Multisensory precision agriculture - improving yields and reducing environmental impact

We are able to separate between stress related to water and N status in wheat plants (workpackage (WP) 1, part 1) by combining temperature measurements (thermal camera) and spectral measurements (spectra-radiometer). The use of thermal cameras is challenging, and we have invested resources into the development of methods and calibration procedures. This work is published. Now we have obtained more data, and recent findings reveals that we are able to separate between the same two stressors by means of spectral measurements alone. These findings are now about to be published. The measurements of N2O-emissions in the field (WP1, part 2) have shown that even marginal changes in pH may have large effects on the emission rates, and that chamber measurements should be performed with parallels (in addition to replicates), due to large, spatial variation. Further, we have found large diurnal variation in measured gas-flux rates, and earlier findings of large emissions related to freezing/thawing events have been confirmed. Moreover, the results appear to support the hypothesis that correct N fertilization leads to lower N2O-emissions per unit of produced grain, than too high or too low rates of applied fertilizer N. The robot being developed in the project (WP2) is now ready, along with the chambers used for the N2O-measurement. The measurements so far have revealed that the response time of the system is much faster than that assumed by traditional methods. The system developed is now being thoroughly tested under field conditions, and field trials are established, involving pH-adjustment by means of selected minerals, where the robot will be used for monitoring N2O-emissions (in cooperation with the project MIGMIN). The robot has become considerable attention internationally, and we have established intention agreements with ICOS and other research groups for producing more robots. The instrument for multiple gas measurements (FTIR) performed well during the initial tests, and a new prototype is now built. The prototype will be tested further. It appears that troublesome, perennial weeds (thistle, sow-thistle and quackgrass) in wheat may be possible to map automatically by means of image analysis (WP3). So far, we have simulated images taken from a combined harvester or a tractor, with a RGB-camera mounted on a 3 m large pole, or mounted on a ground based robot (developed in WP2). In both cases the images were taken at an angle of 45º off nadir. Algorithms for projecting the images into a nadir view are developed, and they work well. Manual weed mapping with GPS (± 10 cm accuracy) is performed on populations, from which images were taken the previous year. The time of these mappings was selected to coincide with the normal time for weed control. This data represents the basis for optimizing the image analysis. Imaging of thistle, sow-thistle and quackgrass have been performed by using the robot. Some analyses of these data remain. In the process of developing methods for pre-symptomatic disease detection (WP4), VOC-profiles are collected from Fusarium species and pathogens involved in the Norwegian leaf blotch complex in wheat. There are significant differences in VOC-profiles between these two groups. Further, we have analysed the VOC-production from both healthy and inoculated wheat plants. Two weeks after plants were inoculated with powdery mildew, they produced four separate chemical compounds, which cannot be detected in the surroundings of healthy plants. For grown-up plants, we have found that mellein was produced from plants inoculated with Septoria leaf blotch only, and that sativen was produced from plants inoculated with Fusarium head blight only. In cooperation with SINTEF, we have optimized the measuring of these compounds, when they are present in low concentrations close to plants grown in a green house. In parallel to this, we have shown that it is possible to detect pre-symptomatic fungal infections with optical sensors. To investigate effects of co-occurring multiple stressors (WP5) under controlled conditions, several project partners cooperated in performing a pot experiment at Hohenheim University. The experiment was run twice, and the wheat plants were measured frequently with seven different sensors (soon to be published). Thereafter, we performed a field trial following a similar design as the pot experiment. In the field trial, we utilized the robot developed in WP2 to perform sensor measurements in an effective way. The data is still subject to analyses.

In this project we will develop and exploit new multi-sensor techniques to reduce the N20-emmissions from wheat production, by increasing the efficiency of fertilizer, weed-, and disease control. To do so we will follow the principle of precision agricult ure, i.e. the application of inputs reflects the spatial variability within a field. Starting with fertilizer, we intend to improve an existing system for site-specific application, which is unable to distinguish between the spectral characteristics of pl ants deficient in N from those related to water stress. The effects of site-specific fertilizer management on the N2O-emissions will also be documented. To increase the currently low efficiency of N2O-flux measurements, we will construct a programmable, u nmanned ground vehicle equipped with a gas monitoring system. This will greatly enhance our ability to quantify annual N2O emissions, as affected by treatments. The work on weeds comprises construction and calibration of image analysis for perennial weed species detection, and translation of sensed data into site-specific control actions. Discriminative power of the image analysis under varying field conditions is the most critical factor. Increased efficiency in disease control will be obtained by pre-sy mptomatic disease detection, based on induced chlorophyll fluorescence and volatile organic compounds (VOC) profile determination. One critical aspect here will be whether recent developments of e.g. electronic noses will make the detection of fungal VOC profiles possible in the field. Finally, we intend to combine our human and technological resources and perform comprehensive integrated analyses to unravel significant interactive effects on plant performance which should be accounted for in an optimized system designed to reduce the environmental footprint of food production. Both, public and private industry sectors as well as the civil society at large may benefit from the project results.