NOTE:

This manuscript is based on a paper presented at the Sustainable Aquaculture Conference held by PACON International at Honolulu, Hawaii from June 11-14, 1995.

Abstract
Introduction
General framework of POND
Functional Modules
Applications of POND
Future Directions
References
Figures

Abstract

Decision-making for pond aquaculture requires knowledge of the effects of management practices on fish performance, soil and water quality. It is also necessary to examine economic and environmental consequences of various practices. A decision support system POND which enables definition of an entire aquaculture facility (in terms of location, ponds, fish populations, and species), and provides analysis capabilities in the form of simulation models and an economics package has been developed. The software requires an IBM-compatible personal computer running the Microsoft Windows operating environment. POND models can be used to examine the implications of management practices such as feeding, fertilization, liming, stocking and water exchange rates on facility-level fish production. Effects of management practices on nutrient levels in the pond environment can also be examined. Fish growth models in POND can be parameterized for single or multiple species. Economic analysis is accomplished by the use of enterprise budgets which account for fixed, depreciable, and variable costs, as well as income based on fish yields predicted by the models. POND provides a useful framework for integrating various components that define a pond aquaculture facility. Applications of decision support systems such as POND for technology transfer, management, planning and research are discussed.

Introduction

Pond aquaculture planners and managers are often confronted with a variety of decisions regarding site locations, target fish species and appropriate practices such as fish feeding, pond fertilization and liming, stocking densities, aeration, and water exchange. These decisions may have considerable effects on both the economics of an aquaculture facility, as well as the environment that surrounds it. The decision-making process typically requires considerable knowledge on the part of the planner or manager. Such knowledge may include an understanding of the principles of pond aquaculture and the implications of various decisions on facility-level economics. Decision-makers usually acquire the required knowledge via formal education or by trial and error. Often, the need for immediate technology may cause decision-makers to apply or recommend pond management practices developed and tested at one location to a new site.

The use of technology that has been found to be suitable for one location may very well be inadequate when applied elsewhere. This may be due to differences in fish production potential caused by the variability in climate, water and soil characteristics among sites, as well as availability and cost of resources used in pond production. For example, a decision as specific as the calculation of feed requirements for a pond requires consideration of fish biomass, natural food availability, and water temperature among other factors. Similarly, calculation of fertilizer application rates requires a basic understanding of soil and water chemistry. In both cases, availability and cost of appropriate inputs should be factored into the decision-making process.

The complexity of decision-making for an aquaculture facility suggests the need for analytical tools that can integrate various components of the knowledge base required to arrive at a decision. Such tools, termed decision support systems, integrate knowledge in the form of mathematical models, rule-based (expert) systems, and/or databases into user-friendly software systems focused on developing, analyzing and optimizing management strategies. These tools have emerged as powerful tools for capturing expert knowledge about particular domains and providing that knowledge in a friendly, easy to use manner to end users. A key component of any decision support system is the knowledge base(s) upon which decisions are made. Expertise exists in many forms, ranging from highly qualitative 'rule of thumb' approaches useful for capturing subjective, historical experience, to databases containing historical information available for 'data mining', to more rigorous and quantitative mathematical algorithms that describe explicit relationships between components of the domain in question.

In agriculture, decision support systems have been developed for the diagnosis of plant diseases (Michalski et al., 1982), crop production (Smith et al., 1985), analyzing marketing alternatives (Uhrig et al., 1986), selection of appropriate crop cultivars (Bolte et al., 1990), and many others. A simple decision support system PONDCLASS that focuses on fertilization and liming recommendations for pond aquaculture has been developed (Lannan, 1993). A more comprehensive software POND that integrates the functionality of PONDCLASS, and provides additional analysis capabilities in the form of simulation models and enterprise budgeting has since been developed (Bolte et al., 1995). This paper presents an overview of POND, describes its functional modules, and highlights applications in pond aquaculture. The development of POND is supported by the Pond Dynamics/Aquaculture Collaborative Research Support Program (PD/A CRSP) funded in part by the U.S. Agency for International Development.

General Framework of POND

The general framework of POND, functionality and application areas of the software are indicated in Figure 1. The software requires an IBM-compatible personal computer running the Microsoft Windows (version 3.1 or higher) operating environment. It requires approximately 1.5MB of available hard disk space and a minimum of 4MB RAM. An 80386 CPU is required, and an 80486 or greater CPU is recommended.

