Early Yield Prediction Research Articles

A Deep–Learning Network for Wheat Yield Prediction Combining Weather Forecasts and Remote Sensing Data

Accurately predicting winter wheat yield before harvest could greatly benefit decision-makers when making management decisions. In this study, we utilized weather forecast (WF) data combined with Sentinel-2 data to establish the deep-learning network and achieved an in-season county-scale wheat yield prediction in China’s main wheat-producing areas. We tested a combination of short-term WF data from the China Meteorological Administration to predict in-season yield at different forecast lengths. The results showed that explicitly incorporating WF data can improve the accuracy in crop yield predictions [Root Mean Square Error (RMSE) = 0.517 t/ha] compared to using only remote sensing data (RMSE = 0.624 t/ha). After comparing a series of WF data with different time series lengths, we found that adding 25 days of WF data can achieve the highest yield prediction accuracy. Specifically, the highest accuracy (RMSE = 0.496 t/ha) is achieved when predictions are made on Day of The Year (DOY) 215 (40 days before harvest). Our study established a deep-learning model which can be used for early yield prediction at the county level, and we have proved that weather forecast data can also be applied in data-driven deep-learning yield prediction tasks.

Multi temporal multispectral UAV remote sensing allows for yield assessment across European wheat varieties already before flowering.

High throughput field phenotyping techniques employing multispectral cameras allow extracting a variety of variables and features to predict yield and yield related traits, but little is known about which types of multispectral features are optimal to forecast yield potential in the early growth phase. In this study, we aim to identify multispectral features that are able to accurately predict yield and aid in variety classification at different growth stages throughout the season. Furthermore, we hypothesize that texture features (TFs) are more suitable for variety classification than for yield prediction. Throughout 2021 and 2022, a trial involving 19 and 18 European wheat varieties, respectively, was conducted. Multispectral images, encompassing visible, Red-edge, and near-infrared (NIR) bands, were captured at 19 and 22 time points from tillering to harvest using an unmanned aerial vehicle (UAV) in the first and second year of trial. Subsequently, orthomosaic images were generated, and various features were extracted, including single-band reflectances, vegetation indices (VI), and TFs derived from a gray level correlation matrix (GLCM). The performance of these features in predicting yield and classifying varieties at different growth stages was assessed using random forest models. Measurements during the flowering stage demonstrated superior performance for most features. Specifically, Red reflectance achieved a root mean square error (RMSE) of 52.4g m-2 in the first year and 64.4g m-2 in the second year. The NDRE VI yielded the most accurate predictions with an RMSE of 49.1g m-2 and 60.6g m-2, respectively. Moreover, TFs such as CONTRAST and DISSIMILARITY displayed the best performance in predicting yield, with RMSE values of 55.5g m-2 and 66.3g m-2 across the two years of trial. Combining data from different dates enhanced yield prediction and stabilized predictions across dates. TFs exhibited high accuracy in classifying low and high-yielding varieties. The CORRELATION feature achieved an accuracy of 88% in the first year, while the HOMOGENEITY feature reached 92% accuracy in the second year. This study confirms the hypothesis that TFs are more suitable for variety classification than for yield prediction. The results underscore the potential of TFs derived from multispectral images in early yield prediction and varietal classification, offering insights for HTP and precision agriculture alike.

Yield prediction models of organic oil rose farming with agricultural unmanned aerial vehicles (UAVs) images and machine learnaing algorithms

