Search

• A new method to isolate only grapevine NDVI from a Sentinel-2 mixed signal is proposed. • UAV imagery was used to retrieve vines fraction cover within Sentinel-2 pixel. • Two-endmembers spectral linear mixture model was assumed. • Pure NDVI values of endmembers was locally solved using a moving window approach. • MAPE < 16 % resulted comparing estimates to UAV-retrieved high resolution NDVI values. Satellite-based multispectral remote sensing in the wine sector is expanding, aiming at improving vineyard management for both environmental sustainability and vine quality/yield. However, vineyards present a discontinuous vegetative surface, with rows of vines alternating with background areas (bare soil or other vegetation). This irregular pattern adversely affects multispectral satellite data from public and research missions (such as Sentinel 2 and Landsat 8/9, etc.), which operate at lower geometric resolutions. When inter-row spaces become overgrown with other vegetation that occasionally requires mowing, the average spectral response of the pixels changes significantly. Consequently, spectral information specific to the vines is obscured by a complex signal, potentially leading to incorrect conclusions if directly analyzed. To address this issue, this study introduces a novel method for recovering the spectral signal of vines, specifically focusing on NDVI from Sentinel-2 data (S2).The approach relies on the estimation of the local NDVI value of vines by a least squares techniques based on the application of a spectral unmixing technique operated in the space domain using a moving window surrounding the pixel for which the estimation is required for. At each moving window step, NDVI values of only grapevines (at satellite resolution i.e., 10 m per 10 m) were estimated and mapped using as main inputs the S2 NDVI values and the grapevine fraction cover values retrieved by high resolution UAV imagery. Results shows a 16 % relative error in NDVI measurements for vines. [ABSTRACT FROM AUTHOR]

• Performance evaluation of vine canopy detection methods using multispectral images. • Deep learning methods achieved the best performance in vineyard canopy detection. • U-Net and Mask R-CNN provide more accurate monitoring to assess vineyard vigor. • Canopy NDVI varies as a function of image analysis techniques applied. In viticulture, the rapid and accurate acquisition of canopy spectral information through ultra-high spatial resolution imagery is increasingly demanded for decision support. The prevalent practice involves creating vigor maps using spectral data obtained from pure vine canopy pixels. Object-Based Image Analysis (OBIA) among conventional methods exhibits a reasonable efficiency in canopy classification due to its feature extraction capabilities. In recent years, deep learning (DL) techniques have demonstrated significant potential in orchard monitoring, leveraging their ability to automatically learn image features. This study assessed the performance of different methodologies, including Mask R-CNN, U-Net, OBIA and unsupervised methods, in identifying pure canopy pixels. The effectiveness of shadow and background detection methods and the impact of misclassified pixels on NDVI were compared. Results were compared with agronomic surveys conducted during the 2021 and 2022 growing seasons, focusing on two distinct phenological stages (BBCH65-BBCH85). Mask R-CNN and U-Net exhibited superior performance in terms of Overall Accuracy (OA), F1-score, and Intersection Over Union (IoU). Among OBIA methods, the Gaussian Mixture Model (GMM) proved to be the most effective classifier for canopy segmentation, and Support Vector Machine (SVM) also demonstrated reasonable stability. Conversely, Random Forest (RF) and K-Means yielded lower accuracy and higher error rates. As a result of the limited accuracy, it is noted for vineyard rows with low vigor canopies that NDVI was overestimated, while for high vigor canopies NDVI was underestimated. Significantly improved determination coefficients were observed for the comparison between Total Leaf Area (TLA) and NDVI data derived from Mask R-CNN and U-Net. Positive correlations were also found with NDVI data from GMM and SVM algorithms. Regarding leaf chlorophyll (Chl) and NDVI correlations, Mask R-CNN and U-Net methods showed superior performance. Additionally, the relationship between TLA and projected canopy area (PCA) was significantly better represented by U-Net and Mask R-CNN, while PCA was not recommended for estimating chlorophyll content. This investigation establishes that improved vine canopy delimitation contributes to improved vineyard vigour monitoring, providing winegrowers with more accurate and reliable agronomic information for management decisions. [ABSTRACT FROM AUTHOR]

• A crop yield predicting model for promoting sustainable production and facilitating local stakeholder decision-making. • LSTM with Attention model can perform well better than panel regression but falls short when trained with small dataset. • Departmental series of peak NDVI images were effective for early soybean yield prediction. • Deep learning models showed superior generalization, predicting well on new, unseen data. The accurate prediction of soybean yield is vital for global food market stabilization and food security. Recent advancements in remote sensing technology have significantly amplified interest in leveraging satellite-based methods for predicting crop yield. These methods offer in-season yield estimates. By utilizing this timely information, decision-makers can formulate strategic, well-informed choices that preemptively mitigate potential food price hikes, ultimately bolstering food security. While simple regression models have been widely utilized for satellite-based yield prediction, researchers have recently begun to explore the use of deep learning algorithms. This study compares the performance of panel regression and deep learning models for in-season soybean yield prediction at the Department (county-equivalent) level in Argentina. Data sources include the latest soybean land use products and MODIS bi-weekly vegetation index products. Results indicate that deep learning models significantly outperform panel regression. Deep learning Long Short-Term Memory (LSTM) models, which incorporate attention mechanism and a series of peak NDVI images, generate more accurate and time-sensitive predictions. Among competing LSTM models, the one with attention mechanism applied to the entire growing season's NDVI data yields the highest prediction accuracy, with a Root Mean Square Error (RMSE) of 505.78 kg/ha and Normalized Root Mean Square Error (NRMSE) of 0.0726. The LSTM model with attention on the three highest NDVI images attains a satisfactory prediction accuracy (RMSE = 627.28 kg/ha, NRMSE = 0.089) six weeks prior to harvest. This study presents a robust model for predicting crop yields, promoting sustainable production of soybeans and facilitating knowledgeable choices among farmers and policymakers. [ABSTRACT FROM AUTHOR]

