Currently, Brazil is the leading producer of sugarcane in the world, with self-sufficiency in the use of ethanol as a biofuel, as well as being one of the largest suppliers of sugar to the world. This study aimed to develop a predictive model for sugarcane production based on data extracted from aerial imagery obtained from drones or satellites, allowing the precise tracking of plant development in the field. A model based on a semiparametric approach associated with the inverse Gaussian distribution applied to vegetation indices (VIs), such as the Normalized Difference Vegetation Index (NDVI) and Visible Atmospherically Resistant Index (VARI), was developed with data from drone images obtained from two field experiments with randomized replications and four sugarcane varieties. These experiments were performed under conditions identical to those applied by sugarcane farmers. Further, the model validation was carried out by scaling up the analyses with data extracted from Sentinel-2 images of several commercial sugarcane fields. Very often, in countries such as Brazil, sugarcane crops occupy extensive areas. Consequently, the development of tools capable of being operated remotely automatically benefits the management of this crop in the field by avoiding laborious and time-consuming sampling and by promoting the reduction of operation costs. The results of the model application in both sources of data, i.e., data from field experiments as well as the data from commercial fields, showed a suitable level of overlap between the data of predicted yield using VIs generated from drone and satellite images with the data of verified yield obtained by measuring the production of experiments and commercial fields, indicating that the model is reliable for forecasting productivity months before the harvest time.

This research investigates how to estimate sugarcane (Saccharum officinarum L.) yield at harvest by using an average satellite image time-series collected during the growth phase. This study aims to evaluate the effectiveness of various modeling approaches, including a heteroskedastic gamma regression model, Random Forest, and Artificial Neural Networks, in predicting sugarcane yield based on satellite-derived vegetation indices and environmental variables. Key covariates analyzed include sugarcane varieties, production cycles, accumulated precipitation during the growth phase, and the mean GNDVI vegetation index. The analysis was conducted in two locations over two consecutive growing seasons. The research emphasizes the integration of satellite data with advanced statistical and machine learning techniques to enhance yield prediction in agricultural systems, specifically focusing on sugarcane cultivation. The results indicate that the heteroskedastic gamma regression model outperformed the other methods in explaining yield variability, particularly in commercial sugarcane fields, achieving a Coefficient Determination ($R^2$) of 0.89. These findings highlight the potential of these models to support informed decision-making and optimize agricultural practices, providing valuable insights for precision farming. Overall, the results of this study represent an initial step toward developing more robust models for predicting sugarcane yield. Future work will involve incorporating additional variables to better assess the impacts of environmental stresses, such as high temperatures and water deficits, on the crop’s agronomic performance.

Accurate crop yield prediction is essential for sugarcane growers, as it enables them to predict harvested biomass, guiding critical decisions regarding acquiring agricultural inputs such as fertilizers and pesticides, the timing and execution of harvest operations, and cane field renewal strategies. This study is based on an experiment conducted by researchers from the Commonwealth Scientific and Industrial Research Organisation (CSIRO), who employed a UAV-mounted LiDAR and multispectral imaging sensors to monitor two sugarcane field trials subjected to varying nitrogen (N) fertilization regimes in the Wet Tropics region of Australia. The predictive performance of models utilizing multispectral features, LiDAR-derived features, and a fusion of both modalities was evaluated against a benchmark model based on the Normalized Difference Vegetation Index (NDVI). This work utilizes the dataset produced by this experiment, incorporating other regressors and features derived from those collected in the field. Typically, crop yield prediction relies on features derived from direct field observations, either gathered through sensor measurements or manual data collection. However, enhancing prediction models by incorporating new features extracted through regressions executed on the original dataset features can potentially improve predictive outcomes. These extracted features, nominated in this work as meta-features (MFs), extracted through regressions with different regressors on original features, and incorporated into the dataset as new feature predictors, can be utilized in further regression analyses to optimize crop yield prediction. This study investigates the potential of generating MFs as an innovation to enhance sugarcane crop yield predictions. MFs were generated based on the values obtained by different regressors applied to the features collected in the field, allowing for evaluating which approaches offered superior predictive performance within the dataset. The kNN meta-regressor outperforms other regressors because it takes advantage of the proximity of MFs, which was checked through a projection where the dispersion of points can be measured. A comparative analysis is presented with a projection based on the Uniform Manifold Approximation and Projection (UMAP) algorithm, showing that MFs had more proximity than the original features when projected, which demonstrates that MFs revealed a clear formation of well-defined clusters, with most points within each group sharing the same color, suggesting greater uniformity in the predicted values. Incorporating these MFs into subsequent regression models demonstrated improved performance, with ${\overline{R}}^2$ values higher than 0.9 for MF Grad Boost M3, MF GradientBoost M5, and all kNN MFs and reduced error margins compared to field-measured yield values. The ${\overline{R}}^2$ values obtained in this work ranged above 0.98 for the AdaBoost meta-regressor applied to MFs, which were obtained from kNN regression on five models created by the researchers of CSIRO, and around 0.99 for the kNN meta-regressor applied to MFs obtained from kNN regression on these five models.

Sugarcane ( Saccharum officinarum L.) is one of the largest crops in Brazil, and its productivity varies according to the environment and management practices adopted. In this study, tons of sugar per hectare (TSH) are estimated using a heteroscedastic gamma (GA) regression model, which considers several explanatory variables, one of which is the normalized difference green vegetation index (GNDVI), obtained from multispectral images in two locations over two consecutive growing seasons. The modeling considers regression structures in the parameters representing the mean and coefficient of variation, respectively. The results show that there is an influence of location, cultivar, cycle, accumulated precipitation, and GNDVI. To verify if the model is well-fitted to the data, the analysis of quantile residuals shows that the model is adequate. Therefore, the results indicate that heteroscedastic GA regression is an alternative model for predicting TSH and can assist in decision-making in sugarcane cultivation.

Brazil is the world’s largest producer of sugarcane (Saccharum officinarum), accounting for approximately 40% of global production, with the state of São Paulo responsible for more than half of the national output due to its high level of mechanization. Despite its economic importance, publicly available datasets integrating information on sugarcane yield and production environment remain scarce.