The alternative solution for predicting yields involves leveraging the latest technological advancements in unmanned aerial vehicles (UAVs) imagery. This approach aims to empower agricultural systems operating in nature-dependent conditions, enabling them to optimize the utilization of limited resources effectively. This study aimed to provide the agricultural data management needed for the adoption of sustainable agriculture systems and the transition of precision to smart agriculture systems by combining technology. The study was carried out in the 2019–2020 vegetation period to determine the effect of the parameters on the yield in the areas where organic farming is grown in Isparta oil rose (Rosa Damascena Mill.) and to develop yield prediction models by combining them with spectral measurements. In the study, the phenological, morphological, and biochemical changes of the Isparta oil rose were examined periodically, and the effects of biotic and abiotic stress factors during the vegetation period were determined in the study and testing area. For the early yield prediction of organic Isparta oil rose was chosen thirteen parameters as soil moisture, pH, electrical conductivity (EC), soil temperature, and nine vegetation indices were derived from UAV-acquired multispectral images. The parameters were functioned in machine learning algorithms, which were multiple linear regression (MLR), multivariate adaptive regression splines (MARS), decision trees (chi-squared automatic interaction detector (CHAID), exhaustive chi-squared automatic interaction detector (ExCHAID), classification and regression tree (CART)), random forest (RF)) and artificial neural network (ANN), to develop the prediction models. Early period (day 69) for organic rose yield prediction models were determined to be useable as MARS 1 (0.907 R2), ExCHAID 1 (0.888 R2), CART 1 (0.931 R2), and RF 1 (0.909 R2).As a result of the study, the models can be used integrated with UAV and soil data for the early prediction of Isparta oil rose yield under organic farming systems, especially during the Agriculture 4.0 transition phase in our country. It is such that models be developed and used in different ecologies for sustainable agricultural systems.

Enhancing WOFOST crop model with unscented Kalman filter assimilation of leaf area index

ABSTRACT The challenging task of early yield prediction is an essential problem for present-day agriculture. It is commonly solved with a crop model along with relevant observation data: field scouting, in-situ sensors, satellite imagery data and information from previous growing seasons. Crop growth simulation models benefit greatly from application of these data; however, only a limited number of established data assimilation procedures receive notable application. Most studies focus on model parameter calibration, machine learning, ensemble Kalman filters (EnKF) or particle filters. These methods are powerful yet computationally expensive, which limits their extensive application. In this study, we bring into consideration a modern KF variant – the unscented Kalman filter (UKF). We implement the UKF data assimilation for leaf area index (LAI) within WOFOST PCSE model. To demonstrate its efficiency, we conduct simulations with EnKF and UKF assimilation of Sentinel-2 LAI data and compare the results to actual historical yield data of five crops on 2740 fields. Also, a field-level numerical experiment is set up to demonstrate the influence of LAI assimilation on the predicted yield. The results indicate the proposed approach performs consistently and significantly improves the accuracy of predicted yields.

Lightweight green citrus fruit detection method for practical environmental applications

Real-time detection of green citrus fruit is crucial for accurate fruit localization and early yield prediction in the citrus growing process. This detection involves three imaging steps: acquisition, transmission, and detection. Image quality during acquisition and transmission also impacts detection accuracy. To enhance accuracy, we propose a lightweight method for detecting green citrus fruit in practical environments. The method applies a random sequence of four operations (blurred kernel, noise, size reduction, and compression) to preprocess original citrus fruit images, simulating real-world blurring and constructing a new dataset. A knowledge model is trained to reconstruct blurry citrus fruit images and address accuracy issues caused by blurring in practical environments. The YOLOv5 backbone network combines a convolutional neural network (CNN) for local information and a transformer for global background features. This fusion encodes both local and global information, learning representations efficiently and improving the detection model. In the YOLOv5 neck network, feature weighting optimizes representation of citrus fruit in space and channels, enhancing accuracy by reducing background weight. The proposed method, evaluated using a constructed dataset, outperforms mainstream lightweight detection models with 93.6% accuracy, 6.3 M parameters, and a 12.4 M model size. Compared to YOLOv5-MobileNetv3-small and YOLOv5-ShuffleNetv2, accuracy improves by 8.5% and 10.98%, respectively. Compared to YOLOv5s, parameters decrease by 9.6%, model size decreases by 8.82%, and accuracy improves by 1.5%. The proposed method demonstrates robust and effective detection of green citrus fruit in natural environments, providing guidance for accurate localization and early yield prediction. Data and methods used in this paper can be found at: https://github.com/PingfuChen/Citrus_detection.git.

