A Parametric Transfer Function for Grain-Flow Within a Conventional Combine Harvester

Grain yield monitoring is an integral tool in the Precision Agriculture management system. When used in conjunction with a satellite-based navigation system, it provides spatial information on output variability, output response to managed inputs and is used to identify limiting resources in the crop production process. Accurately matching measured yield quantities with spatial units within a field is therefore important. At present, a simple linear time shift is employed by all commercial monitoring systems to account for the delay between GPS recorded positions and subsequent yield measurements. This study examines the internal process of grain transport to the sensor by monitoring the flow of strategically coloured grain. The flow is shown to be significantly influenced by mixing induced by threshing and auger transport processes. In contrast to the common assumption that grain moves as a spatially related cohort through to the sensor, the results suggest that a diffusion process is more realistic. A parametric model for the diffusion process is provided which suggests that from each individual yield measurement a maximum 20% of the mass could be assigned to a single spatial unit of the size that is typically allocated. The results imply that for further analyses, the inconclusive spatial origin and artificially smoothed quantities of instantaneous yield measurements should be considered.     

Modelling long-term in situ soil profile evolution: application to the genesis of soil profiles containing stone layers

Models of soil genesis are potentially of great importance in assessing the effects of global change on ecosystems, and may also contribute to our understanding of soil genetic processes. Many quantitative models have so far focused on individual soil genetic processes and are difficult to extrapolate to the landscape scale. A few attempts have been made to model soil evolution as a whole from a pedologic perspective. This study develops a quantitative model of soil formation at the profile scale, taking into account major soil‐forming processes. These include physical and chemical weathering of primary minerals, strain processes, and bioturbation. The model allows the quantification of the evolution of the particle size, mineral composition and bulk density of the soil. The model is applied with varying values of input parameters, and is compared with actual soil genetic processes. Running the model results in the formation of stone‐layered soil profiles. Stone‐line formation by means of bioturbation, as already described in the literature, seems to be adequately simulated. Planned improvements of the model include implementation of other major soil genetic processes such as leaching, organic matter influence, etc. This model will then have to be implemented spatially considering particularly redistribution processes, to reproduce soil formation at the landscape scale.     

A Two-Factor Empirical Deterministic Response Surface Calibration Model for Site-Specific Predictions of Lime Requirement

This paper describes the development of an empirical deterministic two-factor response surface model for the Woodruff lime-requirement buffer (WRF). The model may be used to produce variable-rate lime requirement maps, or to predict lime requirements in real-time. Hence it may be suitable as a component of a decision support system (DSS) for the site-specific management of acid soil. The models' predictions were compared to those of a one-factor response surface, and those of a linear regression. The models tested were validated against soil-CaCO₃ incubations using a statistical jackknifing procedure for error and bias estimations. The Akaike Information Criterion (AIC) was used to ascertain the best model in terms of goodness of fit and parsimony. The two-factor response surface model produced the best lime requirement estimates, followed by the single-factor model, then the conventional linear regression. The advantages of the response surface models are their improved prediction accuracy, and their flexibility in the choice of any target pH (from pH 5.5 to 7) without the need for excessive calibrations. The uncertainty of the model was assessed using data from an agricultural field in Kelso, New South Wales, Australia. Block kriged maps of soil pH measured in 0.01 M CaCl₂ (pH_CaCl2), WRF buffer pH (pH_buffer) and lime requirements to a target pH of 7 were produced, to compare their spatial distributions. Finally the economic and agronomic benefits of site-specific liming were considered.     

Quantification and comparison of wheat yield variation across space and time

Research about spatio-temporal variation of crop yield does not abound. From a precision agriculture (PA) perspective and particularly considering site-specific crop management (SSCM), this is an aberration. There is a serious need to further question how temporal variation of crop yield impacts ones ability to manage spatial variation. The aim of this study is to consider and develop new and existing approaches to this question. Spatio-temporal analysis was undertaken for two wheat fields in South Australia with 3 and 4 years of wheat yield data. Temporal analysis included the calculation of semi-variance across each field between pairs of years for the creation of maps and the calculation of rank correlations between pairs of years. These analyses supported previous notions that the magnitude of temporal variation is large compared with spatial variation. However, some consistence of spatial patterns between years was also observed for each of the fields indicating that considering magnitudes of variation alone is not an exhaustive analysis. A long-term (100 years) temporal analysis using variograms was undertaken for a single point simulated using the Agricultural Production Simulator Model (APSIM). The long-term analysis overcame the fact that 3 or 4 years of yield data are an extremely small sample size for the time dimension. This analysis provided some useful insight into temporal variation such as a large nugget variance accounting for 75% of the temporal variation and the cyclical nature of temporal yield variation. A novel use of pseudo cross semi-variograms was applied to a spatio-temporal analysis of yield variation for the two fields. This analysis provides a preliminary insight into identifying space–time variance equivalents. With greater depth of temporal crop yield data this is a promising perspective from which to identify optimal spatial management strategies.     