The main focus of the POND software is to provide a view of pond dynamics at both the individual pond as well as at the facility level. This involves the ability to define a pond facility, providing capabilities for simulating processes within a pond, and enabling users to impose certain management or planning decisions for a given facility. The term 'facility' is used in POND to describe a physical aquaculture system consisting of a specific geographical location, source water quality, pond(s) associated with the site, fish lot(s) or populations (comprising one or more species) associated with each pond, and a soil type for each pond. Mini-databases are maintained to record user-specified information for each of the above entities. Databases are also available for other functional components of the software such as simulation scenarios, economic information, fertilizer, lime and feed materials, and weather characteristics.

Functional Modules

The functionality of POND encompasses four general areas (Fig. 1): (i) PONDCLASS functionality (i.e., estimation of fertilizer and lime requirements for individual ponds without the use of simulation models), (ii) pond simulation capabilities, (iii) economic analysis, and (iv) parameter estimation.

PONDCLASS functionality

The rationale for fertilizer recommendations proposed by Lannan (1993) is based on observations that fish production in fertilized ponds can be enhanced by appropriate management of net primary productivity. Standard limnological equations are used to assess nutrient requirements for a pond (see Lannan 1993 for details), and fertilizer amendments required to satisfy these nutrient requirements of a pond are calculated based on the availability of nutrients from different fertilizers, after taking into account ambient nitrogen and phosphorus concentrations in the pond water. In pond aquaculture (as in agriculture), it is often more economically efficient to use a mix of fertilizers rather than one fertilizer to satisfy nutrient requirements; this is achieved by the use of a linear programming algorithm (Nelder and Mead 1965) which estimates the least cost combination of fertilizers that will meet nutrient requirements of a pond.

Addition of lime to ponds with acid muds or water of low alkalinity is a widely accepted aquacultural practice. The amount of calcium carbonate required to neutralize the exchange acidity of a pond soil is called the lime requirement, which depends on ion exchange processes that occur on the surface of soil particles. A detailed description of such processes as it applies to pond soils is available in Boyd (1990). Bowman (1992) developed a simple technique based on soil pH to estimate lime requirements. This approach is used to estimate the amount of lime to be added to a pond of a given soil type in the POND program.

Fertilizer recommendations generated by the guidelines of Lannan (1993) have been tested by PD/A CRSP researchers in the Philippines, Thailand and Honduras. Preliminary analysis of the results suggest that fertilizer application rates for the Thailand site are lower than those traditionally used there, because the software takes into account nutrients remaining in ponds from previous inputs (Szyper and Hopkins, 1995). This translates into lower fertilization (and therefore fish production) costs, and is consistent with earlier results from an experiment conducted in the Philippines (Hopkins et al. 1994). In both experiments, fish yields were within the range typically obtained from fertilized ponds. Fish yields from the experiment in Honduras were comparable to control treatments, but fish growth in the PONDCLASS treatment appeared to be limited by elevated ammonia nitrogen concentrations (Teichert-Coddington and Ramos, 1995). Although the liming recommendations proposed by Bowman (1992) have not been tested in the field, recent laboratory experiments with different soils suggest that the lime requirements estimated using his procedure are comparable to those generated by the use of more sophisticated methods that require detailed soil analyses (J. R. Bowman, Oregon State University, personal commn.).

Pond simulation capabilities

Simulation models are the primary analysis tool in POND. These models are essentially collections of mathematical equations that describe fundamental relationships between components within the system of interest. They can be used to predict the response of the system to a particular set of operation conditions or environmental constraints. POND simulation models are deterministic in nature, and are formulated as a set of ordinary differential equations which are solved numerically over time, and provide opportunity to address fish growth, water/sediment quality dynamics, and primary and secondary productivity under different pond conditions. POND supports both daily and diurnal simulations of pond facilities. Details of the simulation models are beyond the scope of the current discussion and are described elsewhere (Bolte et al., 1995; Nath et al., in preparation).

Simulation models were chosen as the main tool for analyzing pond ecosystems for a variety of reasons. They provide an opportunity for knowledge synthesis, whereby a large body of knowledge about ponds can be represented as a single entity, which can then be used to explore 'what-if' scenarios in the pond ecosystem. Simulation model development also imposes a rigorous framework on the model builder. This forces the model builder to clearly articulate knowledge of the fundamental relationships that govern a pond's response to external stimuli, simultaneously exposing gaps in the knowledge base. The rigorous nature of model specification results in a testable hypothesis about the system, i.e., the model can be run for a known set of conditions to determine whether its output is adequate to represent our knowledge of systems processes.

