Phosphate rocks (PRs) can be used as phosphorus (P) fertilizer sources in agricultural systems. Under certain conditions, farmers can apply them in order to supply P to crops at a lower cost than water-soluble P fertilizer. However, before applying PR, a farmer will probably ask questions such as: "Will PRs work on my land?"; and "Would a water-soluble P fertilizer be more costeffective?" Researchers who have worked with PRs will respond by saying: "It will depend on the type of PR that you use and the environmental conditions that occur in your fields". In fact, many factors determine whether a given PR will be an effective P fertilizer in a farmers field. The way these factors interact to influence PR performance is complex for any specific condition. Thus, it is difficult to make general technical recommendations.
Decision-support systems (DSSs) are simple tools that enable research and extension personnel to provide technical recommendations and decision support to farmers. A DSS for PR use (PR-DSS) utilizes available information on the above factors to predict whether a given PR will be effective in a given crop environment. This chapter discusses different types of DSSs for predicting PR performance, including the approaches used to develop a PR-DSS in New Zealand and Australia. It presents examples that illustrate these approaches. Finally, it outlines steps that are being undertaken in the development a more global DSS by FAO, the International Fertilizer Development Center (IFDC) and the International Atomic Energy Agency (IAEA) for use in tropical and subtropical countries for a range of food crops.
Farmers need to know whether the use of PR will: (i) be worthwhile in supplying P to a deficient soil; and (ii) will save money compared with using water-soluble P fertilizer. They could obtain this information from an expert experienced in PR use. However, there are limitations on accessing the information on these technical recommendations. This chapter contends that a DSS is the most effective way of integrating the main factors that determine PR effectiveness, and then of indicating whether the PR will be effective in farmers fields.
A DSS is an interactive computer-based system that helps decision-makers utilize data and models to solve unstructured problems (Sprague and Carlson, 1982). The main aim of such a system is to improve the performance of decision-makers while reducing the time and human resources required for analysing complex situations. A PR-DSS will be able to predict whether a particular PR will be effective in providing P to the crop in the field. A number of these DSSs have been developed to predict PR performance in different environments.
In practice, a PR will be agronomically effective in a farming system if it is able to dissolve fast enough to provide plant-available P at a rate that is adequate for crop growth. Thus, it is a question of whether the rates of PR dissolution can match the rates of P demand by crops. Various authors have reviewed the conditions necessary for this to occur in temperate (Khasawneh and Doll, 1978) and tropical regions (Hammond et al., 1986b; Sale and Mokwunye, 1993). Chapter 5 discusses this aspect in more detail.
The conceptual framework in Figure 24 summarizes the factors that determine the feasibility of using PR for direct application (Heng, 2003). Five key factors determine whether the rate of supply of dissolved P from PRs will match the rate of P demand by the crop. These are: the reactivity of the PR, the properties of the soil, the prevailing climate conditions, the crop type, and the management system employed in growing the crop and applying the PR.
Process determining the feasibility of using PR for direct application
Source: Heng, 2001.
A final issue is the integration effect of time. A common feature of PRs is that the initial plant response to PR may be limited. However, their performance relative to water-soluble fertilizer tends to improve with time. This is because of the continuing dissolution of the PR compared with the declining availability of the P from residues and reaction products from water-soluble P sources. Thus, the PR can have a superior residual effect compared with the watersoluble fertilizer. The farmer sees increasing benefits from the PR as time passes.
The framework in Figure 24 highlights the complexity of the whole system. Apart from the factors that affect the agronomic performance of the PR, many other issues will impinge on the use of PR. These include: the size of the potential PR deposit that would supply the PR; the cost of mining, grinding and distribution; the cost-benefit ratio for all participants in the supply chain; the social, economic and environmental impact; and government policy. In the final analysis, these will determine whether a deposit will be developed in order to enable farmers to use the PR as a fertilizer.
Chapter 10 discusses the issue of whether local farmers will adopt PR for use as a P fertilizer. The DSS instruments described in this chapter do not incorporate the question of farmer adoption and the socio-economic and policy issues outlined above. These DSSs focus solely on the question of whether the PR fertilizer will be effective in a specific agricultural system.
Perhaps the most complete approach for predicting how a PR will dissolve is the mechanistic model constructed by Kirk and Nye (1985a, 1985b, 1986). This model starts from the premise that the diffusion of the dissolving ions away from the PR surface is the rate-limiting step. The model is able to provide quite accurate predictions of PR dissolution under controlled conditions (Anderson and Sale, 1993). However, it cannot determine whether a particular crop will respond to a PR application in the field.