• Predict the normalized difference vegetation index (NDVI) using RGB drone imagery data. • The curvelet coefficients were analysed using Empirical Curvelet Transform method from RGB. • The deep DenseNet model predict the NDVI using the output of Empirical Curvelet Transform method. • This work is helpful to monitor the growth, plant health, and identify crop problems. Predicting accurately the Normalized Difference Vegetation Index (NDVI) trends from RGB images are essential to monitor crops and identify issues related to plant diseases, and water shortages. The current NDVI prediction models are primarily based on traditional machine learning models which lack reliability due to the problem related to atmospheric conditions. To predict NDVI in Prince Edward Island using RGB drone imagery data, this paper proposed a novel framework integrating empirical curvelet transform and DenseNet models. Each channel of RGB drone imagery data was passed through empirical curvelet transform method where the curvelet coefficients were analysed which result in creating a new formula to design NDVI. The output of the new formula was sent to the deep DenseNet to predict the final NDVI. The proposed model was evaluated using quantitative metrics including, Q-Q plot, regression, correlation coefficients, structural similarity (SSIM), peak signal to noise ratio (PSNR) and mean square error (MSE) as well as accuracy (A C C) , sensitivity (S E N) , f1-score, specificity. The obtained results showed that the proposed model outperformed the previous models by scoring the highest values of SSIM = 0.98, and lowest MSE = 120. It is believed that the proposed model is helpful to support farmers in monitoring the growth and plant health as well as to identify crops problems. [ABSTRACT FROM AUTHOR]

• Evaluate the potential using Landsat 8 vegetation index for jujube yield estimation. • Identify the phenology time for making a reliable jujube yield prediction. • Integrate phenology length to improve remote sensing-based yield forecasting model. • The proposed method showed better performance than the leave-one-year-out method. It is challenging to generate a time series of vegetation indices from moderate spatial resolution Landsat Thematic Mapper images (Landsat 8) for crop yield forecasting. In addition, crop yields are correlated with phenology information, especially the fruit filling days. The objectives of this study were to identify the phenology time for making a reliable jujube yield prediction, more importantly, explore an approach that used the length of phenology growth period to improve remotely sensed estimates of inter-annual variability for yields. The best time for making jujube yield prediction was found to be during the fruit filling period, showing higher correlation coefficient (r) between vegetation indices and yields. The average NDVI for 14th and 15th half-months represented a better performance for yield prediction, with a highest r value of 0.87 for NDVI, 0.82 for SAVI, 0.73 for NDWI and 0.73 for EVI, respectively. The potential of using Landsat-NDVI for jujube yield estimation, combined with the phenological length, was preliminarily proved based on 200 observations of individual jujube orchards, showing a validated R2 of 0.85, 0.80 and 0.67, RMSE of 0.61, 0.78 and 0.85 t ha−1 for 2013, 2014 and 2016, respectively. Furthermore, the phenological adjusted model was further evaluated by inter-annual official statistic data, with R2 and RMSE values ranging from 0.38 to 0.53, and 0.31 to 0.47 t ha−1, respectively. The proposed method showed better performance between years when the fruit filling days differed greatly than the leave-one-year-out method, which was verified to well fit to jujube yield monitoring and mapping two months before harvest. [ABSTRACT FROM AUTHOR]

Graphical abstract Highlights • Gross features of the phenology of fields can be detected with Sentinel-2. • High weed infestations can be detected from NDVI time series from Sentinel-2. • Seven phenological classes of cropping practice were mapped for two Sentinel-2 tiles. • Two thirds of the winter crops show some sign of weed infestation. • The thematic accuracy of the land use map is assessed to 69%. Abstract Maps of cropping practice, including the level of weed infestation, are useful planning tools e.g. for the assessment of the environmental impact of the crops, and Northern Italy is an important example due to the large and diverse agricultural production and the high weed infestation. Sentinel-2A is a new satellite with a high spatial and temporal resolution which potentially allows the creation of detailed maps of cropping practice including weed infestation. To explore the applicability of Sentinel-2A for mapping cropping practice, we analysed the Normalised Differential Vegetation Index (NDVI) time series from five weed-infested crop fields as well as the areas designated as non-irrigated agricultural land in Corine Land Cover, which also contributed to an increased understanding of the cropping practice in the region. The analysis of the case studies showed that the temporal resolution of Sentinel-2A was high enough to distinguish the gross features of the cropping practice, and that high weed infestations can be detected at this spatial resolution. The analysis of the entire region showed the potential for mapping cropping practice using Sentinel-2. In conclusion, Sentinel-2A is to some extent applicable for mapping cropping practice with reasonable thematic accuracy. [ABSTRACT FROM AUTHOR]

Stakeholders, policy makers, government planners and agricultural market participants in Central Eastern Europe require accurate and timely information about wheat and maize yield and production. The study site, the lowlands (altitude below 200 m) of the Tisza river catchment is by far the most important wheat and corn producing region in the Carpathian basin, and even in Central Eastern Europe. The conventional sampling of on-field data and data processing for crop forecasting requires significant amounts of time before official reports can be released. Several studies have shown that wheat and maize yield can be effectively forecast using satellite remote sensing. In this study, a freely available MODIS NDVI satellite data based wheat and maize yield forecasting methodology was developed and evaluated for estimating yield losses effected by drought. Wheat and maize yield was derived by regressing reported yield values against time series of 15 different peak-season MODIS-derived NDVI. The lowest RMSE values at the river basin level for both wheat and maize yield forecast versus reported yield were found when using at least six or more years of training data. Wheat forecast for the 2000 to 2015 growing seasons were within 0.819% and 19.08% of final reported yield values. Maize forecast at county level for the 2000 to 2015 growing seasons were within 0.299% and 17.14% of final reported yield values. The Nash–Sutcliffe efficiency index (E 1 ) is positive with E 1  = 0.322 in the case of wheat forecast, and with E 1  = 0.401 in the case of maize forecast, which means the developed and evaluated forecasting method performs acceptable forecast efficiency. Nevertheless the occurrence of extreme drought or extreme precipitation can alter the forecasting efficiency resulting over or underestimation. Overall statement, which based on MODIS NDVI, possible yield losses can easily be forecasted 6–8 weeks before harvesting and applying simple threshold levels, yield losses can be mapped simply. [ABSTRACT FROM AUTHOR]

