Many plant diseases have distinct visual symptoms which can be used to identify and classify them correctly. This paper presents a potato disease classification algorithm which leverages these distinct appearances and the recent advances in computer vision made possible by deep learning. The algorithm uses a deep convolutional neural network training it to classify the tubers into five classes, four diseases classes and a healthy potato class. The database of images used in this study, containing potatoes of different shapes, sizes and diseases, was acquired, classified, and labelled manually by experts. The models were trained over different train-test splits to better understand the amount of image data needed to apply deep learning for such classification tasks.

The precise application of pesticides to fruit crops requires information regarding the tree or vine canopy as a system input in order to control the amount of liquid and air being applied. Variations in canopy volume and density occur due to variety, trellis system, growth stage, training system and season. Current practice is to occasionally change liquid volume but seldom to change airflow. This paper details the development and validation of an ultrasonic sensor system to measure not only canopy volume but also canopy density and presence. Sensors fitted to the sprayer can record, in real time, changes in crop characteristics as the sprayer moves along the row. Signals can then send information to variable output nozzles and adjustable air fans. Trials have been conducted and results have proven to be extremely reliable and accurate. The ability to precisely control the spray results in the optimum application rate, leading to better results in the use of pesticides, less environmental pollution (less drift and less leaf runoff) and improved economic viability for the fruit grower.

In agriculture, tank mixtures with several plant protection products (PPP) are frequently used, which makes it impossible to apply specific plant protection products to specific areas. Therefore a prototype of a field sprayer with a direct injection system was developed as part of a joint project between the company Herbert Dammann GmbH and the Julius Kühn- Institute. This framework enables a real-time site-specific dosing of plant protection products and water from separate containers with direct mixing before the application without residual amounts. Results show a high accuracy in the working range of the selected dosing pumps by means of test bench measurements and field tests Field tests for site-specific weed control in winter wheat demonstrate the applicability of the system under practical conditions and enable a precise application of various herbicides.

Assessment of Tar Spot Complex (TSC) severity in maize breeding experiments is conducted visually and may sometimes result in inconsistencies due to human interpretation. Disease scoring using remote sensing technologies may help bring more precision to the phenotyping process. An experiment for assessment of grain yield losses due to TSC was conducted at the Aguafria Experimental Station of the International Center for Wheat and Maize Improvement – CIMMYT in Mexico. Twenty-five maize genotypes were planted in spring of 2016 under a fungicide treatment to control TSC development and no fungicide treatment in a square lattice design with three replications. Four flights were carried out using an Unmanned Aerial Vehicle (UAV) equipped with a multispectral (550, 660, 735, 790 nm) and a thermal camera, simultaneously with the visual disease scorings and the yield was measured after harvesting. The preliminary results of the study indicated that the use of remote sensing in disease resistance phenotyping may be as effective as visual disease scoring since both correlate highly with the grain yield. Structural and chlorophyll vegetation indices (VIs) proved to be a good alternative for the estimation of yield losses caused by TSC in experimental field conditions, which may be potentially used for screening for resistance to this disease in maize genotypes, hypothetically reducing the need for visual disease scoring in the field.

A ground-based hyperspectral imaging system covering the spectral range of 384–1034 nm was used for Sclerotinia Stem Rot (SSR) detection. Two sample sets of oilseed leaves were collected. Four vegetation indices were extracted and evaluated by analysis of variance (ANOVA) combined with linear discriminant analysis (LDA) for the two sample sets. Discriminant models were built using the 4 vegetation indices. The discriminant results of the two sample sets were good with classification accuracies of the calibration set and the prediction set over 85%. The overall results indicated that vegetation indices calculated from ground-based hyperspectral imaging could be used as reliable and accurate indices for SSR detection.

In Spain, the use of annual cover crops is a crop management practice for irrigated vineyards that allows controlling vineyard vigor and yield, which also leads to improve the crop quality. Recently, Cynodon dactylon (bermudagrass) has been reported to infest those cover crops and colonize the grapevine rows, resulting in significant yield and economic losses due to the competition for water and nutrients. From timely unmanned aerial vehicle (UAV) imagery, the objective of this research was to map C. dactylon patches in order to provide an optimized site-specific weed management. A quadrocopter UAV equipped with a point-and-shoot camera was used to collect a set of aerial red-green-blue (RGB) images over a commercial vineyard plot, coinciding with the dormant period of C. dactylon (February 2016). Object-based image analysis (OBIA) techniques were used to develop an innovative algorithm for early discrimination and mapping of C. dactylon, which had the ability to solve the limitation of spectral similarity of this weed with cover crops or bare soil. As a general result, the classified maps of the studied vineyard showed four main classes, i.e. vine, cover crop, C. dactylon and bare soil, with 85% overall accuracy. These weed maps allow developing new strategies for site-specific control of C. dactylon populations in the context of precision viticulture.