The major limitation to the use of complex mechanistic models is the need for a large number of input parameters that are relatively difficult to determine. For example, the Kirk and Nye model requires ten parameters in order to define the soil environment. These include: the concentration and activities of phosphate and calcium (Ca) ions in the soil solution, soil pH, pH buffering capacity, P buffering capacity, coefficients for P adsorption isotherms, soil bulk density, volumetric water content, and percentage clay. This approach is too complex for farm-level use of a PR-DSS.
The New Zealand DSS
Researchers in New Zealand have developed a DSS for using PRs on high-rainfall areas with permanent pastures grazed by sheep and/or cattle. This system is based around five components (Perrott, 2003). The first is an empirical model that predicts the suitability of a particular pastoral site for PR use, given by the mean diffusion (Dm) coefficient of phosphate in soil at a specific site. This is based on direct dissolution measurements of a standard PR (0.075-0.150 mm Sechura PR) at 90 pastoral sites in New Zealand over a two-year period, which provided regression data to predict the Dm variable. Soil type, pH, exchangeable magnesium (Mg), soil drainage and rainfall are used as parameters to predict this variable.
The second component is a laboratory test that determines the reactivity of the PR by measuring the equilibrium P concentration sustained by the PR (CR) in a simulated soil solution maintained at a constant pH (5.5) using an automatic titrator. The CR value, together with the particle density, the P concentration, and the range and amount of particle sizes of the PR, provides a measure of the intrinsic reactivity of the PR (Perrott, 2003).
The third component is a mechanistic model developed from a simplified version of the Kirk and Nye model (Watkinson, 1994a). It calculates the annual dissolution rate of the particular PR at the particular pastoral site, using the predetermined PR and site variables. This annual dissolution rate is then expressed as a rate constant for use in an exponential dissolution model (fourth component) and the annual release of P from the PR is then treated as a plant-available P input into the labile soil P pool, for the OverseerTM, which is an integrated fertilizer DSS (fifth component). OverseerTM is a tool for making site-specific recommendations on the use of P, potassium (K) and sulphur (S) fertilizers on the basis of economic and environmental criteria (Metherell and Perrott, 2003).
The New Zealand DSS is specific for reactive PRs on grazed pasture systems on New Zealand soil types. While the system is limited to this form of land use in this part of the world, the approach has much to recommend it. It attempts to define and quantify the sequential steps in the PR dissolution process. By determining empirical predictors based on simple field experiments, it simplifies complex soil-site-climate interactions that determine the likelihood of PRs dissolving rapidly in a particular pastoral environment. By determining the amount of dissolved PR entering the plant labile pool each year, the modelling process enables the PR to be compared with water-soluble P fertilizers, which also contribute plant-available P to the labile pool in the soil.
The New Zealand DSS is the product of many years of research by chemists, agronomists and computer specialists. The importance of the grazing industries to New Zealand and the need for nutrient inputs to maintain the productivity of these industries have sustained this research effort.
The PR-DSS being developed by the IFDC
The IFDC has also developed a preliminary version of a PR-DSS for estimating the agronomic effectiveness of freshly applied PR with respect to water-soluble phosphate (WSP) fertilizers (Hellums et al., 1992; Chien et al., 1999; Singh et al., 2003). The development is based on IFDC work in West Africa. The model incorporates the effect of PR sources (PR solubility), soil pH, soil texture, organic matter, crop type, and moisture/rainfall regime in predicting the relative agronomic effectiveness (RAE) of the PR with respect to WSP fertilizers. The current version of the model is not able to determine whether P is limiting or what the P fertilizer rate should be. It assumes that other nutrients and pests are non-limiting. It does not consider any socio-economic evaluation.
An example of the use of the current IFDC PR-DSS
A task for the IFDC PR-DSS was to predict how effective Minjingu PR from the United Republic of Tanzania might be for growing a maize crop on a soil at Kabete, Kenya (Singh et al., 2003). The DSS required the following inputs: the neutral ammonium citrate (NAC) solubility of the moderately reactive PR (8.45 percent P2O5 of the PR), the soil pHwater (5.02), the organic carbon percent of the soil (1.44 percent), the clay content (25 percent), the sand content (38 percent), the growing-season rainfall (800 mm), and the crop type (maize). The DSS predicted the RAE for Minjingu PR would be 85 percent. Singh et al. (2003) reported that the observed RAE values at this site ranged from 68 to 100 percent, with the value for the fourth annual application ranging from 80 to 90 percent.