Phenology is one of the most reliable indicators of vegetation dynamics. Assessing and monitoring the dynamics of phenology is relevant to support several decisions in order to improve the efficiency of several farming practices. An open source application – QPhenoMetrics - implemented in QGIS software that estimates vegetation phenology metrics is presented, using Earth Observation Systems (EOS) based time-series of Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) as proxies for phenology. QPhenoMetrics is characterized by freely-usable and updatable code, acceptance of satellite images or text formats, time-series analysis toolbox allowing the selection of region of interest with statistical quality assessment for Vegetation Indices (VI), and estimation of ensemble metrics. The application is structured in three components: (i) input data; (ii) pre-processing of the VI time-series and several fitting methods and (iii) computation of the phenological metrics. QPhenoMetrics produces a plot with the VI time-series and corresponding phenology metrics, and a spreadsheet is created with a list of NDVI or EVI values estimated using the selected fitting method. To evaluate the application, two main Portuguese crops, vineyards and maize, and MOD13 data from MODIS sensor during 2011–2012 were considered. QPhenoMetrics was validated with vineyard phenology observations (2007–2011). A comparative analysis with software products TimeSat and Spirits was also performed. It was concluded that QPhenoMetrics can be very useful for common users to extract phenology information for 16 daily MODIS data in HDF format, text files with NDVI/EVI data and ASCII files, through a simple and intuitive graphic interface. Furthermore, the user can evaluate the quality assessment of VI of the images used. QPhenoMetrics is an effective open source tool that in addition to being free, is readily modifiable by user according to the study requirements. [ABSTRACT FROM AUTHOR]

Cover crops may be a factor in the variability of soil fertility in areas under a no-tillage system (NTS); one which is not taken into account when fertilising the cash crop. The aim of this work was to evaluate the effect of the variability caused by the dry matter and nutrient cycling of the cover crop on the yield of the succeeding cash crop, by means of the NDVI, calculated by a portable ground sensor and using an unmanned aerial vehicle (UAV). The study was carried out in two areas of a Oxisol cultivated under NTS, where soil samples were taken, for chemical analysis and to evaluate the dry matter and nutrient accumulation in a winter cover crop of black oats. NDVI readings were taken at the R5 and R5.5 stages in soybean. A portable sensor was used seeking a comparison with the readings made by the UAV. It could be seen from the results, that the cover crop has an influence on the main crop under NTS. In the present case, the dry weight of the black oats as well as the accumulated nutrients of nitrogen, phosphorus and magnesium, showed the highest correlations with grain yield in the soybean, whereas for the chemical properties of the soil under evaluation, organic matter had the most influence on the grain yield. The NDVI assessed using the portable ground sensor and with the UAV, was efficient in evaluating the effects of variability in the crop of black oats and in the cycled nutrients. [ABSTRACT FROM AUTHOR]

• A comparison of crop masks to spatially aggregate NDVI for yield forecasting in EU. • Forecasts improve when using annual crop group-specific than generic crop masks. • Improvements concern both accuracy and timeliness of yield forecasts. • Annual crop group-specific masks allow better interpretation of NDVI time series. • Agricultural Decision Support Systems (DSS) can benefit from the obtained results. Operational crop yield forecasting services typically provides crop yield forecasts based on regression models between official yields and agro-environmental variables, among which meteorological data, crop simulation model or satellite-derived indicators. The reliability with which these variables infer on yields depends, among many other factors, also on their aggregation in the space domain, for example on the type of the crop masks utilized in the aggregation process from point to regional scale. This work investigates how the yield explanatory power of satellite-derived indicators is changing moving from time-stable arable land masks to annual crop group-specific masks. We compare in particular the linkage between time series of regional crop yield in Europe and the Normalized Difference Vegetation Index (NDVI) from Moderate Resolution Imaging Spectroradiometer (MODIS), when generic arable land masks or crop group-specific and year-specific information are applied to aggregate pixel values at regional level. Regional (Eurostat level NUTS-2) crop yield statistics were collected from official databases for the period 2003–2019, while NDVI data were derived from MODIS daily products at 250 m spatial resolution for the same reference period. Regional NDVI profiles were retrieved by averaging single pixels time series according to the information of five crop masks, including generic arable land masks, crop group-specific static masks (separating winter and spring crops from summer crops), and annual crop group-specific masks (distinguishing between the two crop groups and varying in time). A compared correlation analysis between yield data and regional temporal NDVI profiles was performed assuming a linear regression model. Coefficient of determination R2 and Root Mean Squared Error (RMSE) were computed to assess the models' errors and to analyze the effect of the aggregation on the different crop masks. Results indicated an improvement in yield estimation when using annual crop group-specific indicators with respect to generic and static products. Although not homogeneously distributed throughout Europe, advantages have been highlighted both in terms of accuracy and timeliness of the prediction. In most regions, the introduction of annual masks allowed to reduce RMSE values by 0.3 t/ha and advanced forecast times by up to 30 days. The added value in the use of annual crop group-specific masks concerned the two European most cultivated crops, namely soft wheat and grain maize. [ABSTRACT FROM AUTHOR]

