Monitoring the dynamics in wheat crops requires near-term observations with high spatial resolution due to the complex factors influencing wheat growth variability. We studied the prospects for monitoring the biophysical parameters and nitrogen status in wheat crops with low-cost imagery acquired from unmanned aerial vehicles (UAV) over an 11 ha field. Flight missions were conducted at approximately 50 m in altitude with a commercial copter and camera system—three missions were performed between booting and maturing of the wheat plants and one mission after tillage. Ultra-high resolution orthoimages of 1.2 cm·px⁻¹ and surface models were generated for each mission from the standard red, green and blue (RGB) aerial images. The image variables were extracted from image tone and surface models, e.g., RGB ratios, crop coverage and plant height. During each mission, 20 plots within the wheat canopy with 1 × 1 m² sample support were selected in the field, and the leaf area index, plant height, fresh and dry biomass and nitrogen concentrations were measured. From the generated UAV imagery, we were able to follow the changes in early senescence at the individual plant level in the wheat crops. Changes in the pattern of the wheat canopy varied drastically from one mission to the next, which supported the need for instantaneous observations, as delivered by UAV imagery. The correlations between the biophysical parameters and image variables were highly significant during each mission, and the regression models calculated with the principal components of the image variables yielded R² values between 0.70 and 0.97. In contrast, the models of the nitrogen concentrations yielded low R² values with the best model obtained at flowering (R² = 0.65). The nitrogen nutrition index was calculated with an accuracy of 0.10 to 0.11 NNI for each mission. For all models, information about the surface models and image tone was important. We conclude that low-cost RGB UAV imagery will strongly aid farmers in observing biophysical characteristics, but it is limited for observing the nitrogen status within wheat crops.

Soil pH is a key parameter for crop productivity, therefore, its spatial variation should be adequately addressed to improve precision management decisions. Recently, the Veris pH Manager^TM, a sensor for high-resolution mapping of soil pH at the field scale, has been made commercially available in the US. While driving over the field, soil pH is measured on-the-go directly within the soil by ion selective antimony electrodes. The aim of this study was to evaluate the Veris pH Manager^TM under farming conditions in Germany. Sensor readings were compared with data obtained by standard protocols of soil pH assessment. Experiments took place under different scenarios: (a) controlled tests in the lab, (b) semicontrolled test on transects in a stop-and-go mode, and (c) tests under practical conditions in the field with the sensor working in its typical on-the-go mode. Accuracy issues, problems, options, and potential benefits of the Veris pH Manager^TM were addressed. The tests demonstrated a high degree of linearity between standard laboratory values and sensor readings. Under practical conditions in the field (scenario c), the measure of fit (r²) for the regression between the on-the-go measurements and the reference data was 0.71, 0.63, and 0.84, respectively. Field-specific calibration was necessary to reduce systematic errors. Accuracy of the on-the-go maps was considerably higher compared with the pH maps obtained by following the standard protocols, and the error in calculating lime requirements was reduced by about one half. However, the system showed some weaknesses due to blockage by residual straw and weed roots. If these problems were solved, the on-the-go sensor investigated here could be an efficient alternative to standard sampling protocols as a basis for liming in Germany.

Deforestation and draining of the peatlands in equatorial SE Asia has greatly increased their flammability, and in September–October 2015 a strong El Niño-related drought led to further drying and to widespread burning across parts of Indonesia, primarily on Kalimantan and Sumatra. These fires resulted in some of the worst sustained outdoor air pollution ever recorded, with atmospheric particulate matter (PM) concentrations exceeding those considered “extremely hazardous to health” by up to an order of magnitude. Here we report unique in situ air quality data and tropical peatland fire emissions factors (EFs) for key carbonaceous trace gases (CO₂, CH₄ and CO) and PM_2.5 and black carbon (BC) particulates, based on measurements conducted on Kalimantan at the height of the 2015 fires, both at locations of “pure” sub-surface peat burning and spreading vegetation fires atop burning peat. PM_2.5 are the most significant smoke constituent in terms of human health impacts, and we find in situ PM_2.5 emissions factors for pure peat burning to be 17.8 to 22.3 g·kg⁻¹, and for spreading vegetation fires atop burning peat 44 to 61 g·kg⁻¹, both far higher than past laboratory burning of tropical peat has suggested. The latter are some of the highest PM_2.5 emissions factors measured worldwide. Using our peatland CO₂, CH₄ and CO emissions factors (1779 ± 55 g·kg⁻¹, 238 ± 36 g·kg⁻¹, and 7.8 ± 2.3 g·kg⁻¹ respectively) alongside in situ measured peat carbon content (610 ± 47 g-C·kg⁻¹) we provide a new 358 Tg (± 30%) fuel consumption estimate for the 2015 Indonesian fires, which is less than that provided by the GFEDv4.1s and GFASv1.2 global fire emissions inventories by 23% and 34% respectively, and which due to our lower EF_CH4 produces far less (~3×) methane. However, our mean in situ derived EF_PM2.5 for these extreme tropical peatland fires (28 ± 6 g·kg⁻¹) is far higher than current emissions inventories assume, resulting in our total PM_2.5 emissions estimate (9.1 ± 3.5 Tg) being many times higher than GFEDv4.1s, GFASv1.2 and FINNv2, despite our lower fuel consumption. We find that two thirds of the emitted PM_2.5 come from Kalimantan, one third from Sumatra, and 95% from burning peatlands. Using new geostationary fire radiative power (FRP) data we map the fire emissions’ spatio-temporal variations in far greater detail than ever before (hourly, 0.05°), identifying a tropical peatland fire diurnal cycle twice as wide as in neighboring non-peat areas and peaking much later in the day. Our data show that a combination of greatly elevated PM_2.5 emissions factors, large areas of simultaneous, long-duration burning, and very high peat fuel consumption per unit area made these Sept to Oct tropical peatland fires the greatest wildfire source of particulate matter globally in 2015, furthering evidence for a regional atmospheric pollution impact whose particulate matter component in particular led to millions of citizens being exposed to extremely poor levels of air quality for substantial periods.