Simulation models are also useful tools for predicting system response to conditions that are either too complex or expensive to explore experimentally. Because model-based experiments can be completed in seconds on a computer, rather than months or years in the field, models provide opportunity for exploring a much larger set of operating conditions, environments, management strategies, and constraints than traditional experiments allow. Further, results of such numerical experiments are useful in evaluating model assumptions, and refining the models.

It is important to note that ponds are complex, rather unpredictable ecosystems. Therefore, ecosystem-level validation of the models in POND is a difficult task. Further, data are often not available to validate certain model components. However, some POND models (e.g., fish growth) have been validated with favorable results (Bolte et al., 1995). Validation of other components is an on-going activity.

POND simulations are dynamic, providing time series analysis for a range of variables that are dependent on the level of resolution requested prior to simulation. During the simulation, POND accumulates time series data for each variable; these data may be viewed in plots or tables at the end of the simulation run. Simulation models in pond are organized hierarchically into three levels. This scheme allows users to perform different kinds of analyses based on data availability and output resolution requirements. Each of these levels are described below.

Level 1: Level 1 models are fairly simple and require minimal data inputs. State variables simulated are fish growth (based on a bioenergetics model), water temperature and pond volume (Fig. 2). Weather data that are required for the latter two variables may either be generated by a simple weather generator embedded in POND or read from ASCII files provided by the user. Consumption of natural food by fish is assumed to be a function of fish biomass and appetite. Fertilizer application rates are user-specified, but POND allows supplementary feeding schedules to be either optionally generated or specified by the user.

Level 2: Level 2 models provide a substantially more sophisticated view of pond dynamics, allowing prediction of phytoplankton, zooplankton and nutrient dynamics (total carbon, nitrogen and phosphorus) in the pond water, in addition to all the functionality of Level 1 (Fig. 3). At Level 2, steady state bacterial concentrations are maintained, and nutrient exchange between pond water and sediments is assumed to depend on the concentration gradient between these two components. Fish can feed from a pool of natural food resources and/or artificial feed. Consumption of natural food (phytoplankton, zooplankton and bacterial pools) by fish is predicted on the basis of a resource competition model (Tilman, 1982), and is dependent on the concentrations of the natural food pools and artificial feed (if specified), food preference of fish species, and fish appetite. At this level, both fertilizer and feed requirements may be user-specified or optionally generated by the model.

Level 3: Level 3 models inherit all the functionality of Level 2, and provide additional capabilities for simulating bacterial kinetics, and detailed pond water/sediment quality dynamics (Fig. 4). Additional state variables for pond water include dissolved oxygen and alkalinity. Further, state variables are also maintained for organic, ammonia and nitrate nitrogen, as well as organic and inorganic phosphorus and carbon in both pond water and sediments. User-specified fertilization and feeding regimes, coupled to pond process-based nutrient mass balances, are used to estimate nutrient consumption and production rates.

Economic analysis

POND enables economic analyses of facilities to be accomplished by the use of enterprise budgets. These budgets allow definition of various types of cost and income items, and associated interest and depreciation expressions, which can be used to assess the overall economic viability of a particular production enterprise. Three cost categories are supported: (i) fixed, (ii) depreciable and (iii) variable costs. Fixed costs do not change over the course of facility operation (e.g., construction cost for a pond). Related to fixed costs are depreciable costs; this category applies to items such as equipment, which may have a redeemable value after some period of time. POND incorporates depreciation schedules that describe the loss of value of the depreciable asset over time. Variable costs are neither fixed or depreciable, and typically vary according to the scale of production (e.g., labor costs, fertilizer and feed costs).

To generate an enterprise budget, income sources are also required. POND allows the specification of any number of income sources, based on either a per unit area, per unit of production, or per facility basis. Income sources relating to fish production are automatically fed to the economics package at the end of a simulation run. Finally, interest rates used for calculating fixed and variable investment costs are to be provided by the user. Once all the items to be included in the enterprise budget are specified and a simulation completed, the economics package in POND summarizes costs on an areal (e.g., per ha), per unit of production (e.g., per kg of fish produced) or overall facility basis, balances these costs against income, and reports the results in a tabular form. By including and/or excluding particular costs/incomes, or adjusting cost/income details, users can quickly 'experiment' with various facility configurations and/or management strategies to examine their effects on the economic performance of the facility.

