Search Results (22)

The aim of this study is to determine the reliability of regular and spatial cross-validation methods in predicting subfield-scale maize yields using phenological measures derived by Sentinel-2. Three maize fields from eastern Croatia were monitored during the 2023 growing season, with high-resolution ground truth yield data collected using combine harvester sensors. Sentinel-2 time series were used to compute two vegetation indices, Enhanced Vegetation Index (EVI) and Wide Dynamic Range Vegetation Index (WDRVI). These features served as inputs for three machine learning models, including Random Forest (RF) and Bayesian Generalized Linear Model (BGLM), which were trained and evaluated using both regular and spatial 10-fold cross-validation. Results showed that spatial cross-validation produced a more realistic and conservative estimate of the performance of the model, while regular cross-validation overestimated predictive accuracy systematically because of spatial dependence among the samples. EVI-based models were more reliable than WDRVI, generating more accurate phenomenological fits and yield predictions across parcels. These results emphasize the importance of spatially explicit validation for subfield yield modeling and suggest that overlooking spatial structure can lead to misleading conclusions about model accuracy and generalizability. Full article

As one of the world’s leading grain producers, Brazil stands out in soybean and corn production. Accurate estimation of key crop phenological stages is essential for agricultural decision-making, especially considering Brazil’s vast territory, climatic diversity, and increasing frequency of extreme weather events. This study investigated the applicability of the NDVI, EVI, WDRVI, and NDWI, derived from Harmonized Landsat Sentinel-2, to identify crop sowing and harvest dates at the field scale. We extracted the vegetative peak from each vegetation index time series and identified the left and right inflection points around the peak to delineate the crop season. A double-logistic function and a derivative approach were applied to identify the Start of Season, Peak of Season, and End of Season. For both soybeans and corn, the RMSE ranged from 5 to 8 days for sowing dates, while for harvest dates it ranged from 6 to 15 days for corn. Despite these differences, all vegetation indices exhibited robust performance, with Spearman correlation values between 0.56 and 0.84. Our findings indicate that the use of different indices does not have a significant impact on the results, as long as the adjustment of temporal parameters for the phenological metrics is appropriate for each index. Full article

Cropland suitability calculations quantify natural suitability according to abiotic conditions, thus making them crucial for sustainable land management. However, since ground-truth yield data are extremely scarce, there is a need to improve knowledge on the optimal proxy metric from satellite imagery, which represents cropland suitability and enables global applicability. This study evaluated four frequently used vegetation indices from Sentinel-2 image time-series (normalized difference vegetation index, enhanced vegetation index, enhanced vegetation index 2, and wide dynamic range vegetation index) with three phenology metrics for correlation analysis with maize and soybean yield. Four years (2019–2022) in two study areas (Iowa and Illinois) were utilized in this research, and 1000 ground-truth crop yield samples were created for each combination of study year and area. The combination of wide dynamic range vegetation index (WDRVI) and maximum vegetation index phenology metric (MAX) was an optimal proxy for maize yield prediction, while enhanced vegetation index 2 (EVI2) and MAX produced the highest correlation for soybean, producing Pearson’s correlation coefficient means of 0.506 and 0.519, respectively. This study improved our knowledge of the optimal proxy metric for cropland suitability by combining multiple large ground-truth crop yield datasets with 30 m spatial resolution satellite imagery, which can be further improved with the use of novel vegetation indices with improved resistance to a saturation effect. Full article

Traditional disease severity monitoring is subjective and inefficient. This study employs a Parrot multispectral sensor mounted on an unmanned aerial vehicle (UAV) to apply machine learning algorithms, such as random forest, for categorizing defoliation levels in R7-stage soybean plants. This research assesses the effectiveness of vegetation indices, spectral bands, and relative vegetation cover as input parameters, demonstrating that machine learning approaches combined with multispectral imagery can provide a more accurate and efficient assessment of Asian soybean rust in commercial soybean fields. The random forest algorithm exhibited satisfactory classification performance when compared to recent studies, achieving accuracy, precision, recall, F1-score, specificity, and AUC values of 0.94, 0.92, 0.92, 0.92, 0.97, and 0.97, respectively. The input variables identified as most important for the classification model were the WDRVI and MPRI indices, the red-edge and NIR bands, and relative vegetation cover, with the highest Gini importance index. Full article