Field-Scale Experiments for Site-Specific Crop Management. Part II: A Geostatistical Analysis

Part II analyses approach C experiments. Field-scale experiments were applied to four wheat fields in the Western Australian wheat belt. Different experimental designs were used two two-dimensional sine-waves, a chequerboard, and a two-factor strip arrangement. In each experiment, the yield associated with a particular treatment was predicted by kriging to where the other treatments were located. Different forms of kriging were investigated. Co-located cokriging, using the previous-season yield map as a covariate, was the most promising. The kriged data were then modelled with polynomial yield response functions. The outcome was a map for each field that described the optimum application of experimental input. The requirements varied continuously across the field, and could justify future site-specific crop management. The two-factor strip experiment was the most successful of those presented; the field on which it was used showed relatively strong responses to the applied inputs. The other sites were affected by lack of rain and/or design flaws. The underlying philosophy is sound, but the method proposed is time-consuming and inefficient. We hope that this paper can stimulate further research on the subject.     

A continuum approach to soil classification by modified fuzzy k-means with extragrades

The need for a more continuous approach to soil classification is discussed, and methods based on the mathematical constructs known as fuzzy sets are considered most appropriate for this. A centroidal grouping method, fuzzy k-means with extragrades, which quantifies the intragrading and extragrading of soil individuals is described. An example of the application of this technique to an area of 4800 ha at Wesepe in The Netherlands is presented. The results show that the technique could create a classification that reflects the main pedological features of the area in a continuous way. Although there may be problems in selecting the optimal number of groups and degree of fuzziness, we conclude that the method is most promising and worthy of consideration when any type of quantitative soil classification is required.     

Models relating soil pH measurements in water and calcium chloride that incorporate electrolyte concentration

Soil pH is the most routinely measured soil property for assessing plant nutrient availability. Nevertheless, there are various techniques for soil pH measurement, which vary with regard to the solution used and the soil‐to‐solution ratio. Soil pH is commonly measured in water or 0.01 m CaC1₂. Soil pH in CaCl₂ is usually preferred as it is less affected by soil electrolyte concentration and provides a more consistent measurement. Therefore there is a need to convert measurement values between the two methods. Previous models reported linear and curvilinear relationships between the two measurements. However, the pH difference between measurements in water and CaCl₂ is related to the soil solution electrolyte concentration. We observed that the pH difference between the two methods became smaller with increasing soil electrical conductivity (EC). We therefore developed models that relate pH in CaCl₂ and water and incorporate EC values. We calibrated a linear and a non‐linear model (artificial neural networks, ANN) using 9817 soil samples from Queensland, Australia. Soil pH in water and CaCl₂ and EC were measured with a 1:5 soil‐to‐solution ratio. The results show that incorporating EC in the prediction model improves the prediction of pH in CaCl₂ significantly. We validated these models using 4576 independent samples obtained from a diverse range of soils across Australia. Although the linear and ANN models performed similarly, the ANN (which has a curvilinear relationship) provided a better prediction and aligns with the theory that for acid and alkaline pH values, the difference between pH in water and CaCl₂ is less than that for pHs between 4.5 and 7.     

Creating Field Extent Digital Elevation Models for Precision Agriculture

This paper presents a comparison of techniques for the interpolation of elevation data collected using a global positioning system. The techniques examined included global and local regression models and kriging of the residuals, global and local kriging, and the TOPOGRID tool in Arc Info. The results demonstrated the superiority of the TOPOGRID tool.     

Evaluation of a local regression kriging approach for mapping apparent electrical conductivity of soil (ECa) at high resolution

Apparent electrical conductivity of soil (ECa) is a property frequently used as a diagnostic tool in precision agriculture, and is measured using vehicle‐mounted proximal sensors. Crop‐yield data, which is measured by harvester‐mounted sensors, is usually collected at a higher spatial density compared to ECa. ECa and crop‐yield maps frequently exhibit similar spatial patterns because ECa is primarily controlled by the soil clay content and the interrelated soil moisture content, which are often significant contributors to crop‐yield potential. By quantifying the spatial relationship between soil ECa and crop yield, it is possible to estimate the value of ECa at the spatial resolution of the crop‐yield data. This is achieved through the use of a local regression kriging approach which uses the higher‐resolution crop‐yield data as a covariate to predict ECa at a higher spatial resolution than would be prudent with the original ECa data alone. The accuracy of the local regression kriging (LRK) method is evaluated against local kriging (LK) and local regression (LR) to predict ECa. The results indicate that the performance of LRK is dependent on the performance of the inherent local regression. Over a range of ECa transect survey densities, LRK provides greater accuracy than LK and LR, except at very low density. Maps of the regression coefficients demonstrated that the relationship between ECa and crop yield varies from year to year, and across a field. The application of LRK to commercial scale ECa survey data, using crop yield as a covariate, should improve the accuracy of the resultant maps. This has implications for employing the maps in crop‐management decisions and building more robust calibrations between field‐gathered soil ECa and primary soil properties such as clay content.