yield forecasting and explored satellite images used for carrot yield prediction. Wei [22] generated carrot yield maps using planetScope hyperspectral data with a high accuracy (R 2 = 0.68-0.82), and a study by de Lima Silva [23] achieved a good carrot predicted yield accuracy (R 2 = 0.68) using planetScope CubeSat Platform.…”

“…[34][35][36]. Examples include sunflower [37], sugarcane [38], rice [39], corn [40], wheat [41], carrot [20][21][22], etc. The ML algorithms, including neural network (NN), stacked autoencoder (SAE), recurrent NN (RNN), graph NN (GNN), and restricted Boltzmann machine (RBM), have been widely used in agricultural applications [32].…”