Predicting crop yield at preharvest is pivotal for agricultural policy and strategic decision making. Despite global agricultural targets, labour-intensive surveys for yield estimation pose challenges. Using unmanned aerial vehicle (UAV)-based multispectral sensors, this study assessed crop phenology and biotic stress conditions using various spectral vegetation indices. The goal was to enhance the accuracy of predicting key agricultural parameters, such as leaf area index (LAI), soil and plant analyser development (SPAD) chlorophyll, and grain yield of maize. The study’s findings demonstrate that during the kharif season, the wide dynamic range vegetation index (WDRVI) showcased superior correlation coefficients (R), coefficients of determination (R²), and the lowest root mean square errors (RMSEs) of 0.92, 0.86, and 0.14, respectively. However, during the rabi season, the atmospherically resistant vegetation index (ARVI) achieved the highest R and R² and the lowest RMSEs of 0.83, 0.79, and 0.15, respectively, indicating better accuracy in predicting LAI. Conversely, the normalised difference red-edge index (NDRE) during the kharif season and the modified chlorophyll absorption ratio index (MCARI) during the rabi season were identified as the predictors with the highest accuracy for SPAD chlorophyll prediction. Specifically, R values of 0.91 and 0.94, R² values of 0.83 and 0.82, and RMSE values of 2.07 and 3.10 were obtained, respectively. The most effective indices for LAI prediction during the kharif season (WDRVI and NDRE) and for SPAD chlorophyll prediction during the rabi season (ARVI and MCARI) were further utilised to construct a yield model using stepwise regression analysis. Integrating the predicted LAI and SPAD chlorophyll values into the model resulted in higher accuracy compared to individual predictions. More exactly, the R² values were 0.51 and 0.74, while the RMSE values were 9.25 and 6.72, during the kharif and rabi seasons, respectively. These findings underscore the utility of UAV-based multispectral imaging in predicting crop yields, thereby aiding in sustainable crop management practices and benefiting farmers and policymakers alike. Full article

Globally, rice is one of the most important staple food crops. The most significant metric for evaluating the rice growth and productivity is the Leaf Area Index (LAI), which can be effectively monitored using remote sensing data. Hyperspectral remote sensing provides contiguous bands at narrow wavelengths for mapping LAI at various rice phenological stages, and it is functionally related to canopy spectral reflectance. Hyperspectral signatures for different phases of rice crop growth was recorded using Airborne Visible Near-Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) along with corresponding ground based observations. Ground-based hyperspectral canopy spectral reflectance measurements were recorded with FieldSpec 3 Hi-Res spectroradiometer (ASD Inc., Forsyth County, GA, USA; spectral range: 350–2500 nm) and LAI data from 132 farmer’s fields in Southern India. Among 29 hyperspectral vegetation indices tested, 8 were found promising for mapping rice LAI at various phenological stages. Among all the growth stages, the elongation stage was the most accurately estimated using vegetation indices that exhibited a significant correlation with the airborne hyperspectral reflectance. The validation of hyperspectral vegetation indices revealed that the best fit model for estimating rice LAI was mND₇₀₅ (red-edge, blue, and NIR bands) at seedling and elongation, SAVI (red and NIR bands) at tillering and WDRVI (red and NIR bands) at booting stage. Full article