The growth of precision agriculture has allowed farmers access to more data and greater efficiency for their farms. With consistently tight profit margins, farmers need ways to take advantage of the advancement of technology to lower their costs or increase their revenue. One area where these advancements can prove beneficial are in the measurement of vegetation indices such as the Normalized Difference Vegetation Index (NDVI) and Normalized Difference Red Edge Index (NDRE). Color maps representing these vegetation indices can be used to identify problem areas, plant health, or even places where spot applications are needed. These color maps help farmers to visualize these areas. Currently, a multi-thousand dollar multispectral camera, typically attached to an Unmanned Aerial Vehicle (UAV) during flight, is required for measuring these indices. This makes obtaining NDVI and NDRE somewhat cost prohibitive for most farmers. This work demonstrates a solution to this cost issue. The solution involves the use of a conditional Generative Adversarial Network known as Pix2Pix. By using Pix2Pix along with training data from UAV flights of corn, soybeans, and cotton, this paper highlights the potential for predicting comparable NDVI and NDRE with a low-cost Red-Green-Blue (RGB) camera. This paper proposes and assesses a cost-efficient method that can comparably predict these vegetation indices, resulting in cost-savings in the range of $5000 per UAV system. • Pix2Pix used with aerial crop imagery to estimate vegetation indices. • Deep Learning for estimation of vegetation indices from standard imagery. • Using aerial crop imagery to train a network capable of producing accurate vegetation index estimates. • Leveraging Artificial Intelligence for cost savings. • NDVI estimation with Pix2Pix. [ABSTRACT FROM AUTHOR]

Identifying the limited available resource, use of improved irrigation technology and better water resources management have been suggested as mechanisms for increased production. To optimize the agricultural production, knowing the available land for irrigation is mandatory. This study aimed to assess the suitable land for surface irrigation using different spatial parameters (normalized difference vegetation index, soil, slope, proximity, rainfall deficit, and land use). Land suitability was assessed by considering the individual parameters and then by overlaying. The overlay suitability analysis was performed by using weighted overlay tool within spatial analyst tool in ArcGIS. After excluding the irrigation constraints suitable land for surface irrigation was identified. Based on the individual parameters analysis the normalized difference vegetation index gives the largest land size which is 95 % of the total land as suitable. While the slope factor gives the lowest figure which is 57 % of the total land is suitable for irrigation. The weighted overlay analysis, gives only 28.46 % of the total land in the basin was suitable for surface irrigation. More than 93 % of this suitable land is grouped in to small scale while only about 7 % is used for medium to large scale irrigation. [ABSTRACT FROM AUTHOR]

• 1. Use of active illumination, through own LED-technology sources operating on specific wavelengths. • 2. Possibility to perform measurement sessions regardless of ambient lighting conditions. • 3. multispectral images where pixel values are reflectance. • 4. system able to mapping NDVI and other spectral indices (EVI, SAVI, GI, etc). • 5. Measurements can be obtained both in the form of images and in numerical format. Current Proximal Sensing technologies are based on multispectral imaging systems able to capture images in a few spectral bands, usually centred in VIS and NIR regions, to derive vegetation indices. However, most of such systems lack an internal radiometric calibration to estimate the actual reflectance of the observed target, making them sensitive to the local radiative environment and requiring a per-session calibration against a reference target. To overcome such dependence, the instrument described adopts an active illumination of the target surface, allowing the monitoring of soil and low vegetation surfaces by a radiometrically pre-calibrated imaging camera. The system, driven by a micro-controller, is contained in a light-tight box, actively illuminated by LED sources with different emission bands (R, G, B and IR). These features make measurements with the proposed instrument independent of local lighting conditions and the need for the in situ, per-session calibrations that are mandatory with the conventionally adopted instrumentation for field measurements. The capture of multispectral images is performed at a relatively high resolution (1269x972 px) in the broadband visible (VIS, 420–670 nm), monochromatic (R, G, B) and near-infrared (NIR, 820 nm) domains, and is suitable to characterise different surfaces, such as soil, turf, and low-vegetation. A web-based user interface allows the control and the parametrisation of measurements, together with the archival of the acquired images both in RAW format and as calibrated spectral reflectance or already processed as the most common vegetation indices (NDVI, EVI, SAVI). The system is intended for use in a research environment and the technical management of recreation or cultivated surfaces, either for direct use or as a calibrated reference in support of proximal-sensing or drone-based applications. [ABSTRACT FROM AUTHOR]

Corn height measured manually has shown promising results in improving the relationship between active-optical (AO) sensor readings and crop yield. Manual measurement of corn height is not practical in US commercial corn production, so an alternative automatic method must be found in order to capture the benefit of including canopy height into in-season yield estimates and from there into in-season nitrogen (N) fertilizer applications. One existing alternative to measure canopy height is an acoustic height sensor. A commercial acoustic height sensor was utilized in these experiments at two corn growth stages (V6 and V12) along with AO sensors. Eight corn N rate sites in North Dakota, USA, were used to compare the acoustic height sensor as a practical alternative to manual height measurements as an additional parameter to increase the relationship between AO sensor readings and corn yield. Six N treatments, 0, 45, 90, 134, 179, and 224 kg ha −1 , were applied before planting in a randomized complete block experimental design with four replications. Height measurement using the acoustic sensor provided an improved yield relationship compared to manual height at all locations. The level of improvement of the relationship between AO readings multiplied by acoustic sensor readings and yield was greater at V6 growth stage compared to the V12 growth stage. At V12, corn height measured manually and with the acoustic sensor multiplied by AO readings provided similar improvement to the relationship with yield compared to relating AO readings alone with yield at most locations. The acoustic height sensor may be useful in increasing the usefulness of AO sensor corn yield prediction algorithms for use in on-the-go in-season N application to corn particularly if the sensor height is normalized within site before combining multiple locations. [ABSTRACT FROM AUTHOR]