Parameter estimation

Users of the POND program may often be interested in tailoring the models to one or more fish species at a given location. This may be accomplished by calibrating (adjusting) model parameters such that the simulations result in fish growth profiles that are consistent with the user's experience or match their growth data adequately. Although the task of calibration can be accomplished manually, it is tedious because of the large number of parameters in the models. Therefore, a parameter estimation routine is embedded in the POND program; this routine generates best-fit model parameters by comparing the results of multiple runs of the models with user-provided fish growth data. The parameter estimation routine allows users to adapt the software to different species, locations and/or management practices.

Applications of POND

Decision support systems such as POND can provide valuable information in the context of pond management, planning, extension (including technology transfer) and research. Specific applications where POND is likely to be useful within each of these broad focus areas are discussed in greater detail below. In general, Level 1 models in POND are intended for applied management and rapid analysis of pond facilities. Level 2 models allow for more detailed pond analysis, management optimization and numerical experimentation, and Level 3 models are useful for exploring fundamental aspects of pond dynamics (e.g., detailed nutrient transformations in pond water and sediments, atmospheric diffusion, etc) in addition to Level 2 functionality.

Species/Facility customization

The POND framework is generic in that it can be adapted for different species and culture conditions. This feature may be useful for pond managers who wish to explore the use of alternate species or want to compare model output and recommendations (e.g., feeding or fertilization rates) to their current practices. Such analyses may also be important for planning, research and extension activities (e.g., feasibility studies for different species and/or locations).

Economic optimization

The simulation and economic analysis capabilities of POND can be useful for economic optimization. From a management perspective, such analyses may focus on identifying suitable practices (e.g., levels of fertilization and feeding, water exchange) for an already exisiting facility. From a planning perspective, optimization may provide useful information for feasibility studies that focus both on facility configurations (e.g., combinations of ponds, lots and species) and management strategies. Optimization may also be of interest to researchers involved in the comparison of economic benefits from different pond aquaculture systems. Currently, automatic facility-level economic optimization procedures are not fully supported in POND.

Lot management

POND managers may be interested in exploring the effects of stocking single or multiple fish species at different densities in ponds. It may also be of interest to examine the effects of stocking and harvesting lots at different dates on facility-level economics. Such analyses may be useful in the context of scheduling pond operations and assessing resource needs (e.g., fingerling requirements). Extension agents may also find it beneficial to use POND as a tool in recommending appropriate stocking densities to farmers based on local conditions.

Estimation of feed requirements

Artificial feeds often represent the single most important component of variable costs in an aquaculture facility. This is especially true of large-scale commercial operations. Therefore, assessment of feed requirements (in terms of both quantity and quality) and subsequent effects on facility-level economics (and water/sediment quality) will likely be useful for a wide variety of personnel involved in pond aquaculture.

Estimation of fertilizer and lime requirements

As indicated earlier, fertilizer requirements for ponds can either be generated by the use of guidelines proposed by Lannan (1993), specified by the user or generated by the models in POND. Each of these options serve different needs. Thus, the fertilization guidelines of Lannan (1993) allow pond managers to adjust fertilization rates on a routine basis. This is beneficial from the viewpoint of reducing fertilizer waste (if any) and provides for interactive pond management. The use of fertilizer application rates defined by pond managers as input data to the POND models should provide opportunity to compare model outputs (e.g., fish growth, plankton growth, and/or nutrient concentrations) with actual pond information. Such analyses will likely provide opportunity for future model refinement. On the other hand, generation of fertilization schedules as one of the model outputs is useful from the viewpoint of assessing fertilizer requirements for an entire facility and gauging the viability of certain pond aquaculture systems (e.g., subsistence farming) from a planning perspective.

Lime requirements recommended by POND are likely to be more applicable to newer ponds without extensive organic matter deposits (J.R. Bowman, Oregon State University, personal commn.). Older ponds may have different lime requirements because the original nature of the soil is modified as its organic matter content accumulates.

Water and sediment quality management

Analyses of the effects of management practices on pond water/sediment quality are important from planning, extension and management perspectives in terms of resource flows, facility-level economics and verifying whether effluent standards (e.g., nitrogen, phosphorus and organic carbon levels) are met. It may also be possible to assess short-term aeration requirements for ponds by the use of Level 3 models. Because many pond processes are not fully understood, the POND models may also be used to guide experimental work that specifically focuses on these processes.