Ensuring food security in precision agriculture demands early prediction of corn yield in the USA at international, regional, and local levels. Accurate corn yield estimation can play a crucial role in averting famine by offering insights into food availability during the growing season. To address this, we propose a Concatenate-based 2D-CNN-BILSTM model that integrates Sentinel-1, Sentinel-2, and Soil GRIDS (global gridded soil information) data for corn yield estimation in Iowa State from 2018 to 2021. This approach utilizes Sentinel-2 features, including spectral bands (Blue, Green, Red, Red Edge 1/2/3, NIR, n-NIR, and SWIR 1/2), and vegetation indices (NDVI, LSWI, DVI, RVI, WDRVI, SAVI, VARIGREEN, and GNDVI), alongside Sentinel 1 features (VV, VH, difference VV, and VH, and RVI), and soil data (Silt, Clay, Sand, CEC, and pH) as initial inputs. To extract high-level features from this data each month, a dedicated 2D-CNN was designed. This 2D-CNN concatenates high-level features from the previous month with low-level features of the subsequent month, serving as input features for the model. Additionally, to incorporate single-time soil data features, another 2D-CNN was implemented. Finally, high-level features from soil, Sentinel-1, and Sentinel-2 data were concatenated and fed into a BILSTM layer for accurate corn yield prediction. Comparative analysis against random forest (RF), Concatenate-based 2D-CNN, and 2D-CNN models, using metrics like RMSE, MAE, MAPE, and the Index of Agreement, revealed the superiority of our model. It achieved an Index of Agreement of 84.67% with an RMSE of 0.698 t/ha. The Concatenate-based 2D-CNN model also performed well with an RMSE of 0.799 t/ha and an Index of Agreement of 72.71%. The 2D-CNN model followed closely with an RMSE of 0.834 t/ha and an Index of Agreement of 69.90%. In contrast, the RF model lagged with an RMSE of 1.073 t/ha and an Index of Agreement of 69.60%. Integration of Sentinel 1–2 and Soil-GRIDs data with the Concatenate-based 2D-CNN-BILSTM model significantly improved accuracy. Combining soil data with Sentinel 1–2 features reduced the RMSE by 16 kg and increased the Index of Agreement by 2.59%. This study highlighted the potential of advanced machine learning (ML)/deep learning (DL) models in achieving precise and reliable predictions, which could support sustainable agricultural practices and food-security initiatives. Full article

New agronomic opportunities for more informed agricultural decisions and enhanced crop management have been made possible by drone-based near-ground remote sensing. Obtaining precise non-destructive information regarding crop biophysical characteristics at spatial and temporal scales is now possible. Drone-mounted multispectral and thermal sensors were used to assess crop phenology, condition, and stress by profiling spectral vegetation indices in crop fields. In this study, vegetation indices, viz., Atmospherically Resistant Vegetation Index (ARVI), Modified Chlorophyll Absorption Ratio Index (MCARI), Wide Dynamic Range Vegetation Index (WDRVI), Normalized Red–Green Difference Index (NGRDI), Excess Green Index (ExG), Red–Green Blue Vegetation Index (RGBVI), and Visible Atmospherically Resistant Index (VARI) were generated. Furthermore, Pearson correlation analysis showed a better correlation between WDRVI and VARI with LAI (R = 0.955 and R = 0.982) ground truth data. In contrast, a strong correlation (R = 0.931 and R = 0.844) was recorded with MCARI and NGRDI with SPAD chlorophyll ground truth data. Then, the best-performing indices, WDRVI and MCARI in cotton, and VARI and NGRDI in rice, were further used to generate the yield model. This study for determining LAI and chlorophyll shows that high spatial resolution drone imageries are accurate and fast. As a result, finding out the LAI and chlorophyll and how they affect crop yield at a regional scale is helpful. The widespread use of unmanned aerial vehicles (UAV) and yield prediction were technical components of large-scale precision agriculture. Full article

