High spatial variability of soil properties restricts the benefits of process-oriented modelling for management recommendations on field scale due to rare information about the soil inventory and its distribution. Geo-electrical mapping provides with a certain spatial pattern, but results are influenced by various factors. The model HERMES was applied to 60 of altogether 80 soil sampling points of a well-documented field in North-Rhine Westphalia characterised by a wide range of soil texture. Validation of HERMES resulted in satisfactory root mean square errors (RMSE) for yield (0.76 t ha-1), water (34.70 mm) and nitrogen in soil (51.44 kg ha-1) over the whole simulation period. At the same field, values of conducted electrical conductivity (ECa) mapping ranged from 20 to 90 mS m-1 and have been assigned to the soil sampling points to proof statistical relations. Clay and sand contents of three soil layers were highly correlated with measured ECa values. Derived regression models show R2 values between 0.73 and 0.89. The subsequent calculation of soil texture at points of ECa mapping produced an improved resolution of this key value to initialize model simulation.

An accurate 3D model of an outdoor scene can be used in many different scenarios of precision agriculture, for instance to analyse the silhouette of a tree crown canopy for precision spraying, to count fruit for fruit yield prediction or to simply navigate a vehicle between the plant rows. Instead of using stereovision, limited by the problems of different light intensities, or by using expensive multi-channel 3D range finder (LIDAR scanner), limited by the number of channels, this work investigates the possibility of using two single channel LIDAR scanners mounted on a small robot to allow a real-time 3D object reconstruction of the robot environment. The approach used readings captured by two LIDAR scanners, SICK LMS111 and SICK TiM310, where the first one was scanning horizontally and the second one vertically. In order to correctly map the 3D points of the readings from the vertical sensor into a 3D space, a custom SLAM algorithm based on image registration techniques was used to calculate the new positions of the robot. The approach was tested in an indoor and outdoor environment, proving its accuracy with an error rate of 0.02 m±0.02 m for vertical and −0.01 m±0.13 m for the horizontal plane.

This research was carried out in an experimental vineyard grown in artificially transformed karst terrain (Croatia). The experimental design included four water treatments in three replicates: 1) fully watered or based on 100% evapotranspiration (ETc) application; 2) regulated deficit irrigation based on 75% and 50% ETc applications; and 3) non-watered. Hyperspectral images of grapevines were taken in the summer of 2016 using two spectral-radiance (W·sr−1·m−2) calibrated cameras, covering wavelengths from 409 to 988 nm and 950 to 2509 nm. The four treatments were grouped into two new sets: 1) drought (NO); and 2) watered (the remaining three treatments). The watered group was then split into two new sets: 1) 50% treatment; and 2) 75+100% treatment. The images were analyzed using Partial least squares – discriminant analysis (PLS-DA), and Single Vector Machines (SVM). The PLS-DA demonstrated the capability to determine levels of grapevine drought or watered groups with 70 - 80% accuracy. A similar success rate was achieved in distinguishing the 50% group from the 75+100% group.

Nitrogen (N) losses and ammonia emissions from livestock operations represent a significant water and air pollutant. It has been repeatedly demonstrated that reduction in dietary protein input results in improved N utilization efficiency and reduced N losses with manure in dairy cows. Reducing dietary protein can also benefit the producer by reducing feed cost. These interventions, however, have to be balanced with the risk of decreased productivity. If animal requirements for metabolizable protein (MP) are not met, productivity cannot be sustained. Supplementation with rumen-protected (RP) amino acids (AA) limiting milk production and milk protein synthesis may compensate for the lack of MP in dairy cow diets. Long-term experiments at Penn State University demonstrated reduced urinary N losses, blood urea, milk urea N, and ammonia emissions from manure with MP-deficient v. MP-adequate diets. In some trials, however, milk production and milk protein yield were decreased with the MP-deficient diets. Supplementation of these diets with RPAA (Lys, Met and His) increased dry matter intake and milk yield of the cows to levels comparable with MP-adequate diets and can be a successful strategy for reducing N losses and maintaining productivity in lactating dairy cows fed low-protein diets.

