Subsoil compaction is a major yield limiting factor for most agricultural crops. Tillage is the most efficient method to quickly treat compacted subsoil, but it is also expensive, increases erosion, and accelerates nutrient cycling. The use of real-time electrical conductivity (EC) and near-infrared (NIR) reflectance values to differentiate compacted areas from uncompacted areas was studied. This method has potential to reduce monetary and time investments inherent in traditional grid sampling and the resultant deep tillage of an entire field. EC and NIR reflectance are both very sensitive to spatial variability of soil attributes. The objective of this research was to determine whether the amount of soil moisture affects the efficacy of EC and NIR spectroscopy (at 2151.9 nm) in identifying subsoil compaction through correlation analysis, and also to determine whether a minimum level of compaction was necessary for these same methods to detect compaction in three different soil textures across a variable water gradient. Bulk density measurements were taken in late 2007 from plots traversing an induced soil moisture gradient, and low, medium, and high levels of compaction at three locations with different soil textures. A Veris Technologies (Salina, KS) Near-Infrared Spectrophotometer equipped with an Electrical Conductivity Surveyor 3150 was used to measure and geo-reference EC and NIR reflectance data over the same plots. Analysis of the data for a correlation between compaction (bulk density values) and EC, as well as compaction and NIR reflectance, produced clear results. It was found that electrical conductivity is not significantly different between compacted or uncompacted soils even when tested at all moisture extremes and in different soil textures in Utah. Also, NIR spectroscopy was unsuccessful at identifying subsoil compaction because all tested procedures to induce a spectrometer into the soil resulted in changes the physical properties of the soil.

At no point in history have we demanded so much from our agricultural land whilst simultaneously leaving so little room for management error. Of the many possible environmental impacts from agriculture, soil and water salinisation has some of the most long-lived and deleterious effects. Despite its importance, however, land managers are often unable to make informed decisions of how to manage the risk of salinisation due to a lack of data. Furthermore, there remains no universally agreed method for salinity risk mapping. This thesis addresses these issues by investigating new methods for producing high-resolution predictions of soil salinity, soil physical properties and groundwater depth using a variety of traditional and emerging ancillary data sources. The results show that the methodologies produce accurate predictions yielding natural resource information at a scale and resolution not previously possible. Further to this, a new methodology using fuzzy logic is developed that exploits this information to produce high-resolution salinity risk maps designed to aid both agricultural and natural resource management decisions. The methodology developed represents a new and effective way of presenting salinity risk and has numerous advantages over conventional risk models. The incorporation of fuzzy logic provides a meaningful continuum of salinity risk and allows for the incorporation of uncertainty. The method also allows salinity risk to be calculated relative to any vegetation community and shows where the risk is coming from (root-zone or groundwater) allowing more appropriate management decisions to be made. The development of this methodology takes us a step closer to closing what some have called our greatest gap in agricultural knowledge. That is, our ability to manage the salinity risk at the subcatchment scale.

The recommended procedures for implementing COTMAN; a cotton management expert system; suggest frequent crop scouting at numerous locations for each field. Machine based height sensors coupled with the ability to spatially record height values make it possible to locate regions of a field that are height representative of the entire field. A machine based height measurement system called HMAP was used to assess plant height in various fields in the 2003 growing season while the same fields were monitored with COTMAN. The plant height data was used to determine an optimal COTMAN sampling scheme for each field consisting of significantly fewer sampling locations than recommended by COTMAN. It was possible to ascertain equivalent information from COTMAN using two sites selected from height data in place of six sites selected per COTMAN recommendations. The HMAP system was extended to monitor rate of growth in real time in addition to plant height by comparing historical plant height data recorded on previous field passes to current height values. The rate of growth capable HMAP system will make it possible to track cotton growth and development with an automated system.