The fraction of absorbed photosynthetically active radiation (FPAR), which represents the capability of vegetation-absorbed solar radiation to accumulate organic matter, is a crucial indicator of photosynthesis and vegetation growth status. Although a simplified semi-empirical FPAR estimation model was easily obtained using vegetation indices (VIs), the sensitivity and robustness of VIs and the optimal inversion method need to be further evaluated and developed for canola FPAR retrieval. The objective of this study was to identify the robust hybrid inversion model for estimating the winter canola FPAR. A field experiment with different sow dates and densities was conducted over two growing seasons to obtain canola FPARs. Moreover, 29 VIs, two machine learning algorithms and the PROSAIL model were incorporated to establish the FPAR inversion model. The results indicate that the OSAVI, WDRVI and mSR had better capability for revealing the variations of the FPAR. Three parameters of leaf area index (LAI), solar zenith angle (SZA) and average leaf inclination angle (ALA) accounted for over 95% of the total variance in the FPARs and OSAVI exhibited a greater resistance to changes in the leaf and canopy parameters of interest. The hybrid inversion model with an artificial neural network (ANN-VIs) performed the best for both datasets. The optimal hybrid inversion model of ANN-OSAVI achieved the highest performance for canola FPAR retrieval, with R² and RMSE values of 0.65 and 0.051, respectively. Finally, the work highlights the usefulness of the radiation transfer model (RTM) in quantifying the crop canopy FPAR and demonstrates the potential of hybrid model methods for retrieving the canola FPAR at each growth stage. Full article

The water cycle around the globe is significantly impacted by the moisture in the soil. However, finding a quick and practical model to cope with the enormous amount of data is a difficult issue for remote sensing practitioners. The traditional methods of measuring soil moisture are inefficient at large sizes, which can be replaced by remote sensing techniques for obtaining soil moisture. While determining the soil moisture, the low return frequency of satellites and the lack of images pose a severe challenge to the current remote sensing techniques. Therefore, this paper suggested a novel technique for Soil Moisture Retrieval. In the initial phase, image acquisition is made. Then, VI indexes (NDVI, GLAI, Green NDVI (GNDVI), and WDRVI features) are derived. Further, an improved Water Cloud Model (WCM) is deployed as a vegetation impact rectification scheme. Finally, soil moisture retrieval is determined by the hybrid model combining Deep Max Out Network (DMN) and Bidirectional Gated Recurrent Unit (Bi-GRU) schemes, whose outputs are then passed on to enhanced score level fusion that offers final results. According to the results, the RMSE of the Hybrid Classifier (Bi-GRU and DMN) method was lower (0.9565) than the RMSE of the Hybrid Classifier methods. The ME values of the HC (Bi-GRU and DMN) were also lower (0.728697) than those of the HC methods without the vegetation index, the HC methods without the presence of water clouds, and the HC methods with traditional water clouds. In comparison to HC (Bi-GRU and DMN), the HC method without vegetation index has a lower error of 0.8219 than the HC method with standard water cloud and the HC method without water cloud. Full article

Banana Fusarium wilt (BFW) is a devastating disease with no effective cure methods. Timely and effective detection of the disease and evaluation of its spreading trend will help farmers in making right decisions on plantation management. The main purpose of this study was to find the spectral features of the BFW-infected canopy and build the optimal BFW classification models for different stages of infection. A RedEdge-MX camera mounted on an unmanned aerial vehicle (UAV) was used to collect multispectral images of a banana plantation infected with BFW in July and August 2020. Three types of spectral features were used as the inputs of classification models, including three-visible-band images, five-multispectral-band images, and vegetation indices (VIs). Four supervised methods including Support Vector Machine (SVM), Random Forest (RF), Back Propagation Neural Networks (BPNN) and Logistic Regression (LR), and two unsupervised methods including Hotspot Analysis (HA) and Iterative Self-Organizing Data Analysis Technique Algorithm (ISODATA) were adopted to detect the BFW-infected canopies. Comparing to the healthy canopies, the BFW-infected canopies had higher reflectance in the visible region, but lower reflectance in the NIR region. The classification results showed that most of the supervised and unsupervised methods reached excellent accuracies. Among all the supervised methods, RF based on the five-multispectral-band was considered as the optimal model, with higher overall accuracy (OA) of 97.28% and faster running time of 22 min. For the unsupervised methods, HA reached high and balanced OAs of more than 95% based on the selected VIs derived from the red and NIR band, especially for WDRVI, NDVI, and TDVI. By comprehensively evaluating the classification results of different metrics, the unsupervised method HA was recommended for BFW recognition, especially in the late stage of infection; the supervised method RF was recommended in the early stage of infection to reach a slightly higher accuracy. The results found in this study could give advice for banana plantation management and provide approaches for plant disease detection. Full article