The goal of this work was to cluster maize plants perception points under six different growth stages in noisy 3D point clouds with known positions. The 3D point clouds were assembled with a 2D laser scanner mounted at the front of a mobile robot, fusing the data with the precise robot position, gained by a total station and an Inertial Measurement Unit. For clustering the single plants in the resulting point cloud, a graph-cut based algorithm was used. The algorithm results were compared with the corresponding measured values of plant height and stem position. An accuracy for the estimated height of 1.55 cm and the stem position of 2.05 cm was achieved.

The probabilistic Rasch model is used to get objective measures of production potential at some locations of an olive orchard located in Badajoz, southwestern Spain. Nine soil properties (soil apparent electrical conductivity, clay, sand, and silt content, organic matter, total nitrogen, available phosphorous and potassium, and cation exchange capacity), taken at 40 locations in the field, were considered and, after being integrated in the model, a ranking of all locations according to the soil production potential and the influence on the production potential of each individual soil property were determined. Moreover, those soil samples or properties which had any anomaly where highlighted; this information can be necessary to conduct site-specific treatments, leading to a more cost-effective and sustainable field management. Additionally, estimates using geostatistical algorithms were utilised to map soil production potential and to delineate with a rational basis the management zones in the field.

Infield route planning is used to optimise field operations in order to decrease operational costs and environmental impacts. Route planners must be able to plan operations within real fields and account for real situations such as irregular shapes and obstacles. Therefore, a representative set of fields is required to robustly test the route planner. Instead of choosing randomly, which may result in a non-representative sample of the diversity of fields; a stratification strategy was used to separate the field dataset into strata. Proportional sampling from each stratum provided a representative sample of 217 fields, out of the original set of 603,218 from the Danish field database.

Aflatoxin contamination of food can cause liver cancer in humans and animals. Identification of aflatoxin risk areas allows farmers to adapt management strategies before planting, during growth and at harvest. Aflatoxin contamination is driven by high temperatures and drought conditions and crops grown on light textured soil in the south eastern USA are at particular risk. Aflatoxin assessment is expensive so a role of extension services in precision farming is to identify the areas most at risk of contamination so that farmers can adapt irrigation or planting strategies. This paper extends a county-level risk factors approach developed by Kerry et al. (2017) by investigating the use of NDVI and thermal IR data to indicate drought stress and thus aflatoxin contamination risk at the sub-county level.

This work was part of the EU RobustMilk project. In this work package, we have focused on two aspects of robustness, micro- and macro-environmental sensitivity and applied these to somatic cell count (SCC), one aspect of milk quality. We showed that it is possible to combine both categorical and continuous descriptions of the environment in one analysis of genotype by environment interaction. We also developed a method to estimate genetic variation in residual variance and applied it to both simulated and a large field data set of dairy cattle. We showed that it is possible to estimate genetic variation in both micro- and macro-environmental sensitivity in the same data, but that there is a need for good data structure. In a dairy cattle example, this would mean at least 100 bulls with at least 100 daughters each. We also developed methods for improved genetic evaluation of SCC. We estimated genetic variance for some alternative SCC traits, both in an experimental herd data and in field data. Most of them were highly correlated with subclinical mastitis (>0.9) and clinical mastitis (0.7 to 0.8), and were also highly correlated with each other. We studied whether the fact that animals in different herds are differentially exposed to mastitis pathogens could be a reason for the low heritabilities for mastitis, but did not find strong evidence for that. We also created a new model to estimate breeding values not only for the probability of getting mastitis but also for recovering from it. In a progeny-testing situation, this approach resulted in accuracies of 0.75 and 0.4 for these two traits, respectively, which means that it is possible to also select for cows that recover more quickly if they get mastitis.

This article describes several indicators and methods for estimating nitrogen flows and balance sheets. It reviews their strengths and weaknesses according to user types and their objectives. The nitrogen balance sheets, based on the difference between inputs and outputs are by far the most-often used estimator of nitrogen management on livestock farms. Among those, the soil surface balance helps in guiding the fertilisation and the farmgate balance is logically higher. The indicators of practices are easy to use but are poor predictors of nitrogen losses, whereas the indicators of emissions allow for estimating nitrogen losses using either direct measurements or models thereby providing a dynamic assessment of these flows. The coupling of balance sheets and indicators of emissions offers a certain potential for improvement of diagnosis and support of decision making. Finally, the indicators of impacts, including the life cycle analyses are currently the most popular for qualifying the impacts of animal production.