The overarching goal of this dissertation project was to address several fundamental aspects of applying site-specific crop management for fiber quality in cotton production. A two-year (2005 and 2006) field study was conducted at the IMPACT Center, a portion of the Texas A&M Research farm near College Station, Texas, to explore the spatial variability of cotton fiber quality and quantify its relationship with in-season soil moisture content. Cotton samples and in-situ soil moisture measurements were taken from the sampling locations in both irrigated and dry areas. It was found that generally low variability (CV < 10%) existed for all of the HVI (High Volume Instrument) fiber parameters under investigation. However, an appreciable level of spatial dependence among fiber parameters was discovered. Contour maps for individual fiber parameters in 2006 exhibited a similar spatial pattern to the soil electrical conductivity map. Significant correlations (highest r = 0.85) were found between most fiber parameters (except for micronaire) and in-season soil moisture in the irrigated areas in 2005 and in the dry area in 2006. In both situations, soil moisture late in the season showed higher correlation with fiber parameters than that in the early-season. While this relationship did not hold for micronaire, a non-linear relationship was apparent for micronaire in 2006. This can be attributed to the boll retention pattern of cotton plants at different soil moisture levels. In addition, a prototype wireless- and GPS-based system was fabricated and developed for automated module-level fiber quality mapping. The system is composed of several subsystems distributed among harvest vehicles, and the main components of the system include a GPS receiver, wireless transceivers, and microcontrollers. Software was developed in C language to achieve GPS signal receiving, wireless communication, and other auxiliary functions. The system was capable of delineating the geographic boundary of each harvested basket and tracking it from the harvester basket to the boll buggy and the module builder. When fiber quality data are available at gins or classing offices, they can be associated with those geographic boundaries to realize fiber quality mapping. Field tests indicated that the prototype system performed as designed. The resultant fiber quality maps can be used to readily differentiate some HVI fiber parameters (micronaire, color, and loan value) at the module level, indicating the competence of the system for fiber quality mapping and its potential for site-specific fiber quality management. Future improvements needed to make system suitable for a full-scale farming operation are suggested.

The overarching goal of this dissertation project was to address several fundamental aspects of applying site-specific crop management for fiber quality in cotton production. A two-year (2005 and 2006) field study was conducted at the IMPACT Center, a portion of the Texas A&M Research farm near College Station, Texas, to explore the spatial variability of cotton fiber quality and quantify its relationship with in-season soil moisture content. Cotton samples and in-situ soil moisture measurements were taken from the sampling locations in both irrigated and dry areas. It was found that generally low variability (CV < 10%) existed for all of the HVI (High Volume Instrument) fiber parameters under investigation. However, an appreciable level of spatial dependence among fiber parameters was discovered. Contour maps for individual fiber parameters in 2006 exhibited a similar spatial pattern to the soil electrical conductivity map. Significant correlations (highest r = 0.85) were found between most fiber parameters (except for micronaire) and in-season soil moisture in the irrigated areas in 2005 and in the dry area in 2006. In both situations, soil moisture late in the season showed higher correlation with fiber parameters than that in the early-season. While this relationship did not hold for micronaire, a non-linear relationship was apparent for micronaire in 2006. This can be attributed to the boll retention pattern of cotton plants at different soil moisture levels. In addition, a prototype wireless- and GPS-based system was fabricated and developed for automated module-level fiber quality mapping. The system is composed of several subsystems distributed among harvest vehicles, and the main components of the system include a GPS receiver, wireless transceivers, and microcontrollers. Software was developed in C language to achieve GPS signal receiving, wireless communication, and other auxiliary functions. The system was capable of delineating the geographic boundary of each harvested basket and tracking it from the harvester basket to the boll buggy and the module builder. When fiber quality data are available at gins or classing offices, they can be associated with those geographic boundaries to realize fiber quality mapping. Field tests indicated that the prototype system performed as designed. The resultant fiber quality maps can be used to readily differentiate some HVI fiber parameters (micronaire, color, and loan value) at the module level, indicating the competence of the system for fiber quality mapping and its potential for site-specific fiber quality management. Future improvements needed to make system suitable for a full-scale farming operation are suggested.

The ability to map the profit made across a cotton field would enable producers to see in detail where money is being made or lost on their farms. This ability, which requires sitespecific knowledge of yield, fiber quality, and input costs would further enable them to implement precise field management practices to ensure that they receive the highest return possible on each portion of a field and do not waste materials and other inputs throughout the field. Investigators at Texas A&M previously developed a wireless-GPS system that tracks where a module of cotton comes from within a field. This system is a necessary component in mapping fiber quality, which is a major determiner of price and thus profit. Three drawbacks to the previous wireless-GPS system are that (1) a person must manually trigger the system to send wireless communications when a field machine dumps its load of cotton, (2) multiple field machines of the same type (e.g., two cotton pickers) cannot be used simultaneously on the same system within the same field, and (3) no software is available to automatically produce fiber-quality maps after the data are downloaded from the gin. The first two drawbacks, the need for an automatic communication-triggering system and the needed capability for multiple field machines of the same type are the problems addressed in this work. To solve the first problem, a sensing and control system was added to a harvester to automatically indicate when the machine is dumping a basket load of cotton so that wireless messages can be automatically sent from the harvester to subsequent field machines without human intervention. This automated communication-triggering system was incorporated into the existing wireless- GPS system, rigorously field tested, and ultimately proven to operate as designed. Linking data collected with this system together with classing information will enable producers to create fiber-quality maps, and linking fiber-quality maps with yield and input-cost maps will enable them to create profit maps. Additionally, a radio-frequency identification (RFID) system was integrated with the wireless-GPS system to allow for multiple field machines of the same type. The RFID system was also rigorously field tested and proven to operate as designed. Finally, the entire system was field tested as a whole and operated according to design. Thus, the wireless-GPS module tracking system now operates without human intervention and works with multiple field machines of each type, two additional capabilities required for practical use in large farming operations.