Near-ground geophysical soil sensors provide valuable information for precision agriculture applications. Indeed, their readings can be used as proxy for many soil parameters. On-the-go soil sensor surveys are, typically, carried out intensively (e.g., every 2 m) over many parallel transects. Two types of soil sensors measurements are considered in this paper: apparent electrical conductivity (4 fields in California, USA) and reflectance (1 field in Italy). Two types of spatial interpolations are carried out, universal kriging (model-based) and inverse distance weighting (deterministic). Interpolation quality assessment is usually carried out using leave-one-out ( loo ) resampling. We show that loo resampling on transect sampling datasets returns overly-optimistic, low interpolation errors, because the left-out data point has values very close to that of its neighbors in the training dataset. This bias in the map quality assessment can be reduced by removing the closest neighbors of the validation observation from the training dataset, in a (spatial) h-block (SHB) fashion. The results indicate that, for soil sensor data acquired along parallel transects: ( i ) the SHB resampling is a useful tool to test the performance of interpolation techniques and ( ii ) the optimal (i.e., rendering the same errors of un-sampled locations between transects) SHB threshold distance ( h.dist ) for neighbor-exclusion is proportional to the semi-variogram range and partial sill. This procedure provides research scientists with an improved means of understanding the error of soil maps made by interpolating soil sensor measurements. [ABSTRACT FROM AUTHOR]

The accurate assessment of the vigor and disease impact is a major challenge in precision viticulture. It is essential for managing phytosanitary treatments. Up to now, some remote sensing techniques such as aerial imagery and handheld optical sensors have been applied to grapevine characterization. However each technique provides limited, specific information about foliage. To broaden the characterization of the foliage, we developed a proximal integrated, multispectral imaging sensor that operates in the visible and near-infrared bands. It is mounted on a track-laying tractor equipped with a Greenseeker-RT-100, coupled with a GPS-RTK. As the sensor is very sensitive to the ambient light, a radiometric calibration is required: it allows producing absolute reflectance images, using a color chart. If the chart is hidden by leaves, for instance, the images are corrected using the linear interpolation method. The adaptive radiometric method is evaluated as a function of the number of neutral patches selected on the color chart during the linear regression process and the efficiency of the spatial interpolation method is assessed using a leave-one-out-cross-validation (LOOCV) method. The radiometric calibration is validated by comparison of NDVI maps produced by imagery and by the Greenseeker, a commercial system. In the early stage of berry formation, we examined and quantified the spatial patterns and demonstrated a low-cost imagery method that is capable of analyzing correctly the vigor. This corroborates the efficiency of the calibration method encouraging the use of multi-spectral imagery for other vineyard applications, such as the characterization of physiological status. [ABSTRACT FROM AUTHOR]

Cotton root rot is a very serious and destructive disease of cotton grown in the southwestern United States. Accurate information regarding its spatial and temporal distribution within fields is important for effective management of the disease. The objectives of this study were to examine the consistency and variation of cotton root rot infections within cotton fields over 10-year intervals using airborne multispectral imagery and to assess the feasibility to use historical imagery to create prescription maps for site-specific management of the disease. Airborne multispectral images collected from a 102-ha cotton field in 2001 and 2011 and from a 97-ha field in 2002 and 2012 in south Texas were used in this study. The images were rectified and resampled to the same pixel size between the two years for each field. The normalized difference vegetation index (NDVI) images were generated and unsupervised classification was then used to classify the NDVI images into root rot-infected and non-infected zones. Change detection analysis was performed to detect the consistency and change in root rot infection between the two growing seasons for each field. Results indicate that the spatial patterns of the disease were similar between the two seasons, though variations existed for each field. To account for the potential expansion and temporal variation of the disease, buffer zones around the infected areas were created. The buffered maps between the two years agreed well. The results from this study demonstrate that classification maps derived from historical images in conjunction with appropriate buffer zones can be used as prescription maps for site-specific fungicide application to control cotton root rot. [ABSTRACT FROM AUTHOR]

Site-specific management strategies in wheat fields can be strongly enhanced with sensor technology that detects spatial changes of wheat biomass. The objectives of this study were to propose a multi-sensor approach combining a digital camera system that measures plant coverage with an arbitrary crop height measuring instrument for estimating wheat biomass. Digital images, fresh and dry biomass, and crop height measurements were taken at 180 sample points distributed over 4 fields between the BBCH growth stages 30–75. Plant coverage percentage was calculated by separating plant pixels from background pixels of NDVI and NIR images using image segmentation based on partitioning clustering. Performance of three clustering algorithms (k-means, partition around mediods (PAM), and fuzzy c-means) was analyzed. Plant coverage from image clustering was further related to fresh and dry biomass with and without crop height measurements using simple and multiple regression models. The performance of the three clustering algorithms was similar for estimating wheat biomass. NDVI image segmentation was highly obstructed by scattering effects especially at later BBCH stages through the presence of wheat ears, stems, and tilted leaves, whereas NIR image segmentation was generally good except with images that were taken at locations with very low plant coverage and dry soil crusts. Consequently, NIR image clustering yielded more accurate estimates of fresh and dry biomass ( R 2 = 0.79/0.68) than NDVI image clustering ( R 2 = 0.66/0.56) among the individual measurement runs on average. Still cloud conditions had some influence on NIR clustering. By pooling the complete set of sampling points from all measurement runs into a global model, the combination of image clustering with crop height was helpful. For fresh biomass, global model quality changed from R 2 = 0.15 or R 2 = 0.46 without crop height to R 2 = 0.63 or R 2 = 0.82 with crop height for NDVI or NIR, respectively. For dry biomass, crop height was itself a strong predictor with R 2 = 0.86, whereas the model improvement by including image clustering of plant coverage was nearly negligible. In conclusion, the combination of a camera sensor using image clustering with a sensor able to measure crop height such as LIDAR or ultrasound systems seems to be a promising way to reach for more accurate and robust estimations of wheat biomass especially when measurements over multiple fields and dates are considered without the need of re-calibration. [ABSTRACT FROM AUTHOR]