Combining biophysical parameters with thermal and RGB indices using machine learning models for predicting yield in yellow rust affected wheat crop.

Evaluating crop health and forecasting yields in the early stages are crucial for effective crop and market management during periods of biotic stress for both farmers and policymakers. Field experiments were conducted during 2017-18 and 2018-19 with objective to evaluate the effect of yellow rust on various biophysical parameters of 24 wheat cultivars, with varying levels of resistance to yellow rust and to develop machine learning(ML) models with improved accuracy for predicting yield by integrating thermal and RGB indices with crucial plant biophysical parameters. Results revealed that as the level of rust increased, so did the canopy temperature and there was a significant decrease in crop photosynthesis, transpiration, stomatal conductance, leaf area index, membrane stability index, relative leaf water content, and normalized difference vegetation index due to rust, and the reductions were directly correlated with levels of rust severity. The yield reduction in moderate resistant, low resistant and susceptible cultivars as compared to resistant cultivars, varied from 15.9-16.9%, 28.6-34.4% and 59-61.1%, respectively. The ML models were able to provide relatively accurate early yield estimates, with the accuracy increasing as the harvest approached. The yield prediction performance of the different ML models varied with the stage of the crop growth. Based on the validation output of different ML models, Cubist, PLS, and SpikeSlab models were found to be effective in predicting the wheat yield at an early stage(55-60 days after sowing) of crop growth. The KNN, Cubist, SLR, RF, SpikeSlab, XGB, GPR and PLS models were proved to be more useful in predicting the crop yield at the middlestage(70days after sowing) of the crop, while RF, SpikeSlab, KNN, Cubist, ELNET, GPR, SLR, XGB and MARS models were found good to predict the crop yield at late stage (80days after sowing). The study quantified the impact of different levels of rust severity on crop biophysical parameters and demonstrated the usefulness of remote sensing and biophysical parameters data integration using machine-learning models for early yield prediction under biotically stressed conditions.

Feasibility of Early Yield Prediction per Coffee Tree Based on Multispectral Aerial Imagery: Case of Arabica Coffee Crops in Cauca-Colombia

Crop yield is an important factor for evaluating production processes and determining the profitability of growing coffee. Frequently, the total number of coffee beans per area unit is estimated manually by physically counting the coffee cherries, the branches, or the flowers. However, estimating yield requires an investment in time and work, so it is not usual for small producers. This paper studies a non-intrusive and attainable alternative to predicting coffee crop yield through multispectral aerial images. The proposal is designed for small low-tech producers monitored by capturing aerial photos with a MapIR camera on an unmanned aerial vehicle. This research shows how to predict yields in the early stages of the coffee tree productive cycle, such as at flowering by using aerial imagery. Physical and spectral descriptors were evaluated as predictors for yield prediction models. The results showed correlations between the selected predictors and 370 yield samples of a Colombian Arabica coffee crop. The coffee tree volume, the Normalized Difference Vegetation Index (NDVI), and the Coffee Ripeness Index (CRI) showed the highest values with 71%, 55%, and 63%, respectively. Further, these predictors were used as the inputs for regression models to analyze their precision in predicting coffee crop yield. The validation stage concluded that Linear Regression and Stochastic Descending Gradient Regression were better models with determination coefficient values of 56% and 55%, respectively, which are promising for predicting yield.

Correlation study of yield attributing traits in maize (Zea mays L.)