It is also possible to use expert systems to develop a DSS. They do not require the years of dedicated research needed to develop mechanistic models. Instead, they require the input of a skilled knowledge engineer, who spends substantial time working with PR specialists, and other experts who have years of experience in the field. Australian researchers used this approach in order to construct a DSS to advise on PR use on pastures. The task was to determine where PRs might be effective on pastures grazed by sheep and/or cattle in high-rainfall regions in eastern and southern Australia.
An expert system was built on the findings of a large, national project that managed to establish 25 effective, replicated field experiments on representative soils in the target regions. The project generated annual substitution values of triple superphosphate (TSP) for the highly reactive North Carolina PR at each site for a four-year period. The SV50 parameter is the amount of P (kilograms) supplied as TSP that produced 50 percent of the maximum yield response to P at a particular site, divided by the amount of P (kilograms) from North Carolina PR required to produce the same yield.
The project provided a picture of the performance of PRs in a range of pastoral environments that differed in rainfall, soil properties, and pasture types. These involved environments where North Carolina PR was: (i) as effective as TSP in the first year; (ii) almost as effective as TSP over time; and (iii) completely ineffective compared with TSP. The effect of time was allowed for as PR effectiveness was considered after 1 and 4 years of annual PR applications. A knowledge engineer then used this empirical information, together with scientific understanding of the way the four environmental variables determine PR effectiveness, to construct an expert system called RPR Adviser that could advise Australian pastoral farmers on PR use (Gillard et al., 1997).
The approach taken in building the RPR Adviser was to assign a scalar weighting to each climate or soil factor that appeared to affect PR performance. These included: the reactivity of the PR, rainfall, soil pH, soil texture, the likelihood of leaching of P from water soluble fertilizer, P sorption capacity, and pasture composition. Each factor was assigned independently, with the overall prediction being achieved by multiplying all the factor weightings together.
The expert system was able to provide accurate predictions of the SV50 values that occurred (the correlation coefficient between predicted and observed SV50 values was 0.92) (Gillard et al., 1997). It was also possible to validate results from a number of independent field experiments by using pasture yield responses to PR and superphosphate fertilizer. The required input variables for the expert system were simple and did not require specialized laboratory analyses. The site variables were: annual rainfall, the likelihood that soluble P will leach from the rootzone, and pasture composition. The required soil variables were: pH, Colwell P, field texture and soil colour, with colour and texture being surrogates to estimate P sorption capacity for the soil.
The researchers found some useful hints for using the computerized DSS. As there was no commercial intellectual property involved, the RPR Adviser has been made available free-of-charge at: www.latrobe.edu.au/www/rpr/. This has facilitated its widespread use. In addition, by having notes and string messages, it is possible to explain how a decision has been derived. Good help facilities can help the user to use the system with confidence.
Farmers in Australia received a more simplified DSS in the form of a checklist of questions. The questions gave an indication of conditions likely to result in effective PR use on Australian pastures. If all the questions could be answered in the affirmative, then there was a high likelihood that reactive PRs would be useful P fertilizers at the specific pasture site.
An example of the use of the RPR Adviser
A beef producer in southern Gippsland, a high-rainfall district in southeast Australia, wanted to determine whether a reactive PR might be an effective P fertilizer, compared with single superphosphate, for maintaining the productivity of white-clover, perennial ryegrass pastures. The farmer downloaded the RPR Adviser from the Internet and answered the questions asked by the DSS. The answers were that the annual rainfall was 1 000 mm, the PR to be used was highly reactive, the soil was not red or lateritic (indicating that the P sorption capacity was not high), the soil had a deep sandy texture, and the soil pHwater was 5.3. Unsure of what was meant by a question about the extent of leaching in the soil, the farmer checked the help file and received a brief explanation. The farmer then gave the answer that moderate leaching was possible. The DSS responded by indicating that the highly reactive PR would be as effective as superphosphate in the first year if the farmer changed to the RPR. This prediction is consistent with RPR performance in similar environments across southern Australia. The DSS then advised the beef producer to make an economic assessment of the choice based on the cost per kilogram of P applied. Furthermore, the DSS also warned of a possible shortage of S in that pasture soil if straight RPR were used without added S.