Broccoli is a highly perishable vegetable. In academia, shelf life of broccoli is generally monitored using colorimetric devices based on diffuse reflection. However, actual quality and storability of a single broccoli curd after harvest remains uncertain before changes are noticed by the human eye. The aim of the present study was to evaluate broccoli quality loss during storage based on differences in reflectance spectra and to judge quality stages prior to color changes. Reflection spectra of broccoli harvested throughout one cultivation period in northern Germany (June–September) and stored at 16 °C and 95% relative humidity for four days were analyzed by principal component analysis (PCA). Apart from an overall increase of reflection in the visible range, changes of the spectral signature were limited to the red edge region. Quality and storability of broccoli could sufficiently be rated by the help of the normalized difference vegetation index (NDVI) and the inflection point of the red edge region, respectively. The inflection point was more sensitive to quality changes, while the NDVI was additionally influenced by differences in fertilizer application during cultivation. Analysis of hyperspectral images of broccoli curds showed an increase in both means and variance of Euclidean distances during storage. This applied to the overall VIS spectrum but was most prominent in the range of 530–650 nm. The variation could adequately be demonstrated by the help of composite images, when reflection intensities of three different wavelengths (530, 590, and 650 nm) were placed as RGB image channels. Means of Euclidean distances were highly correlated with means of the colorimetric parameter b ∗ ( R 2 = 0.92). Detected variation within the images was most likely based on different senescence status of single florets. However, classification procedures failed to sort the derived information into image classes because reflection intensity increased rather gradually. Further examinations at lower storage temperatures are required in order to clarify if early quality losses might be sooner detected by Euclidean distances in hyperspectral images compared to CIELAB-readings. [ABSTRACT FROM AUTHOR]

Recent precision agriculture research has focused on the utilization of management zones (MZs) as a method for variable-rate fertilizer applications. This study tries to make use of an active canopy sensor to define MZs in contrast with the ones using soil properties. To achieve this objective, a 20-m grid-sampling scheme was imposed on the field with 101 points, and we collected soil samples from the top 20 cm at each point before tobacco planting in April 2013. The normalized difference vegetative index (NDVI) data at five growth stages of the tobacco growth cycle were measured by using a GreenSeeker handheld crop sensor at the location of each sample point. Yield mapping was carried out at harvest. Through stepwise multiple analysis, we identified the key yield-limiting factors of soil properties and stages of NDVI measured respectively. For soil property data, OM, AP and Fe were used for yield-based MZs while for canopy parameters, NDVI during knee-high and flowering were utilized. Fuzzy c-means clustering algorithm was performed in order to determine the optimal number of the clusters. Results showed that the optimal number of MZs with regard to the two methods were both five and variance analysis indicated the heterogeneity of soil properties and yield among them. Consequently, it is feasible to use an active canopy sensor to delineate management zones for tobacco-planting fields. [ABSTRACT FROM AUTHOR]

• Adequate biomass is required for cover crops to provide significant ecological benefits. • Satellite imagery predicted field level cover crop biomass accurately. • Satellite imagery more accurately estimated biomass than handheld sensors. • At least one vegetation index for each satellite system accurately predicted biomass. • NDVI was not the best predictor of cover crop biomass for any satellite. Cost-share programs based on measures of participation rather than performance are available to farmers who plant cover crops. However, cover crops only provide significant ecological benefits like reduced nutrient loss when adequate biomass is established. The purpose of this study was to determine whether satellite imagery can effectively estimate cover crop biomass in fields with diverse species composition, and whether increased spatial resolution and satellite imaging frequency can increase biomass estimation accuracy. Aboveground biomass samples of 1 m2 were collected for 86 sites within 26 agricultural fields containing unique cover crop species composition to assess biomass production. In-field sensors were used to measure normalized difference vegetation index (NDVI) and groundcover percentage. Three satellites (Landsat-8 [30 m resolution], Sentinel-2 [10 m resolution], and PlanetScope [3 m resolution]) were used to calculate eight vegetation indices (VIs) for comparison with cover crop biomass. Multiple linear regression, correlation coefficients, and root mean square error (RMSE) were used to perform hierarchical clustering to rank VIs calculated from each satellite for biomass estimation accuracy. Satellites predicted cover crop biomass at the field level very accurately (r2 up to 0.79), demonstrating the potential of large-scale biomass estimation at relatively low cost compared to in-field sampling. All satellite-VI pairs estimated biomass more accurately than the in-field sensors. Performance of VIs varied by satellite, but each satellite had at least one VI that performed very well for both site-level and field-averaged data. When using PlanetScope or Landsat-8 imagery, the perpendicular vegetation index provided the most accurate cover crop biomass estimation on a per-site basis and ratio vegetation index performed best using Sentinel-2 imagery. PlanetScope was the only satellite to provide useable imagery for every site due to increased revisit period; however, its increased spatial resolution did not increase estimation accuracy overall compared to Landsat-8 or Sentinel-2. [ABSTRACT FROM AUTHOR]

Abstract: In this study the suitability of thermal imaging for phenotyping was investigated as part of a breeding experiment carried out by the International Maize and Wheat Improvement Centre (CIMMYT) at Tlaltizapán experimental station in Mexico. Different subtropical maize genotypes with two replications were screened with respect to their tolerance to water stress. Thermal images of the canopy of 92 different maize genotypes were acquired on two different days in the time interval between anthesis and blister stages (grain filling 1), whereby each picture contained five plots of different genotypes and canopy temperatures calculated for each plot. Significantly, lower canopy temperatures were found in well-watered genotypes compared with water-stressed genotypes. Furthermore significant differences (p <0.001) between genotypes under water stress were detected using thermal images. A close correlation (p <0.01–0.001) between canopy temperature or modified Crop water stress index with NDVI and SPAD values was obtained. It may be concluded that genotypes better adapted to drought conditions exhibited lower temperatures. Thermography is a potentially promising method to accelerate the screening process and thereby enhance phenotyping for drought adaptation in maize. [Copyright &y& Elsevier]