The goal of the project was to supply growers with knowledge on how incorporation of machine vision technology can affect the wild blueberry crop, disease pressures, and the overall savings of select agrochemical inputs. A machine vision system was developed and mounted on a rear sprayer boom in front of the sprayer nozzles capable of targeting the agrochemical application on an as-needed basis. Results showed that plants that received the proper fungicide application were less prone to premature leaf drop resulting in larger stem diameters and higher bud counts and harvestable fruit yield. Fungicide application savings using the smart sprayer for spot-application was 12% as compared to a uniform application.

Soil application of herbicides for preemergence (PRE) weed control in grain sorghum is vital to control weeds. Efficacy of soil-applied herbicides is impacted by herbicide adsorption which is influenced by soil organic matter (SOM) and texture. With precision agriculture technologies, variable rate applications (VRA) can be utilized to maximize herbicide effectiveness. In 2016, algorithms were developed for two locations to use VRA of two tank mixed herbicides based on SOM and soil electrical conductivity (EC) collected by a Veris MSP3 system. Drone imagery provided an effective way to evaluate the efficacy of herbicide applications along with visual assessment. VRA applications of herbicide tank mixes provided equal weed control compared to flat rate applications.

The WEED-PAM® is a chlorophyll fluorescence sensor. It has already been applied in the detection of herbicide resistant Alopecurus mysuroides populations with promising results. Yet more work needs to be done in order to validate the system’s capabilities in different species. In this study, field experiments were conducted at three sites to clarify the capability of this sensor to detect herbicide resistant Apera spica-venti populations. The plants were treated with five different herbicides: three ALS-, one ACCase- and one PS II- inhibitor. Five days after the herbicide treatment, sensor data were gathered. These data were compared with a visual assessment, performed 21 days after the herbicide application. The populations that exhibited a strong resistance to the ALS and PS II inhibitors could be differentiated from the sensitive ones. Yet the Apera spica-venti population with a low resistance level to the ACCase inhibitors cannot be differentiated from the sensitive population.

Nitrogen fertilization of silage maize in Central Italy is typically carried out with two applications at early stages of crop development: 2nd (V2) and 6th (V6) leaf respectively. In such conditions, the crop has not yet fully covered the soil and proximal or remote sensing of the canopy is hindered by the strong soil background signal. There is thus great interest in rapid and inexpensive approaches to N fertilization prescription. Therefore, an indirect method for inferring information on yield potential and soil variability, through a field-based clustering of multi-temporal satellite data, has been developed using archive Landsat images to identify temporally constant patterns. This method is potentially useful for the creation of prescription maps. The usefulness of the method was evaluated during an N fertilisation field trial in Maccarese (Central Italy), in 2016. At the V2 stage, both uniform and variable rate applications were performed and compared. A pseudo-cross variogram and a standardized ordinary co-kriging methodology was used to highlight spatially variable significant differences among the treatments.

There is evidence that well managed winter cereal cover crops can scavenge a goodly amount of post summer cereal harvest residual nitrogen (N), reducing nitrate-N losses to leaching or runoff. The objective of this study was to compare nitrate-N phytoremediation areas derived from five sources of information: site specific, non-site specific, or a combination. The non-site specific source was a single “composite” soil nitrate sample. The site specific sources were: a) a dense soil nitrate-N grid sampling; and b) a N removal map calculated from yield and grain N concentration, both determined at the same grid density as soil nitrate-N. The source combinations were: a) a yield map and a single grain N concentration value taken from published information; and b) a yield map and a single field “composite” grain N concentration value. The results indicated that the published grain N value was inferior to measured grain N values, and that the maize (Zea mays L.) yield map best serves as a stratification tool, delineating similar crop performance areas. Random soil sampling within those areas further optimizes residual nitrate-N recovery management. Site specific technologies can guide establishment of N scavenging cover crops to simultaneously improve resource use efficiency and water quality.

Optimising oilseed rape canopy size through correct management is crucial for maximising yield. Plant growth regulators (PGRs) and nitrogen (N) fertiliser are generally applied at a flat rate, however variable applications may be useful for the optimisation of canopy size. The aim of this paper was to understand the potential for spectral reflectance indices to predict green area index (GAI) and crop N content in winter oilseed rape, with specific focus on the Fritzmeier Isaria Crop Sensor. Three large oilseed rape chessboard experiments were set up in 2015 and 2016 in the UK. The results show good correlations between the Isaria indices and both GAI and crop N content, suggesting that the Isaria may be a useful tool for variably applying PGRs and N fertiliser to oilseed rape.

A range of precision farming technologies are used commercially for variable rate applications of nitrogen (N) for cereals, yet these usually adjust N rates from a pre-set value, rather than predicting economically optimal N requirements on an absolute basis. This paper reports chessboard experiments set up to examine variation in N requirements, and to develop and test systems for its prediction, and to assess its predictability. Results showed very substantial variability in fertiliser N requirements within fields, typically >150 kg ha−1, and large variation in optimal yields, typically >2 t ha−1. Despite this, calculated increases in yield and gross margin with N requirements perfectly matched across fields were surprisingly modest (compared to the uniform average rate). Implications are discussed, including the causes of the large remaining variation in grain yield, after N limitations were removed.