Heng (2003) has outlined a conceptual framework (Figure 24) for the further development of a more global PR-DSS. It is possible to prescribe a number of steps and a series of guidelines for the development of this kind of system. Such a system is now required in order to provide advice on PR effectiveness for food and cash crops in the tropics and subtropics, particularly in developing countries.
Phosphate-bearing minerals vary widely in their inherent characteristics. Therefore, the characterization of the agronomic potential of PRs is the first and essential step in evaluating their suitability for direct application and for the development of a comprehensive DSS to evaluate PR use. The specific tasks include: (i) collate data on PR solubility analysed with NAC (second extraction) with standard procedures (specified shaker speed and time); (ii) correlate the solubility data obtained with other available data for reactivity tests of the same PRs; (iii) conduct a regression analysis to establish relationship between the two methods under study; (iv) analyse outliers using standard procedures; and (v) analyse all remaining samples using standard procedures.
The Joint FAO/IAEA Division of Nuclear Techniques in Food and Agriculture and the IFDC will be sharing their data set of PR properties in order to create a global database for a PR-DSS. Over the years, the IFDC has conducted numerous field trials in Africa, Asia and Latin America and it has accumulated a large database. Similarly, through the international PR research network implemented from 1993-99 in both developing and developed countries, the Joint FAO/IAEA Division of Nuclear Techniques in Food and Agriculture has collected valuable data on the agronomic effectiveness of P fertilizers, with a variety of PR sources and numerous field crops in a wide range of environments (Zapata, 2000, 2003; IAEA, 2002).
A standard characterization of soil samples from target regions where PRs might be considered for use is required in order to provide reliable input information for the DSS. This will involve collating measurements for key soil properties that affect the dissolution of PRs. The standard characterization of both PRs and soil properties was undertaken by the Joint FAO/IAEA Division of Nuclear Techniques in Food and Agriculture from the research network project on "The use of nuclear and related techniques for evaluating the agronomic effectiveness of phosphatic fertilizers, in particular rock phosphates" (Truong and Zapata, 2002; Montange and Zapata, 2002).
The PR performance information that is used to construct a DSS needs to cover the full range of PRs, climates, soil types and farming systems that are likely to be encountered in the target region. For example, if a DSS is required for certain regions in Africa, then all of the possible findings on PR performance in those regions need to be incorporated in the DSS. This will include socio-economic factors that are an integral part of the farming system. These may well determine whether a local PR might be used in a specific cropping situation. There is an emerging view that DSS applications need to focus on real problems facing ordinary people, and attempt to develop solutions to them (Matthews et al., 2002). The provision of cost-effective plant nutrient sources to combat nutrient-depleted soils is a complex problem throughout the developing world.
The input information for the DSS needs to be simple and able to be provided by likely users. It is not practical to request complex soil measurements, such as pH buffering capacity, as an input variable for the DSS if the user has no idea of the measurement and it can only be obtained in a research laboratory. Soil colour, texture, pH and rainfall are examples of readily obtainable input variables.
There also needs to be close collaboration between the people building the DSS and the researchers who are evaluating PRs in the field. The former need to receive all possible information about PR performance and the environment in which the PR was used, while the latter need to provide all possible information on PR performance for incorporation in the DSS. Researchers planning a field evaluation of a PR should consult with the DSS developers in order to determine whether there might be an additional treatment to include in the field experiment that could be useful in developing the DSS further.
Both the IFDC and FAO/IAEA have agreed to collaborate in the development of a global PRDSS. Plans for this development are underway in terms of developing a simplified PR database from the large amount of information available. Further development will require the processing of additional data in order to establish functional and statistical relationships and then their inclusion in the PR-DSS.
A network of pilot testing and validation field experiments will also be established in a wide range of agro-ecological zones. These will be located at selected sites in Africa, Asia and Latin America and gather information at the cropping-system level, including agronomic management and socio-economic data. This will enable enhanced prediction of expected results and improved decision-making. The resulting DSS will be further developed as a Web-based system, delivering outputs over the Internet. This will facilitate resource sharing, adherence to corporate standards and use of tools and standards developed by the World Agricultural Information Center.
The availability of a global PR-DSS will be a useful research and extension tool for researchers, extension workers, farmers, planners, and agribusiness dealers. It will assist in promoting the use of PR resources in tropical and subtropical developing countries.