Doctor of Philosophy / Precision Agriculture (PA) strives towards holistic production and environmental management. A fundamental research challenge is the continuous expansion of ideas about how PA can contribute to sustainable agriculture. Some associated pragmatic research challenges include quantification of spatio-temporal variation of crop yield; crop growth simulation modelling within a PA context and; evaluating long-term financial and environmental outcomes from site-specific crop management (SSCM). In Chapter 1 literature about managing whole farms with a mind towards sustainability was reviewed. Alternative agricultural systems and concepts including systems thinking, agro-ecology, mosaic farming and PA were investigated. With respect to environmental outcomes it was found that PA research is relatively immature. There is scope to thoroughly evaluate PA from a long-term, whole-farm environmental and financial perspective. Comparatively, the emphasis of PA research on managing spatial variability offers promising and innovative ways forward, particularly in terms of designing new farming systems. It was found that using crop growth simulation modelling in a PA context is potentially very useful. Modelling high-resolution spatial and temporal variability with current simulation models poses a number of immediate research issues. This research focused on three whole farms located in Australia that grow predominantly grains without irrigation. These study sites represent three important grain growing regions within Australia. These are northern NSW, north-east Victoria and South Australia. Note-worthy environmental and climatic differences between these regions such as rainfall timing, soil type and topographic features were outlined in Chapter 2. When considering adoption of SSCM, it is essential to understand the impact of temporal variation on the potential value of managing spatial variation. Quantifying spatiotemporal variation of crop yield serves this purpose; however, this is a conceptually and practically challenging undertaking. A small number of previous studies have found that the magnitude of temporal variation far exceeds that of spatial variation. Chapter 3 of this thesis dealt with existing and new approaches quantifying the relationship between spatial and temporal variability in crop yield. It was found that using pseudo cross variography to obtain spatial and temporal variation ‘equivalents’ is a promising approach to quantitatively comparing spatial and temporal variation. The results from this research indicate that more data in the temporal dimension is required to enable thorough analysis using this approach. This is particularly relevant when questioning the suitability of SSCM. Crop growth simulation modelling offers PA a number of benefits such as the ability to simulate a considerable volume of data in the temporal dimension. A dominant challenge recognised within the PA/modelling literature is the mismatch between the spatial resolution of point-based model output (and therefore input) and the spatial resolution of information demanded by PA. This culminates into questions about the conceptual model underpinning the simulation model and the practicality of using point-based models to simulate spatial variability. iii The ability of point-based models to simulate appropriate spatial and temporal variability of crop yield and the importance of soil available water capacity (AWC) for these simulations were investigated in Chapter 4. The results indicated that simulated spatial variation is low compared to some previously reported spatial variability of real yield data for some climate years. It was found that the structure of spatial yield variation was directly related to the structure of the AWC and interactions between AWC and climate. It is apparent that varying AWC spatially is a reasonable starting point for modelling spatial variation of crop yield. A trade-off between capturing adequate spatio-temporal variation of crop yield and the inclusion of realistically obtainable model inputs is identified. A number of practical solutions to model parameterisation for PA purposes are identified in the literature. A popular approach is to minimise the number of simulations required. Another approach that enables modelling at every desired point across a study area involves taking advantage of high-resolution yield information from a number of years to estimate site-specific soil properties with the inverse use of a crop growth simulation model. Inverse meta-modelling was undertaken in Chapter 5 to estimate AWC on 10- metre grids across each of the study farms. This proved to be an efficient approach to obtaining high-resolution AWC information at the spatial extent of whole farms. The AWC estimates proved useful for yield prediction using simple linear regression as opposed to application within a complex crop growth simulation model. The ability of point-based models to simulate spatial variation was re-visited in Chapter 6 with respect to the exclusion of lateral water movement. The addition of a topographic component into the simple point-based yield prediction models substantially improved yield predictions. The value of these additions was interpreted using coefficients of determination and comparing variograms for each of the yield prediction components. A result consistent with the preceding chapter is the importance of further validating the yield prediction models with further yield data when it becomes available. Finally, some whole-farm management scenarios using SSCM were synthesised in Chapter 7. A framework that enables evaluation of the long-term (50 years) farm outcomes soil carbon sequestration, nitrogen leaching and crop yield was established. The suitability of SSCM across whole-farms over the long term was investigated and it was found that the suitability of SSCM is confined to certain fields. This analysis also enabled identification of parts of the farms that are the least financially and environmentally viable. SSCM in conjunction with other PA management strategies is identified as a promising approach to long-term and whole-farm integrated management.