The use of site-specific data can enhance management decisions in the field. Three different uses of site-specific data were evaluated and their outcomes are promising. Historical yield data from yield monitors and height data from the HMAP (plant height mapping) system were used to select representative areas within the field, and areas of average conditions were used as sampling sites for COTMAN, a cotton management expert system. This proved to be effective, with predicted cutout dates and date of peak nodal development similar to the standard COTMAN approach. The HMAP system was combined with historical height data for variable rate application of mepiquat chloride, based on the plant growth rate. The system performance was evaluated, but weather conditions in 2004 did not allow a true evaluation of varying mepiquat chloride. A series of multi-spectral images were normalized utilizing the soil line transformation (SLT) technique and normalized difference vegetation index (NDVI) was calculated from the transformed images, from the raw image and for the true reflectance images. The SLT technique was effective in tracking the change in true reflectance NDVI in some images, but not all. Changes to the soil line extraction program are suggested so that it more effectively determines soil lines.

The recommended procedures for implementing COTMAN; a cotton management expert system; suggest frequent crop scouting at numerous locations for each field. Machine based height sensors coupled with the ability to spatially record height values make it possible to locate regions of a field that are height representative of the entire field. A machine based height measurement system called HMAP was used to assess plant height in various fields in the 2003 growing season while the same fields were monitored with COTMAN. The plant height data was used to determine an optimal COTMAN sampling scheme for each field consisting of significantly fewer sampling locations than recommended by COTMAN. It was possible to ascertain equivalent information from COTMAN using two sites selected from height data in place of six sites selected per COTMAN recommendations. The HMAP system was extended to monitor rate of growth in real time in addition to plant height by comparing historical plant height data recorded on previous field passes to current height values. The rate of growth capable HMAP system will make it possible to track cotton growth and development with an automated system.

Precision agriculture (PA) can be defined as a set of technologies that have helped propel agriculture into the computerized information-based world, and is designed to help farmers get greater control over the management of farm operations. Because of its potential to spatially reduce yield variability within the field through variable rate application of nutrients it is thought to be a production risk management instrument. Subsurface drip irrigation (SDI) is another production risk management technology that is generating interest from the farming community as a result of new technological improvements that facilitate equipment maintenance and reduces water consumption.In the first article the production risk management potential of these two technologies was investigated both for each technology and for a combination of the two. Simulated yield data for corn, wheat and soybeans were obtained using EPIC, a crop growth simulation model. Mathematical programming techniques were used in a standard E-V framework to reproduce the production environment of a Kentucky commercial grain farmer in Henderson County. Results show that for risk averse farmers, the lowest yield variability was obtained with the SDI technology. The highest profit level was obtained when the two technologies were combined.Investment in two sets of equipments (PA and SDI) to maximize profitability and reduce risk could however expose many farm operations to financial risk. In the second article, a discrete stochastic sequential programming (DSSP) model was used to analyze the impact of PA and/or SDI equipment investment on the farm's liquidity and debt to asset ratio.In the last article, the cotton sector in Benin, West Africa, was utilized to study the transferability of PA technology to a developing country. Properly introduced, precision agriculture (PA) technology could help farmers increase profitability, improve management practices, and reduce soil depletion. An improved production system could also help farmers better cope with the policy risk related to cotton production. Results from the two models show that PA is less profitable for the risk neutral farmer but more profitable for the risk averse one when compared to conventional production practices. The adoption of the new technology also has very little impact on the choice of crop rotation made by the farmer.

This thesis examines the use of precision agriculture data, specifically yield maps, for makingsite-specific economic decisions for improved farm management. The adoption of precisionagriculture on farms has allowed producers to collect a greater quantity and more specificinformation about production than ever before. With such information, site-specific decisions canbe made. Incorporating economic data with yield map data, two primary decision examples aredeveloped: defining areas of production and nonproduction and managing temporal risk spatiallyacross a field. Included with the production/ nonproduction decision are the effects that landtenure arrangements and risk aversion levels have on the decision. The risk maps are developedusing break-even analysis, the coefficient of variation, and a mean-variance framework, all based ona twenty year average of temporal net returns, measured spatially. The risk maps are repeatedincorporating a crop insurance option, a commonly used risk management tool. Results show thatdeveloping these maps can be used by agricultural producers to help with their decision making. Byincorporating these maps into the decision-making process, decisions can be made to increase farmprofitability.