Search

Soil organic carbon (SOC) is believed to play a crucial role for many soil functions and ecosystem services. Despite much research, a lower threshold of SOC for sustainable crop production has not been identified across soil types. We addressed a comprehensive dataset with yields of winter wheat for nearly a 1000 field nitrogen (N) response experiments performed over 3 decades in Denmark. The soils ranged from coarse sand to clay loam (0.016–0.406 kg clay kg1 minerals) with SOC ranging from 0.005 to 0.039 kg kg1 minerals. All experiments were located on arable soil that was fully fertilized with all other plant nutrients. Based on the yield response at 4–6 levels of applied mineral N, we estimated the potential yield not restricted by any nutrient, Ypot, the yield at no applied N, YN0, the marginal agronomic efficiency (AEN0) in terms of the increase in yield per unit of applied N at zero N, and the level of mineral N, Nopt, needed to obtain Ypot. YN0 increased with Fines20 (proportion of soil minerals < 20 μm). AEN0 decreased with SOC independent of soil type. Nopt decreased with increase in SOC. Ypot increased significantly with a squared expression of Fines20 and decreased significantly with SOC. Ypot was better explained by RH¼Fines20/SOC than SOC per se. Our results indicate a positive effect of SOC in terms of a reduction of mineral N needed to obtain the potential crop yield, and a negative effect on Ypot.We suggest amultigate regulation concept for future studies addressing the non-N effect of SOC. We further draw attention to the importance of the quality of the SOC for better understanding effects on crop yields.

Realistic estimation of grain nitrogen (N; N in grain yield) is crucial for assessing N management in crop rotations, but there is little information on the performance of commonly used crop models for simulating grain N. Therefore, the objectives of the study were to (1) test if continuous simulation (multi-year) performs better than single year simulation, (2) assess if calibration improves model performance at different calibration levels, and (3) investigate if a multi-model ensemble can substantially reduce uncertainty in reproducing grain N. For this purpose, 12 models were applied simulating different treatments (catch crops, CO2 concentrations, irrigation, N application, residues and tillage) in four multi-year rotation experiments in Europe to assess modelling accuracy. Seven grain and seed crops in four rotation systems in Europe were included in the study, namely winter wheat, winter barley, spring barley, spring oat, winter rye, pea and winter oilseed rape. Our results indicate that the higher level of calibration significantly increased the quality of the simulation for grain N. In addition, models performed better in predicting grain N of winter wheat, winter barley and spring barley compared to spring oat, winter rye, pea and winter oilseed rape. For each crop, the use of the ensemble mean significantly reduced the mean absolute percentage error (MAPE) between simulations and observations to less than 15%, thus a multi–model ensemble can more precisely predict grain N than a random single model. Models correctly simulated the effects of enhanced N input on grain N of winter wheat and winter barley, whereas effects of tillage and irrigation were less well estimated. However, the use of continuous simulation did not improve the simulations as compared to single year simulation based on the multi-year performance, which suggests needs for further model improvements of crop rotation effects.

The presented work is aimed at designing a system that will model and recognize actions and its interaction with objects. Such a system is aimed at facilitating robot task learning. Activity modeling and recognition is very important for its potential applications in surveillance, human-machine interface, entertainment, biomechanics etc. Recent developments in neuroscience suggest that all actions are a compositions of smaller units called primitives. Current works based on primitives for action recognition uses a supervised framework for specifying the primitives. We propose a method to extract primitives automatically. These primitives are to be used to generate actions based on certain rules for combining. These rules are expressed as a stochastic context free grammar. A model merging approach is adopted to learn a Hidden Markov Model to t the observed data sequences. The states of the HMM approximates local properties of a long sequence. Observation sequences are used to learn a model expressing the data in a structured way. Based on the learned model, recurring parts in the sequences are identied. Primitives that make up the observation sequences are identied as the recurring and unique parts appearing in the sequences. Extracted primitives are used to make a primitive graph from which a grammar for the observed primitives are derived. This method is further extended to include object context. We observe that human actions and objects can be seen as being intertwined: we can interpret actions from the way the body parts are moving, but as well from how their eect on the involved object. While human movements can look vastly dierent even under minor changes in location, orientation and scale, the use of the object can provide a strong invariant for the detection of motion primitives. Movements that produce the same state change in the object state space are classied to be instances of the same action primitive. This allows us to dene action primitives as sets of movements where the movements of each primitive are connected through the object state change they induce.

There has been a recent interest in segmenting action sequences into meaningful parts (action primitives) and to model actions on a higher level based on these action primitives. Unlike previous works where action primitives are defined a-priori and search is made for them later, we present a sequential and statistical learning algorithm for automatic detection of the action primitives and the action grammar based on these primitives. We model a set of actions using a single HMM whose structure is learned incrementally as we observe new types. Actions are modeled with sufficient number of Gaussians which would become the states of an HMM for an action. For different actions we find the states that are common in the actions which are then treated as an action primitive.