Trial was led at College Agronomy Farm, B. A. College of Agriculture, Anand Agricultural University, Anand, Gujarat, India during rabi season of the years 2019-20 and 2020-21 to study the association of yield attributing traits in maize under INM treatments. Number of cobs per plant sowing highly significant positive correlation with the all-other traits under evaluation. Number of cobs per plant and Number of seeds per cob in the year 2019-20 showing positive significant genotypic correlation with the grain yield per plant whereas phenotypic correlation is non-significant. Evaluation of the maize with the same treatments for the two subsequent years and getting same result, suggesting efficacy of the experimental selection and validating the results. This study will provide early yield prediction by considering effect of INM treatments on yield attributes.

Assessing Durum Wheat Yield through Sentinel-2 Imagery: A Machine Learning Approach

Two modeling approaches for the estimation of durum wheat yield based on Sentinel-2 data are presented for 66 fields across three growing periods. In the first approach, a previously developed multiple linear regression model (VI-MLR) based on vegetation indices (EVI, NMDI) was used. In the second approach, the reflectance data of all Sentinel-2 bands for several dates during the growth periods were used as input parameters in three machine learning model algorithms, i.e., random forest (RF), k-nearest neighbors (KNN), and boosting regressions (BR). Modeling results were examined against yield data collected by a combine harvester equipped with a yield mapping system. VI-MLR showed a moderate performance with R2 = 0.532 and RMSE = 847 kg ha−1. All machine learning approaches enhanced model accuracy when all images during the growing periods were used, especially RF and KNN (R2 > 0.91, RMSE < 360 kg ha−1). Additionally, RF and KNN accuracy remained high (R2 > 0.87, RMSE < 455 kg ha−1) when images from the start of the growing period until March, i.e., three months before harvest, were used, indicating the high suitability of machine learning on Sentinel-2 data for early yield prediction of durum wheat, information considered essential for precision agriculture applications.

Early yield prediction in different grapevine varieties using computer vision and machine learning

Yield assessment is a highly relevant task for the wine industry. The goal of this work was to develop a new algorithm for early yield prediction in different grapevine varieties using computer vision and machine learning. Vines from six grapevine (Vitis vinifera L.) varieties were photographed using a mobile platform in a commercial vineyard at pea-size berry stage. A SegNet architecture was employed to detect the visible berries and canopy features. All features were used to train support vector regression (SVR) models for predicting number of actual berries and yield. Regarding the berries’ detection step, a F1-score average of 0.72 and coefficients of determination (R2) above 0.92 were achieved for all varieties between the number of estimated and the number of actual visible berries. The method yielded average values for root mean squared error (RMSE) of 195 berries, normalized RMSE (NRMSE) of 23.83% and R2 of 0.79 between the number of estimated and the number of actual berries per vine using the leave-one-out cross validation method. In terms of yield forecast, the correlation between the actual yield and its estimated value yielded R2 between 0.54 and 0.87 among different varieties and NRMSE between 16.47% and 39.17% while the global model (including all varieties) had a R2 equal to 0.83 and NRMSE of 29.77%. The number of actual berries and yield per vine can be predicted up to 60 days prior to harvest in several grapevine varieties using the new algorithm.

Sugarcane yield prediction and genotype selection using unmanned aerial vehicle‐based hyperspectral imaging and machine learning

AbstractSugarcane (Saccharum spp. interspecific hybrids), a high biomass perennial crop, in which manual data collection for early yield prediction, through its growth cycle (∼12 mo long), is labor intensive and time consuming. Alternately, aerial imagery can be explored to predict yield‐related components and high‐throughput phenotyping for genetic selection. In this study, aerial imagery and ground data were collected in Stage IV (final stage of genotype selection) of the Florida sugarcane cultivar development program to evaluate the use of unmanned aerial vehicles in yield prediction (tons of cane per hectare [TCH], sucrose concentration, and tons of sugar per hectare [TSH]) in multiple new genotypes (13 in plant cane crop, nine in first ratoon crop). Aerial imagery data were collected using hyperspectral sensor, and yield data were collected through manual sampling of sugarcane stalks at harvest. The gradient‐boosting regression tree model was selected based on low mean absolute percentage error on multiple dates (April, July, and September) to determine the best timing of yield predictions. Results showed that yield was predicted with greater accuracy in July in plant crop and April in the first ratoon crop. Also, sucrose percentage was predicted with greater accuracy (94% in plant crop and 93% in first ratoon crop) than TCH and TSH. Although only two out of the top five genotypes were common in both selection methods (measured vs. predicted yields) in Stage IV, high accuracy in TCH and sucrose percentage shows that aerial imagery may be useful in making genotype selection in early stages when actual yield estimation is not feasible.