Weed maps should be available quickly, reliably, and with high detail to be useful for site-specific management in crop protection and to promote more sustainable agriculture by reducing pesticide use. Here, the optimization of a deep residual convolutional neural network (ResNet-18) for the classification of weed and crop plants in UAV imagery is proposed. The target was to reach sufficient performance on an embedded system by maintaining the same features of the ResNet-18 model as a basis for fast UAV mapping. This would enable online recognition and subsequent mapping of weeds during UAV flying operation. Optimization was achieved mainly by avoiding redundant computations that arise when a classification model is applied on overlapping tiles in a larger input image. The model was trained and tested with imagery obtained from a UAV flight campaign at low altitude over a winter wheat field, and classification was performed on species level with the weed species Matricaria chamomilla L., Papaver rhoeas L., Veronica hederifolia L., and Viola arvensis ssp. arvensis observed in that field. The ResNet-18 model with the optimized image-level prediction pipeline reached a performance of 2.2 frames per second with an NVIDIA Jetson AGX Xavier on the full resolution UAV image, which would amount to about 1.78 ha h⁻¹ area output for continuous field mapping. The overall accuracy for determining crop, soil, and weed species was 94%. There were some limitations in the detection of species unknown to the model. When shifting from 16-bit to 32-bit model precision, no improvement in classification accuracy was observed, but a strong decline in speed performance, especially when a higher number of filters was used in the ResNet-18 model. Future work should be directed towards the integration of the mapping process on UAV platforms, guiding UAVs autonomously for mapping purpose, and ensuring the transferability of the models to other crop fields.

Stripe rust (Pst) is a major disease of wheat crops leading untreated to severe yield losses. The use of fungicides is often essential to control Pst when sudden outbreaks are imminent. Sensors capable of detecting Pst in wheat crops could optimize the use of fungicides and improve disease monitoring in high-throughput field phenotyping. Now, deep learning provides new tools for image recognition and may pave the way for new camera based sensors that can identify symptoms in early stages of a disease outbreak within the field. The aim of this study was to teach an image classifier to detect Pst symptoms in winter wheat canopies based on a deep residual neural network (ResNet). For this purpose, a large annotation database was created from images taken by a standard RGB camera that was mounted on a platform at a height of 2 m. Images were acquired while the platform was moved over a randomized field experiment with Pst-inoculated and Pst-free plots of winter wheat. The image classifier was trained with 224 × 224 px patches tiled from the original, unprocessed camera images. The image classifier was tested on different stages of the disease outbreak. At patch level the image classifier reached a total accuracy of 90%. To test the image classifier on image level, the image classifier was evaluated with a sliding window using a large striding length of 224 px allowing for fast test performance. At image level, the image classifier reached a total accuracy of 77%. Even in a stage with very low disease spreading (0.5%) at the very beginning of the Pst outbreak, a detection accuracy of 57% was obtained. Still in the initial phase of the Pst outbreak with 2 to 4% of Pst disease spreading, detection accuracy with 76% could be attained. With further optimizations, the image classifier could be implemented in embedded systems and deployed on drones, vehicles or scanning systems for fast mapping of Pst outbreaks.

Weed detection with aerial images is a great challenge to generate field maps for site-specific plant protection application. The requirements might be met with low altitude flights of unmanned aerial vehicles (UAV), to provide adequate ground resolutions for differentiating even single weeds accurately. The following study proposed and tested an image classifier based on a Bag of Visual Words (BoVW) framework for mapping weed species, using a small unmanned aircraft system (UAS) with a commercial camera on board, at low flying altitudes. The image classifier was trained with support vector machines after building a visual dictionary of local features from many collected UAS images. A window-based processing of the models was used for mapping the weed occurrences in the UAS imagery. The UAS flight campaign was carried out over a weed infested wheat field, and images were acquired between a 1 and 6 m flight altitude. From the UAS images, 25,452 weed plants were annotated on species level, along with wheat and soil as background classes for training and validation of the models. The results showed that the BoVW model allowed the discrimination of single plants with high accuracy for Matricaria recutita L. (88.60%), Papaver rhoeas L. (89.08%), Viola arvensis M. (87.93%), and winter wheat (94.09%), within the generated maps. Regarding site specific weed control, the classified UAS images would enable the selection of the right herbicide based on the distribution of the predicted weed species.