Future Directions

As indicated earlier, current research on POND focuses on validation of different models in the software. It is also anticipated that techniques will be developed to enable more comprehensive economic optimization to be undertaken. Further, an effort is underway to combine facility-level pond and crop simulations for a thorough analysis of integrated farming systems. Our experience with the software suggests that it is likely to be a useful tool for facility-level analysis. Applications in regional-scale pond aquaculture development, as well as in socioeconomic and environmental impact studies of such development may be possible by linking the simple models in POND to regional-scale analysis tools such as geographical information systems (GIS).

References

Bolte, J.P., Hannaway, D.B., Shuler, P.E. and Ballerstedt, P.J. 1990. An intelligent frame system for cultivar selection. A.I. Applications in Natural Resource Management, 5(3):21-31.

Bolte, J.P., Nath, S.S., and Ernst, D.H. 1995. POND: A decision support system for pond aquaculture. Twelfth Annual Administrative Report, PD/A CRSP, Corvallis, OR. pp. 48-67.

Bowman, J.R., 1992. Classification and management of earthen aquaculture ponds with emphasis on the role of the soil. PhD Dissertation, Oregon State University, Corvallis, OR. 209 pp.

Boyd, C. E. 1990. Water quality in ponds for aquaculture. Alabama Agricultural Experimental Station, Auburn University, AL. 482 pp.

Hopkins, K.D., Lopez, E., and Szyper, J.P. 1994. Intensive fertilization of tilapia ponds in the Philippines. Eleventh Annual Administrative Report, PD/A CRSP, Corvallis, OR. pp. 16-20.

Lannan, J. E. 1993. Users guide to PONDCLASS: Guidelines for fertilizing aquaculture ponds. PD/A CRSP, Oregon State University, Corvallis, OR. 60 pp.

Michalski, R.S., Davis, J.H., Bisht, V.S. and Sinclair, J.B. 1982. PLANT/DS: an expert consulting system for the diagnosis of soybean diseases. Proceedings of the 1982 European Conference on Artificial Intelligence, Orsay, France, July 12-14, 1982. pp. 133-138.

Nath, S.S., Bolte, J.P., and Ernst, D.E. In preparation. Systems modelling for pond aquaculture.

Nelder, J.A. and Mead, R. 1965. A simplex method for function minimization. Comput. J., 7: 308-313.

Smith, R.D., Barrett, J.R. and Peart, R.M. 1985. Crop production management with expert systems. ASAE Paper No. 85-5521. ASAE, St. Joseph, MI.

Szyper, J.P. and Hopkins, K.D. 1995. Management of carbon dioxide balance for stability of total alkalinity and phytoplankton stocks in fertilized fish ponds. Twelfth Annual Administrative Report, PD/A CRSP, Corvallis, OR. pp. 28-32.

Teichert-Coddington, D.R., and Ramos, H. 1995. Nutrient input management by the computer program, PONDCLASS, and by concentration of a key nutrient. Twelfth Annual Administrative Report, PD/A CRSP, Corvallis, OR. pp. 20-27.

Tilman, D. 1982. Resource competition and community structure. Princeton University Press, Princeton, NJ. 296 pp.

Uhrig, J.W., Thieme, R.H. and Peart, R.M. 1986. Grain marketing alternative selection: an expert system approach. Paper presented at the AAEA Symposium: Innovative Extension Delivery Systems: Satellites, Expert Systems and Artificial Intelligence. Reno, Nevada, July 29, 1986.

Figure 1. Architecture of POND indicating databases, functional modules and applications for decision support.

Figure 2. State variables and system inputs/outputs for Level 1 modelling in POND.

Figure 3. State variables and system inputs/outputs for Level 2 modelling in POND.

Figure 4. State variables and system inputs/outputs for Level 3 modelling in POND.

Return to the POND Publications page

Biosystems Analysis Group page

Dr. John Bolte or Shree Nath
Biosystems Analysis Group
Department of Bioresource Engineering
Gilmore Hall
Oregon State University
Corvallis, OR 97331, U.S.A
Phone: (503) 737-6303
Fax: (503) 737-2082
e-mail: boltej@ccmail.orst.edu OR naths@ccmail.orst.edu

Copyright © 1996 Shree Nath, Biosystems Analysis Group, Oregon State University