Early Yield Prediction in Spring Barley from Spectral Reflectance and Weather Data Using Machine Learning

Early Yield Prediction in Spring Barley from Spectral Reflectance and Weather Data Using Machine Learning

Maize Yield Prediction at an Early Developmental Stage Using Multispectral Images and Genotype Data for Preliminary Hybrid Selection

Assessing crop production in the field often requires breeders to wait until the end of the season to collect yield-related measurements, limiting the pace of the breeding cycle. Early prediction of crop performance can reduce this constraint by allowing breeders more time to focus on the highest-performing varieties. Here, we present a multimodal deep learning model for predicting the performance of maize (Zea mays) at an early developmental stage, offering the potential to accelerate crop breeding. We employed multispectral images and eight vegetation indices, collected by an uncrewed aerial vehicle approximately 60 days after sowing, over three consecutive growing cycles (2017, 2018 and 2019). The multimodal deep learning approach was used to integrate field management and genotype information with the multispectral data, providing context to the conditions that the plants experienced during the trial. Model performance was assessed using holdout data, in which the model accurately predicted the yield (RMSE 1.07 t/ha, a relative RMSE of 7.60% of 16 t/ha, and R2 score 0.73) and identified the majority of high-yielding varieties, outperforming previously published models for early yield prediction. The inclusion of vegetation indices was important for model performance, with a normalized difference vegetation index and green with normalized difference vegetation index contributing the most to model performance. The model provides a decision support tool, identifying promising lines early in the field trial.

New selection strategies for determining the traits contributing to increased grain yield in wheat (Triticum aestivum L.) under aluminum stress

Finding the suitable traits that mediate grain yield under stress condition represent an important challenge to plant breeders in order to improve production. Meaningful progress in computational analysis enables us to solve this challenge in a more efficient way. In this frame of study, our objective was to determine the most important traits associated with wheat yield improvement under aluminum stress using attribute weighting as well as supervised algorithms. To meet this goal, we studied, under normal and stress conditions, 167 bread wheat recombinant inbred lines. The lines were obtained from a cross of two semi-dwarf spring wheat varieties featured with considerable yield potential; Seri M82 and Babax. A total of 50 different traits including phenological, morphological, agronomic, physiological and biochemical traits were recorded. Ranking traits by attribute weighting algorithms indicated that among the 50 traits, five and eight traits were respectively the most probable candidates under normal and stress conditions, as highlighted by at least six weighting algorithms. According to the performance of decision tree algorithms employed, traits that were pinpointed traits by attribute weighting can be utilized to efficiently discriminate high-, medium-, and low- yield lines with up to 74.92% and 81.47% under normal and stress conditions, respectively. Moreover, valuable basis can be extrapolated from our decision tree models to guide plant breeders for selecting high yield lines going by the key traits highlighted herein. In addition, having an early prediction of yield helps farmers to make informed decision and take efficient steps to improve grain yield.