The feasibility of using optical sensors and non-nitrogen limiting reference crops to determine post-emergent nitrogen fertilizer requirements of canola was evaluated. Normalized difference vegetation index was well suited for estimating yield potential and nitrogen status. Although sensor-based nitrogen management was generally agronomically feasible for canola, the economic benefits of doing so remain uncertain because of the added cost of applying post-emergent nitrogen. / February 2008

The collection of crop yield data has become much easier with the introduction of technologies such as the Global Positioning System (GPS), ground-based yield sensors, and Geographic Information Systems (GIS). This explosive growth and widespread use of spatial data has challenged the ability to derive useful spatial knowledge. In addition, outlier detection as one important pre-processing step remains a challenge because the technique and the definition of spatial neighbourhood remain non-trivial, and the quantitative assessments of false positives, false negatives, and the concept of region outlier remain unexplored. The overall aim of this study is to evaluate different spatial outlier detection techniques in terms of their accuracy and computational efficiency, and examine the performance of these outlier removal techniques in a site-specific management context. In a simulation study, unconditional sequential Gaussian simulation is performed to generate crop yield as the response variable along with two explanatory variables. Point and region spatial outliers are added to the simulated datasets by randomly selecting observations and adding or subtracting a Gaussian error term. With simulated data which contains known spatial outliers in advance, the assessment of spatial outlier techniques can be conducted as a binary classification exercise, treating each spatial outlier detection technique as a classifier. Algorithm performance is evaluated with the area and partial area under the ROC curve up to different true positive and false positive rates. Outlier effects in on-farm research are assessed in terms of the influence of each spatial outlier technique on coefficient estimates from a spatial regression model that accounts for autocorrelation. Results indicate that for point outliers, spatial outlier techniques that account for spatial autocorrelation tend to be better than standard spatial outlier techniques in terms of higher sensitivity, lower false positive detection rate, and consistency in performance. They are also more resistant to changes in the neighbourhood definition. In terms of region outliers, standard techniques tend to be better than spatial autocorrelation techniques in all performance aspects because they are less affected by masking and swamping effects. In particular, one spatial autocorrelation technique, Averaged Difference, is superior to all other techniques in terms of both point and region outlier scenario because of its ability to incorporate spatial autocorrelation while at the same time, revealing the variation between nearest neighbours. In terms of decision-making, all algorithms led to slightly different coefficient estimates, and therefore, may result in distinct decisions for site-specific management. The results outlined here will allow an improved removal of crop yield data points that are potentially problematic. What has been determined here is the recommendation of using Averaged Difference algorithm for cleaning spatial outliers in yield dataset. Identifying the optimal nearest neighbour parameter for the neighbourhood aggregation function is still non-trivial. The recommendation is to specify a large number of nearest neighbours, large enough to capture the region size. Lastly, the unbiased coefficient estimates obtained with Average Difference suggest it is the better method for pre-processing spatial outliers in crop yield data, which underlines its suitability for detecting spatial outlier in the context of on-farm research.