Remote sensing-based crop monitoring has evolved unprecedentedly to supply multispectral imagery with high spatial-temporal resolution for the assessment of crop evapotranspiration (ETc). Several methodologies have shown a high correlation between the Vegetation Indices (VIs) and the crop coefficient (Kc). This work analyzes the estimation of the crop coefficient (Kc) as a spectral function of the product of two variables: VIs and green vegetation cover fraction (f_v). Multispectral images from experimental maize plots were classified to separate pixels into three classes (vegetation, shade and soil) using the OBIA (Object Based Image Analysis) approach. Only vegetation pixels were used to estimate the VIs and f_v variables. The spectral Kc_fv:VI models were compared with Kc based on Cumulative Growing Degree Days (CGDD) (Kc-c_GDD). The maximum average values of Normalized Difference Vegetation Index (NDVI), WDRVI, amd EVI2 indices during the growing season were 0.77, 0.21, and 1.63, respectively. The results showed that the spectral Kc_fv:VI model showed a strong linear correlation with Kc-c_GDD (R² > 0.80). The model precision increases with plant densities, and the Kc_fv:NDVI with 80,000 plants/ha had the best fitting performance (R² = 0.94 and RMSE = 0.055). The results indicate that the use of spectral models to estimate Kc based on high spatial and temporal resolution UAV-images, using only green pixels to compute VI and f_v crop variables, offers a powerful and simple tool for ETc assessment to support irrigation scheduling in agricultural areas. Full article

Corn yields vary spatially and temporally in the plots as a result of weather, altitude, variety, plant density, available water, nutrients, and planting date; these are the main factors that influence crop yield. In this study, different multispectral and red-green-blue (RGB) vegetation indices were analyzed, as well as the digitally estimated canopy cover and plant density, in order to estimate corn grain yield using a neural network model. The relative importance of the predictor variables was also analyzed. An experiment was established with five levels of nitrogen fertilization (140, 200, 260, 320, and 380 kg/ha) and four replicates, in a completely randomized block design, resulting in 20 experimental polygons. Crop information was captured using two sensors (Parrot Sequoia_4.9, and DJI FC6310_8.8) mounted on an unmanned aerial vehicle (UAV) for two flight dates at 47 and 79 days after sowing (DAS). The correlation coefficient between the plant density, obtained through the digital count of corn plants, and the corn grain yield was 0.94; this variable was the one with the highest relative importance in the yield estimation according to Garson’s algorithm. The canopy cover, digitally estimated, showed a correlation coefficient of 0.77 with respect to the corn grain yield, while the relative importance of this variable in the yield estimation was 0.080 and 0.093 for 47 and 79 DAS, respectively. The wide dynamic range vegetation index (WDRVI), plant density, and canopy cover showed the highest correlation coefficient and the smallest errors (R = 0.99, mean absolute error (MAE) = 0.028 t ha⁻¹, root mean square error (RMSE) = 0.125 t ha⁻¹) in the corn grain yield estimation at 47 DAS, with the WDRVI index and the density being the variables with the highest relative importance for this crop development date. For the 79 DAS flight, the combination of the normalized difference vegetation index (NDVI), normalized difference red edge (NDRE), WDRVI, excess green (EXG), triangular greenness index (TGI), and visible atmospherically resistant index (VARI), as well as plant density and canopy cover, generated the highest correlation coefficient and the smallest errors (R = 0.97, MAE = 0.249 t ha⁻¹, RMSE = 0.425 t ha⁻¹) in the corn grain yield estimation, where the density and the NDVI were the variables with the highest relative importance, with values of 0.295 and 0.184, respectively. However, the WDRVI, plant density, and canopy cover estimated the corn grain yield with acceptable precision (R = 0.96, MAE = 0.209 t ha⁻¹, RMSE = 0.449 t ha⁻¹). The generated neural network models provided a high correlation coefficient between the estimated and the observed corn grain yield, and also showed acceptable errors in the yield estimation. The spectral information registered through remote sensors mounted on unmanned aerial vehicles and its processing in vegetation indices, canopy cover, and plant density allowed the characterization and estimation of corn grain yield. Such information is very useful for decision-making and agricultural activities planning. Full article