Qualitative crop condition survey reveals spatiotemporal production patterns and allows early yield prediction

Large-scale continuous crop monitoring systems (CMS) are key to detect and manage agricultural production anomalies. Current CMS exploit meteorological and crop growth models, and satellite imagery, but have underutilized legacy sources of information such as operational crop expert surveys with long and uninterrupted records. We argue that crop expert assessments, despite their subjective and categorical nature, capture the complexities of assessing the "status" of a crop better than any model or remote sensing retrieval. This is because crop rating data naturally encapsulates the broad expert knowledge of many individual surveyors spread throughout the country, constituting a sophisticated network of "people as sensors" that provide consistent and accurate information on crop progress. We analyze data from the US Department of Agriculture (USDA) Crop Progress and Condition (CPC) survey between 1987 and 2019 for four major crops across the US, and show how to transform the original qualitative data into a continuous, probabilistic variable better suited to quantitative analysis. Although the CPC reflects the subjective perception of many surveyors at different locations, the underlying models that describe the reported crop status are statistically robust and maintain similar characteristics across different crops, exhibit long-term stability, and have nation-wide validity. We discuss the origin and interpretation of existing spatial and temporal biases in the survey data. Finally, we propose a quantitative Crop Condition Index based on the CPC survey and demonstrate how this index can be used to monitor crop status and provide earlier and more precise predictions of crop yields than official USDA forecasts released midseason.

Using Bayesian growth models to predict grape yield

Background and aims: Seasonal differences in vine yield need to be managed to ensure appropriate fruit composition at harvest. Differences in yield are the result of changes in vine management (e.g., the number of nodes retained after harvest) and weather conditions (in particular, temperature) at key vine development stages. Early yield prediction enables growers to manage vines to achieve target yields and prepare the required infrastructure for the harvest.Methods and results: Bunch mass data was collected during the 2016/17, 2017/18 and 2018/19 seasons from a commercial vineyard on the Wairau Plains, Marlborough, New Zealand (41o2’23”S; 173o51’15”E). A Bayesian growth model, assuming a double sigmoidal curve, was used to predict the yield at the end of each season. The accuracy of the prediction was investigated using the Monte-Carlo simulation for yield prediction at various growth stages assuming different prior information.Conclusion: The results show that the model is sensitive to prior assumption and that having a non-informative prior may be more beneficial than having an informative prior based on one unusual year.

Automated early yield prediction in vineyards from on-the-go image acquisition

Early grapevine yield assessment provides information to viticulturists to help taking management decisions to achieve the desired grape quality and yield amount. In previous works, image analysis has been explored to this effect, but with systems performing either manually, on a single variety or close to harvest-time, when there are few rectifiable agronomic aspects. This study presents a solution based on image analysis for the non-invasive and in-field yield prediction in vines of several varieties, at phenological stages previous to veraison, around 100 days from harvest. To this end, an all-terrain vehicle (ATV) was modified with equipment to autonomously capture images of 30 vine segments of five different varieties at night-time. The images were analysed with a new image analysis algorithm based on mathematical morphology and pixel classification, which yielded overall average Recall and Precision values of 0.8764 and 0.9582, respectively. Finally, a model was calibrated to produce yield predictions from the number of detected berries in images with a Root-Mean-Square-Error per vine of 0.16 kg. This accuracy makes the proposed methodology ideal for early yield prediction as a very helpful tool for the grape and wine industry.

Early Yield Prediction Using Image Analysis of Apple Fruit and Tree Canopy Features with Neural Networks