The purpose of this study was to determine factors influencing cotton producer adoption of Precision Agriculture in the cotton belt according to members of the American Cotton Producers of the National Cotton Council. The National Research Council’s Board on Agriculture defines Precision Agriculture (PA) as “a management strategy that uses information technologies to bring data from multiple sources to bear on decisions associated with crop production.” For the purpose of this study, Precision Agriculture technologies included yield monitors, global positioning units, variable rate applicators, and similar components. Many studies have found that adoption of Precision Agriculture can be profitable for agricultural producers. However, the fact that Precision Agriculture is relatively new and unproven hinders rapid adoption by agricultural producers. According to the National Research Council Board of Agriculture widespread adoption relies on economic gains outweighing the costs of the technology. This study attempted to find the factors associated with adoption of these technologies in the cotton belt. The sample population consisted of cotton producer representatives from the leading cotton-producing states. A Delphi approach was utilized to establish a consensus of cotton producer perceptions of the advantages of adopting Precision Agriculture technologies. Advantages included more accurate farming (i.e., row spacing, reduced overlap, and cultivation). Barriers to adoption were also documented, questioning employee capability to operate equipment, learning curve, technology complexity, and uncertain return on investment.

Two wild blueberry fields were selected to evaluate the impact of variable rate (VR) fertilization on crop productivity, surface and subsurface water quality. Management zones were delineated based on slope variability, and different fertilizer rates were applied according to prescription maps. Runoff collectors were place in the fields to measure the nutrient losses in surface runoff, while lysimeters were installed to evaluate the impact of VR fertilization (VRF) on subsurface water quality. The VR treatment significantly decreased phosphorus and nitrogen loadings in surface runoff as compared to uniform treatment. The concentrations of nutrients in subsurface water samples were also significantly lower for VR treatment as compared to uniform treatment. The excessive nutrients enhanced vegetative growth in low lying areas of uniform fertilization, while berry yield was less. Based on these results, it can be concluded that VRF in wild blueberry fields improved the crop productivity and potential environmental impacts. / This study was conducted to evaluate the impact of variable rate fertilization on crop productivity and surface and subsurface water quality in wild blueberry fields. Result illustrated that variable rate fertilization significantly reduce the nutrients loading in surface and subsurface water, and improved blueberry yield.

The feasibility of using optical sensors and non-nitrogen limiting reference crops to determine post-emergent nitrogen fertilizer requirements of canola was evaluated. Normalized difference vegetation index was well suited for estimating yield potential and nitrogen status. Although sensor-based nitrogen management was generally agronomically feasible for canola, the economic benefits of doing so remain uncertain because of the added cost of applying post-emergent nitrogen.

The popularity of variable-rate technology (VRT) has grown. However, the limitations and errors ofthis technology are generally unknown. Therefore, a spatial data model was developed to generate "asapplied"surfaces to advance precision agricultural (PA) practices. A test methodology based on ASAEStandard S341.2 was developed to perform uniform-rate (UR) and variable-rate (VR) tests to characterizedistribution patterns testing four VRT granular applicators (two spinner spreaders and two pneumaticapplicators). Single-pass UR patterns exhibited consistent shapes for three of the applicators with patternsshifts observed for the fourth applicator. Simulated overlap analysis showed that three of the applicatorsperformed satisfactorily with most CVs less than 20% while one applicator performed poorly (CVs andgt;25%). The spinner spreaders over-applied at the margins but the pneumatic applicators under-appliedsuggesting a required adjustment to the effective swath spacing. Therefore, it is recommended that CVsaccompany overlap pattern plots to ensure proper calibration of VRT application.Quantification of the rate response characteristics for the various applicators illustrated varying delayand transition times. Only one applicator demonstrated consistent delay and transition times. A sigmoidalfunction was used to model the rate response for applicators. One applicator exhibited a linear responseduring a decreasing rate change. Rate changes were quicker for the two newer VR control systemssignifying advancement in hydraulic control valve technology. This research illustrates the need forstandard testing protocols for VRT systems to help guide VRT software developers, equipmentmanufacturers, and users.The spatial data model uses GIS functionality to merge applicator descriptive patterns with a spatialfield application file (FAF) to generate an 'as-applied' surface representing the actual distribution ofgranular fertilizer. Field data was collected and used to validate the "as-applied" spatial model.Comparisons between the actual and predicted application rates for several fields were madedemonstrating good correlations for one applicator (several R2 andgt; 0.70), moderate success for anotherapplicator (0.60 andlt; R2 andlt; 0.66), and poor relationships for the third applicator (R2 andlt; 0.49). A comparison ofthe actual application rates to the prescription maps generated R2 values between 0.16 and 0.81demonstrating inconsistent VRT applicator performance. Thus, "as-applied" surfaces provide a means toproperly evaluate VRT while enhancing researchers' ability to compare VR management approaches.