Accurately estimating grassland carbon stocks is important in assessing grassland productivity and the global carbon balance. This study used the regression kriging (RK) method to estimate grassland carbon stocks in Northeast China based on Landsat8 operational land imager (OLI) images and five remote sensing variables. The normalized difference vegetation index (NDVI), the wide dynamic range vegetation index (WDRVI), the chlorophyll index (CI), Band6 and Band7 were used to build the RK models separately and to explore their capabilities for modeling spatial distributions of grassland carbon stocks. To explore the different model performances for typical grassland and meadow grassland, the models were validated separately using the typical steppe, meadow steppe or all-steppe ground measurements based on leave-one-out crossvalidation (LOOCV). When the results were validated against typical steppe samples, the Band6 model showed the best performance (coefficient of determination (R²) = 0.46, mean average error (MAE) = 8.47%, and root mean square error (RMSE) = 10.34 gC/m²) via the linear regression (LR) method, while for the RK method, the NDVI model showed the best performance (R² = 0.63, MAE = 7.04 gC/m², and RMSE = 8.51 gC/m²), which were much higher than the values of the best LR model. When the results were validated against the meadow steppe samples, the CI model achieved the best estimation accuracy, and the accuracy of the RK method (R² = 0.72, MAE = 8.09 gC/m², and RMSE = 9.89 gC/m²) was higher than that of the LR method (R² = 0.70, MAE = 8.99 gC/m², and RMSE = 10.69 gC/m²). Upon combining the results of the most accurate models of the typical steppe and meadow steppe, the RK method reaches the highest model accuracy of R² = 0.69, MAE = 7.40 gC/m², and RMSE = 9.01 gC/m², while the LR method reaches the highest model accuracy of R² = 0.53, MAE = 9.20 gC/m², and RMSE = 11.10 gC/m². The results showed an improved performance of the RK method compared to the LR method, and the improvement in the accuracy of the model is mainly attributed to the enhancement of the estimation accuracy of the typical steppe. In the study region, the carbon stocks showed an increasing trend from west to east, the total amount of grassland carbon stock was 79.77 × 10⁴ Mg C, and the mean carbon stock density was 47.44 gC/m². The density decreased in the order of temperate meadow steppe, lowland meadow steppe, temperate typical steppe, and sandy steppe. The methodology proposed in this study is particularly beneficial for carbon stock estimates at the regional scale, especially for countries such as China with many grassland types. Full article

Leaf area per unit surface (LAI—leaf area index) is a valuable parameter to assess vine vigour in several applications, including direct mapping of vegetative–reproductive balance (VRB). Normalized difference vegetation index (NDVI) has been successfully used to assess the spatial variability of estimated LAI. However, sometimes NDVI is unsuitable due to its lack of sensitivity at high LAI values. Moreover, the presence of hail protection with Grenbiule netting also affects incident light and reflection, and consequently spectral response. This study analyses the effect of protective netting in the LAI–NDVI relationship and, using NDVI as a reference index, compares several indices in terms of accuracy and sensitivity using linear and logarithmic models. Among the indices compared, results show NDVI to be the most accurate, and ratio vegetation index (RVI) to be the most sensitive. The wide dynamic range vegetation index (WDRVI) presented a good balance between accuracy and sensitivity. Soil-adjusted vegetation index 2 (SAVI₂) appears to be the best estimator of LAI with linear models. Logarithmic models provided higher determination coefficients, but this has little influence over the normal range of LAI values. A similar NDVI–LAI relationship holds for protected and unprotected canopies in initial vegetation stages, but different functions are preferable once the canopy is fully developed, in particular, if tipping is performed. Full article