(1) Background: Since early yield prediction is relevant for resource requirements of harvesting and marketing in the whole fruit industry, this paper presents a new approach of using image analysis and tree canopy features to predict early yield with artificial neural networks (ANN); (2) Methods: Two back propagation neural network (BPNN) models were developed for the early period after natural fruit drop in June and the ripening period, respectively. Within the same periods, images of apple cv. “Gala” trees were captured from an orchard near Bonn, Germany. Two sample sets were developed to train and test models; each set included 150 samples from the 2009 and 2010 growing season. For each sample (each canopy image), pixels were segmented into fruit, foliage, and background using image segmentation. The four features extracted from the data set for the canopy were: total cross-sectional area of fruits, fruit number, total cross-section area of small fruits, and cross-sectional area of foliage, and were used as inputs. With the actual weighted yield per tree as a target, BPNN was employed to learn their mutual relationship as a prerequisite to develop the prediction; (3) Results: For the developed BPNN model of the early period after June drop, correlation coefficients (R2) between the estimated and the actual weighted yield, mean forecast error (MFE), mean absolute percentage error (MAPE), and root mean square error (RMSE) were 0.81, −0.05, 10.7%, 2.34 kg/tree, respectively. For the model of the ripening period, these measures were 0.83, −0.03, 8.9%, 2.3 kg/tree, respectively. In 2011, the two previously developed models were used to predict apple yield. The RMSE and R2 values between the estimated and harvested apple yield were 2.6 kg/tree and 0.62 for the early period (small, green fruit) and improved near harvest (red, large fruit) to 2.5 kg/tree and 0.75 for a tree with ca. 18 kg yield per tree. For further method verification, the cv. “Pinova” apple trees were used as another variety in 2012 to develop the BPNN prediction model for the early period after June drop. The model was used in 2013, which gave similar results as those found with cv. “Gala”; (4) Conclusion: Overall, the results showed in this research that the proposed estimation models performed accurately using canopy and fruit features using image analysis algorithms.

Detection of apple fruit in an orchard for early yield prediction as a function of crop load

Detection of apple fruit in an orchard for early yield prediction as a function of crop load

Proximal NDVI derived phenology improves in-season predictions of wheat quantity and quality

Automated, low-cost and field-deployable remote sensing tools are well suited for continuously monitoring crop growth and providing growers with timely information about crop performance. Because automated sensors provide information about crop development and performance across time, we examined the hypothesis that ground-based canopy reflectance data might define crop phenology in new ways over the course of the season that can better forecast crop yield, protein, biomass, and grain nitrogen at harvest. This study examines the utility of daily Normalized Difference Vegetation Index (NDVI) data to monitor crop phenology over two complete growing seasons. Spectral reflectance data was collected at a total of sixteen plots under four different applied nitrogen (N) and soil water availability scenarios in rainfed soft white spring wheat (Triticum aestivum L.). Using NDVI at solar noon, four phenological periods were derived from the data using a non-parametric regression locally weighted smoothing parameter (loess) to account for day to day variability, and piecewise linear regression to determine inflection points in the seasonal NDVI curve. The NDVI derived phenological metrics (i.e. the change in NDVI per day, and duration (in days) of each phenological period) were compared against daily NDVI values throughout the season to predict harvest metrics. Daily NDVI data were generally poor predictors of harvest metrics early in the growing season, and reached maximum predictive power at the onset of heading, and the middle of ripening for biomass and yield (R2∼0.50 and ∼0.25 during heading, respectively, and R2∼0.50 during early ripening). Conversely, using both simple and multiple regression analysis, we found that harvest metrics were better explained using the rate and duration of NDVI derived phenological periods. Simple regressions between NDVI derived phenological metrics revealed several physiologically and management relevant correlations including strong, statistically significant (p<0.05) relationships between the rate of tillering and stem extension and total biomass (R2=0.63 and 0.54, respectively), the duration of heading and yield (R2=0.67), the rate of ripening and grain protein concentration (R2=0.45), and the duration of ripening and grain N content (R2=0.43), for example. Using multiple regression analysis, 83% of the variance in yield, 67% in protein concentration, 87% in total biomass, and 80% in grain N was explained by two to three NDVI derived phenological metrics. Further, multiple regression analysis using NDVI derived phenological metrics from the early season (tillering and stem extension) substantially improved early prediction of yield and biomass as compared to daily NDVI data, whereas protein and grain N were primarily driven by metrics associated with the reproductive development of the crop (heading and ripening). This work has implications for improving in-season management decisions and understanding of the phenological drivers of harvest metrics using daily NDVI data as an evaluation tool.