Abstract: In this paper we present a segmentation algorithm, inspired from an image-processing region-merging algorithm, for the delineation of discrete contiguous management zones in agriculture. The algorithm is unique in that it is applicable to high- or low-density irregular data sets, such as yield data. The algorithm is described and a brief example presented using unprocessed sensor-derived grain yield data. A comparison between the segmentation algorithm and a common classification algorithm (k-means clustering) was done using an aerial normalised differences vegetation index (NDVI) image collected on a 200ha vineyard in Olite, Southern Navarre, Spain. Classification was performed as a univariate (NDVI) analysis and a spatially constrained analysis. Segmentation and classification were run to find 2, 4, 6, …, 24 levels and the effectiveness of the outputs determined by how well it explained the variance in vine trunk circumference, a correlated but independent measurement. The results obtained demonstrated that for a given number of manageable (effective) zones the segmentation outputs were equivalent or superior to the classification outputs for partitioning vine circumference variance. The segmentation output also generated more coherent management units that should facilitate differential management. The algorithm presented is a first generation segmentation algorithm and several aspects still need to be developed, in particular methods for eliminating edge effects and converting management zones into management (treatment) classes. The results of the segmentation algorithm presented here would indicate that with further development, segmentation might provide an alternative and possibly preferable approach to delineating management zones. [Copyright &y& Elsevier]

Abstract: Operational airborne and satellite remote sensing in agriculture remains constrained by matching platform availability to suitable daytime weather and illumination conditions, crop development, and availability of ground staff. An ultra low-level aircraft carrying an active NIR/Red CropCircle™ sensor was successfully deployed to record and subsequently map crop vigour via the simple ratio (SR) index over a field of sorghum. Given the logging frequency of ≈20Hz and the presence of alternate rows of bare soil, the Moiré effect reduced the contrast between crop and bare soil skip-rows. Such effects would not be expected to occur in non-skip-row crops. The ultra low-level airborne (ULLA)-SR map derived from the 20m transect records compared favorably with the SR map derived from a meter-resolution airborne digital multispectral image that was re-sampled to a similar spatial resolution. This case study, involving a CropCircle™ sensor mounted in a low-level aircraft demonstrates another deployment option for users of this class of sensor. Moreover, an ULLA configuration offers the potential for greater flexibility in scheduling compared to airborne imaging, given it can be flown at any sun-angle, under cloud, at night, and may easily be incorporated into aircraft already conducting low-level operations, for example crop dusting and reconnaissance, over agricultural fields. [Copyright &y& Elsevier]

Land managers are often confronted with management problems that could be addressed using decision support or management information systems. The use of such tools, however, depends on the availability of appropriately resolved spatial data. One source of such data is multi-spectral remotely sensed imagery. The cost of such imagery, particularly where environmental goals are most important, is often prohibitive. Even when the cost of such imagery can be met, other factors, such as cloud cover for satellite-based systems or sensor scheduling for sophisticated airborne systems, may mean that imagery is not available. This paper investigates the utility of a system that combines near-infrared imagery from a video camera with conventional medium-format aerial photography deployed in a light aircraft platform. Previously, imagery obtained from video cameras has suffered from limited spectral range and from significant image motion effects. These problems were eliminated by the use of an electronic-shutter charge-coupled device video camera with a strong IR response. The systems components and the approach to their operational deployment are described and the options for transforming the raw imagery into survey coverage discussed. The image quality and cost is presented for a site characterisation application where the aim is the generation of normalised difference vegetation index values. It is concluded that the system has significant potential utility for decision support and land-management applications. [Copyright &y& Elsevier]

• One increasingly important application for the spatial analysis of cropping systems, is the estimation of CC via remote sensing NDVI to calibrate simulation models. • In order to explore how correlations between NDVI and CC relate to specific crop species and groups of crop types, we conducted a meta-analysis in which we compiled information that correlated remote and proximal sensing NDVI with field observations of CC for different crop species and types. • Our study proposes generic correlations, which are computationally undemanding, for canopy data assimilation into modelling applications for 13 different agricultural crops. • The proposed models may be considered adequate for many applications, as the RMSE are in line with acceptable levels published in several sensitivity analyses, but there is an effect of irregular sampling sizes that must be considered, as well as sources of systematic error that affect extrapolation of results. The fraction of green canopy cover (CC) is an important feature commonly used to characterize crop growth and for calibration of crop and hydrological models. It is well accepted that there is a relation between CC and NDVI through linear or quadratic models, however a straight-forward empirical approach, to derive CC from NDVI observations, is still lacking. In this study, we conducted a meta-analysis of the NDVI-CC relationships with data collected from 19 different studies (N = 1397). Generic models are proposed here for 13 different agricultural crops, and the associated degree of uncertainty, together with the magnitude of error were quantified for each model (RMSE around 6–18% of CC). We observed that correlations are adequate for the majority of crops as R 2 values were above 75% for most cases, and coefficient estimates were significant for most of the linear and quadratic models. Extrapolation to conditions different than those found in the studies may require local validation, as obtained regressions are affected by non-sampling errors or sources of systematic error that need further investigation. In a case study with wheat, we tested the use of NDVI as a proxy to estimate CC and to calibrate the AquaCrop model. Simulation outcomes were validated with field data collected from three growing seasons and confirmed that the NDVI-CC relationship was useful for modelling research. We highlight that the overall applicability of these relationships to modelling is promising as the RMSE are in line with acceptable levels published in several sensitivity analyses. [ABSTRACT FROM AUTHOR]