Yield monitors have become an indispensable part of precision agriculture systemsbecause of their ability to measure the yield variability. Accurate yield monitor data availabilityis essential for the assessment of farm practices. The current technology of measuring grainyields is prone to errors that can be attributed to mass flow variations caused by the mechanismswithin a grain combine. Because of throughput variations, there are doubts regarding thecorrelation between the mass flow measurement and the actual grain volume produced at aspecific location. Another inaccuracy observed in yield monitor data can be attributed to inexactcut-widths values entered by the machine operator.To effectively address these yield monitor errors, two crop mass flow sensing deviceswere developed and used to correct yield monitor data. The two quantities associated with cropmaterial mass flow that were sensed were tension on the feeder housing drive chain and thehydraulic pressure on the threshing cylinder's variable speed drive. Both sensing approacheswere capable of detecting zero mass flow conditions better than the traditional grain mass flowsensor. The alternative sensors also operate without being adversely affected by materialtransport delays. The feeder housing-based sensor was more sensitive to variations in cropmaterial throughput than the hydraulic pressure sensor. Crop mass flow is not a surrogate forgrain mass flow because of a weak relationship (R2 andlt; 0.60) between the two quantities. The cropmass flow signal does denote the location and magnitude of material throughput variations intothe combine. This delineation was used to redistribute grain mass flow by aligning grain andcrop mass flow transitions using sensor fusion techniques. Significant improvements (?? = 0.05)in yield distribution profile were found after the correction was applied.To address the cut-width entry error, a GIS-based post-processing algorithm wasdeveloped to calculate the true harvest area for each yield monitor data point. Based on theresults of this method, a combine operator can introduce yield calculation errors of 15%. Whenthese two correction methods applied to yield monitor data, the result is yield maps withdramatically improved yield estimates and enhanced spatial accuracy.

In the recent past precision farming has become increasingly popular amongfarmers. However, little has been done to study the business aspect of precision farming,with most research focusing on the production side. This purpose of this thesis is tostudy the feasibility of successfully opening and operating a precision farming firm inKentucky. To determine the feasibility of such a venture a computer model was createdand a producer survey was designed and distributed to farmers in Western and CentralKentucky.The purpose of the computer model was to determine the factors that wouldinfluence the successful operation of a precision farming firm including number of acresserviced per year, pricing of services, the cost of capital to borrow money, and manyother factors. A break-even analysis was performed to determine what kind of annualincreases in business would be required, what price range services should be in, and atwhat interest rate money could be borrowed and a simulated precision farming firm couldstill operate successfully.The producer survey was mailed to 336 farmers in Western and Central Kentuckybecause of their geographical locations and the type of crops that are grown there. Thesurvey response rate was 20 percent and of the 66 surveys that were returned 59 wereappropriate and useful for research. After compiling the results of the surveys,regressions were run to determine any correlation between dependent and independentvariables that affect the adoption rate of precision farming techniques. The results foundthat a negative correlation exists between age adoption rates of precision farming and thata positive correlation exists between farm size and adoption rates of precision farming.After conducting the research, it is believed that given the right economicconditions and management a precision farming firm is very capable of thriving inKentucky. However, the workforce must be very motivated and capable of constantlyrecruiting new clients to adopt precision farming.

The paper describes major features of an unmanned aerial vehicle, designed undersafety and performance requirements for missions of aerial photography and remotesensing in precision agriculture. Unmanned aerial vehicles have vast potential asobservation and data gathering platforms for a wide variety of applications. The goalof the project was to develop a small, low cost, electrically powered, unmanned aerialvehicle designed in conjunction with a payload of imaging equipment to obtainremote sensing images of agricultural fields. The results indicate that this conceptwas feasible in obtaining high quality aerial images.