Field vegetables require large amounts of N and precision N management (PNM) may help increase their productivity, quality and profitability while reducing N leaching in the biosphere. Few studies investigated PNM for field vegetables. This may be explained by the great diversity in crops and cultivars which complicate the broad implementation of PNM discoveries. Field vegetables are often grown in histosols, which have unique properties such as quantity of N mineralized and the spatial pattern of organic soil depth. Finally, research gaps exist in the development of decision support systems adapted to field vegetable crops grown in histosols.

Crop growth models including CERES-Maize and CROPGRO-Soybean have been used in the past to evaluate causes of spatial yield variability and to evaluate economic consequences of variable rate prescriptions. However, these modelling techniques have not been widely used because of an absence of user-friendly software. In this work, a nitrogen prescription model to simulate the consequences of different nitrogen prescriptions using the DSSAT crop growth models is developed. The objective is to describe a site-specific nitrogen prescription and economic optimizer program developed for computing optimum spatial nitrogen rates for maize using the CERES-Maize model. The application of the model is demonstrated on two different fields in Germany and the US. The program simulated optimum N applications that averaged 42% (McGarvey field, US) and 39% (Riech field, Germany) lower than the uniform rates actually applied in the fields. The software is written in Python and will ultimately be distributed in the public domain as a plug-in to the QGIS software.

Little is known about the effectiveness of sensor-based variable rate nitrogen (VRN) fertilizer application at reducing nitrate-N losses. A variable rate in-season nitrogen (VRN) treatment was compared to a conventional pre-plant uniform rate nitrogen treatment in eight Minnesota, USA sub-fields. Water samples collected weekly during 2016 were analyzed for nitrate-N. The Drainmod model was able to accurately represent measured tile discharge and nitrate-N concentration data. Modeled nitrate-N losses in tile discharge were 16% lower in VRN than uniform treatments. These results show that in-season VRN fertilizer N management can significantly reduce nitrate-N loads from tile drained maize fields.

This work proposes a methodology that uses remote sensing (RS) images to obtain optimum nitrogen crop uptake (Nuptake) maps, for the all pixels in the image included in the field during the entire growing season. The Nuptake was determined from relationship between critical nitrogen concentration (Nc) and biomass where biomass was estimated by a crop growth model based on the water use efficiency. The paper proposes the use of this methodology in commercial wheat farm. The results are discussed with respect to field measurements of crop biomass and N concentration on different dates and in zones with different nitrogen treatments from 8 commercial wheat farms in Albacete, Spain during 2015 and 2016.

Over-application of nitrogen (N) in rice (Oryza sativa L.) production in China is common, leading to low N use efficiency (NUE) and high environmental risks. The objective of this work was to evaluate the ability of the CERES-Rice crop growth model to simulate N response in the cool climate of Northeast China, with the long term goal of using the model to develop optimum N management recommendations. Nitrogen experiments were conducted from 2011–2015 in Jiansanjiang, Heilongjiang Province in Northeast China. The CERES-Rice model was calibrated for 2014 and 2015 and evaluated for 2011 and 2013 experiments. Overall, the model gave good estimations of yield across N rates for the calibration years (R2=0.89) and evaluation years (R2=0.73). The calibrated model was then run using weather data from 2001–2015 for 20 different N rates to determine the N rate that maximized the long term marginal net return (MNR) for different N prices. The model results indicated that the optimum mean N rate was 120–130 kg N ha–1, but that the simulated optimum N rate varied each year, ranging from 100 to 200 kg N ha–1. Results of this study indicated that the CERES-Rice model was able to simulate cool season rice growth and provide estimates of optimum regional N rates that were consistent with field observations for the area.

Simple active radiometer sensors, such as RapidScan, enable agronomic decision-making and phenotyping within commercial wheat fields and experiments. The objectives of this study were: 1 - to evaluate the accuracy of quantitative biomass and nitrogen uptake estimation by the RapidScan, and 2 - to evaluate yield loss estimation based on NDVI. The RapidScan sensor was used as an assessment tool in the following studies: (i) over 3 years, 518 wheat samples were monitored during the vegetative growth period for biomass and aboveground nitrogen uptake and (ii) wheat cultivars were tested in an additional 4 field experiments, which were scanned weekly and correlated with grain yield. Results showed that accurate biomass estimation is limited up to about 100 g DM m−2. Grain yield, actual and potential, estimation is highly affected by the emergence date. The results showed that the use of a proximal-sensing technique allows for rapid and accurate crop monitoring and yield estimation, but emphasizing limitations in future use as well.

Intensive agriculture has the objective to increase nutrients use efficiency. Nitrogen (N) is a key nutrient for crops and the estimations of crop N status allow adjusting the fertilization levels to crop requirements, while reducing the environmental costs and optimizing the benefits for farmers. In this work the N status of wheat in a commercial plot has been monitored, varying the N supply taking into account the variability of the soil. The N content in the cover has been monitored simultaneously by sampling at field level and by using vegetation indices based on reflectance in the red-edge band. The results of the field campaign along a crop growth cycle show that the REP, MTCI, AIVI and CCCI calculated from narrow spectral bands show good linear correlations (R2>0.93) with respect to N content (g·m−2). These indices are stable when passing to broad bands as the case of Sentinel 2 with R2>0.9.