• An image mining platform based on a simple image segmentation model (GSM) is proposed. • Ground-based nadir images were taken from wheat canopies, under direct sunlight. • Analyzing GSM graph, shading pattern was evaluated and classified quantitatively. • Using GSM and datamining models, various canopy characters were predicted accurately. • GSM is useful for characterizing canopies and high-throughput phenotyping approaches. Efficient quantification of the sophisticated shading patterns inside the 3D vegetation canopies may improve our understanding of canopy functions and status, which is possible now more than ever, thanks to the high-throughput phenotyping (HTP) platforms. In order to evaluate the option of quantitative characterization of shading patterns, a simple image mining technique named " Green-gradient based canopy Segmentation Model (GSM) " was developed based on the relative variations in the level of RGB triplets under different illuminations. For this purpose, an archive of ground-based nadir images of heterogeneous wheat canopies (cultivar mixtures) was analyzed. The images were taken from experimental plots of a two-year field experiment conducted during 2014–15 and 2015–16 growing seasons in the semi-arid region of southern Iran. In GSM, the vegetation pixels were categorized into the maximum possible number of 255 groups based on their green levels. Subsequently, the mean red and the mean blue levels of each group were calculated and plotted against the green levels. It is evidenced that the yielded graph could be readily used for (i) identifying and characterizing canopies even as simple as one or two equation(s); (ii) classification of canopy pixels in accordance with the degree of exposure to sunlight; and (iii) accurate prediction of various quantitative properties of canopy including canopy coverage (CC), Normalized difference vegetation index (NDVI), canopy temperature, and also precise classification of experimental plots based on the qualitative characteristics such as subjection to water and cold stresses, date of imaging, and time of irrigation. The introduced model may provide a multipurpose HTP platform and open new windows to canopy studies. [ABSTRACT FROM AUTHOR]

• A new visible NDVI (vNDVI) was developed using genetic algorithms. • It was evaluated in three different crops (citrus, grapes, and sugarcane). • The vNDVI accurately estimates NDVI values from RGB data. • It can provide a low-cost alternative for plant phenotyping. Several vegetation indices have been developed, with the normalized difference vegetation index (NDVI) been the most studied and commonly used. To generate an NDVI map, a relatively high-cost multispectral sensor is required; but currently, most UAVs are equipped with low-cost RGB cameras. For that reason, other indices that utilize RGB data have been developed to generate maps similar to NDVI and minimize the data acquisition cost, such as the triangular greenness index (TGI) and the visible atmospheric resistant index (VARI). However, several studies found that these indices cannot be recommended as reliable general-purpose crop health indicators. This study utilizes a genetic algorithm to develop a new visible index (visible NDVI; vNDVI) that estimates NDVI values of vegetation from uncalibrated RGB cameras mounted on UAVs (or other remote sensing platforms). Three experiments were conducted to create and validate the proposed index. First, the NDVI values generated from a multispectral camera were compared with the NDVI values generated by a hyperspectral camera. In the second experiment, the vNDVI formula was created using a genetic algorithm. The third experiment validates the proposed vNDVI, generated from two uncalibrated RGB cameras, in three different crops (citrus, grapes, and sugarcane). The proposed vNDVI proved to be highly accurate on estimating NDVI values by just using RGB cameras, with an overall mean percentage error of 6.89% and a mean average error of 0.052 in all three crops, providing a low-cost alternative for remote sensing and plant phenotyping. [ABSTRACT FROM AUTHOR]

• WDRVI improved the accuracy of Beet LAI, FWL, and FWR predictions with Vis. • WDRVI1 (α = 0.01) provides the most accurate estimation of beet growth indicators. • Drone-mount sensor have great potential for estimating plot-level beet growth indicators. The normalized vegetation index (NDVI) is widely used to monitor the spatial, temporal, physiological, and biophysical characteristics of vegetation. However, if the ground biomass is high, the NDVI becomes rapidly saturated. In the leaf area index (LAI) range of 2 to 6, reflectance in the near-infrared band is significantly higher than that in the red band. When reflectance in the near-infrared band exceeds 40%, the contribution of that reflectance to the NDVI is low. Here, we applied a weight coefficient, α (range 0.05 to 0.5), to the near-infrared band, thus developing a wide-dynamic-range vegetation index (WDRVI). We calculated (α * NIR-RED)/(α * NIR + RED) values; these emphasized the Fresh Weight of Beet LAI, the Fresh Weight of Leaves (FWL), and the Fresh Weight of Roots (FWR). We sought a vegetation index that optimally monitored early-stage sugar beet growth. The WDRVI sensitivities using various α coefficients were higher than those of the NDVI in the LAI range of 2 to 6. The determination coefficients (r2 values) of the sugar beet LAI, FWL, and FWT models established using the WDRVI1 were 0.957, 0.950, and 0.963, respectively. Using the WDRVI1 index model to estimate the accuracy of beet growth indicators can improve 1.05% to 5.07%. Use of the WDRVI reduces beet growth indicator saturation when the ground biomass is high, enhancing the accuracy of growth monitoring. This is useful as the sugar beet has huge ground biomass. [ABSTRACT FROM AUTHOR]

• The CNNs are able to reduce crop yield prediction uncertainty considerably. • RGB images perform better over NDVI images. • Sufficient network depth with regularization were required for better performance. Using remote sensing and UAVs in smart farming is gaining momentum worldwide. The main objectives are crop and weed detection, biomass evaluation and yield prediction. Evaluating machine learning methods for remote sensing based yield prediction requires availability of yield mapping devices, which are still not very common among farmers. In this study Convolutional Neural Networks (CNNs) – a deep learning methodology showing outstanding performance in image classification tasks – are applied to build a model for crop yield prediction based on NDVI and RGB data acquired from UAVs. The effect of various aspects of the CNN such as selection of the training algorithm, depth of the network, regularization strategy, and tuning of the hyperparameters on the prediction efficiency are tested. Using the Adadelta training algorithm, L 2 regularization with early stopping and a CNN with 6 convolutional layers, mean absolute error (MAE) in yield prediction of 484.3 kg/ha and mean absolute percentage error (MAPE) of 8.8% was achieved for data acquired during the early period of the growth season (i.e., in June of 2017, growth phase <25%) with RGB data. When using data acquired later in July and August of 2017 (growth phase >25%), MAE of 624.3 kg/ha (MAPE: 12.6%) was obtained. Significantly, the CNN architecture performed better with RGB data than the NDVI data. [ABSTRACT FROM AUTHOR]