Optimal management of agricultural systems

To remain competitive, many agricultural systems are now being run along business lines. Systems methodologies are being incorporated, and here evolutionary computation is a valuable tool for identifying more profitable or sustainable solutions. However, agricultural models typically pose some of the more challenging problems for optimisation. This chapter outlines these problems, and then presents a series of three case studies demonstrating how they can be overcome in practice. Firstly, increasingly complex models of Australian livestock enterprises show that evolutionary computation is the only viable optimisation method for these large and difficult problems. On-going research is taking a notably efficient and robust variant, differential evolution, out into real-world systems. Next, models of cropping systems in Australia demonstrate the challenge of dealing with competing objectives, namely maximising farm profit whilst minimising resource degradation. Pareto methods are used to illustrate this trade-off, and these results have proved to be most useful for farm managers in this industry. Finally, land-use planning in the Netherlands demonstrates the size and spatial complexity of real-world problems. Here, GIS-based optimisation techniques are integrated with Pareto methods, producing better solutions which were acceptable to the competing organizations. These three studies all show that evolutionary computation remains the only feasible method for the optimisation of large, complex agricultural problems. An extra benefit is that the resultant population of candidate solutions illustrates trade-offs, and this leads to more informed discussions and better education of the industry decision-makers.

Comparison of odour emission rates measured from various sources using two sampling devices

Two commonly used sampling devices (a wind tunnel and the US EPA dynamic emission chamber), were used to collect paired samples of odorous air from a number of agricultural odour sources. The odour samples were assessed using triangular, forced-choice dynamic olfactometry. The odour concentration data was combined with the flushing rate data to calculate odour emission rates for both devices on all sources. Odour concentrations were consistently higher in samples collected with a flux chamber (ratio ranging from 10:7 to 5:1, relative to wind tunnel samples), whereas odour emission rates were consistently larger when derived from wind tunnels (ratio ranging from 60:1 to 240:1, relative to flux chamber values). A complex relationship existed between emission rate estimates derived from each device, apparently influenced by the nature of the emitting surface. These results have great significance for users of odour dispersion models, for which an odour emission rate is a key input parameter.

Odour emissions from anaerobic piggery ponds. 2: Improving estimates of emission rate through recognition of spatial variability

Odour emission rates were measured for seven different anaerobic ponds treating piggery wastes at six to nine discrete locations across the surface of each pond on each sampling occasion over a thirteen month period. Significant variability in emission rates were observed for each pond. Measurement of a number of water quality variables in pond liquor samples collected at the same time and from the same locations as the odour samples indicated that the composition of the pond liquor was also variable. The results indicated that spatial variability was a real phenomenon and could have a significant impact on odour assessment practices. Considerably more odour samples would be required to characterise pond emissions than currently recommended by most practitioners, or regulatory agencies.

Optimal eradication: when to stop looking for an invasive plant

The notion of being sure that you have completely eradicated an invasive species is fanciful because of imperfect detection and persistent seed banks. Eradication is commonly declared either on an ad hoc basis, on notions of seed bank longevity, or on setting arbitrary thresholds of 1% or 5% confidence that the species is not present. Rather than declaring eradication at some arbitrary level of confidence, we take an economic approach in which we stop looking when the expected costs outweigh the expected benefits. We develop theory that determines the number of years of absent surveys required to minimize the net expected cost. Given detection of a species is imperfect, the optimal stopping time is a trade-off between the cost of continued surveying and the cost of escape and damage if eradication is declared too soon. A simple rule of thumb compares well to the exact optimal solution using stochastic dynamic programming. Application of the approach to the eradication programme of Helenium amarum reveals that the actual stopping time was a precautionary one given the ranges for each parameter.

Deriving optimal fishing effort for managing Australia's Moreton Bay multispecies trawl fishery with aggregated effort data

Deriving an estimate of optimal fishing effort or even an approximate estimate is very valuable for managing fisheries with multiple target species. The most challenging task associated with this is allocating effort to individual species when only the total effort is recorded. Spatial information on the distribution of each species within a fishery can be used to justify the allocations, but often such information is not available. To determine the long-term overall effort required to achieve maximum sustainable yield (MSY) and maximum economic yield (MEY), we consider three methods for allocating effort: (i) optimal allocation, which optimally allocates effort among target species; (ii) fixed proportions, which chooses proportions based on past catch data; and (iii) economic allocation, which splits effort based on the expected catch value of each species. Determining the overall fishing effort required to achieve these management objectives is a maximizing problem subject to constraints due to economic and social considerations. We illustrated the approaches using a case study of the Moreton Bay Prawn Trawl Fishery in Queensland (Australia). The results were consistent across the three methods. Importantly, our analysis demonstrated the optimal total effort was very sensitive to daily fishing costs—the effort ranged from 9500–11 500 to 6000–7000, 4000 and 2500 boat-days, using daily cost estimates of $0, $500, $750, and $950, respectively. The zero daily cost corresponds to the MSY, while a daily cost of $750 most closely represents the actual present fishing cost. Given the recent debate on which costs should be factored into the analyses for deriving MEY, our findings highlight the importance of including an appropriate cost function for practical management advice. The approaches developed here could be applied to other multispecies fisheries where only aggregated fishing effort data are recorded, as the literature on this type of modelling is sparse.

Estimating daily methane production in individual cattle with irregular feed intake patterns from short-term methane emission measurements

Spot measurements of methane emission rate (n = 18 700) by 24 Angus steers fed mixed rations from GrowSafe feeders were made over 3- to 6-min periods by a GreenFeed emission monitoring (GEM) unit. The data were analysed to estimate daily methane production (DMP; g/day) and derived methane yield (MY; g/kg dry matter intake (DMI)). A one-compartment dose model of spot emission rate v. time since the preceding meal was compared with the models of Wood (1967) and Dijkstra et al. (1997) and the average of spot measures. Fitted values for DMP were calculated from the area under the curves. Two methods of relating methane and feed intakes were then studied: the classical calculation of MY as DMP/DMI (kg/day); and a novel method of estimating DMP from time and size of preceding meals using either the data for only the two meals preceding a spot measurement, or all meals for 3 days prior. Two approaches were also used to estimate DMP from spot measurements: fitting of splines on a 'per-animal per-day' basis and an alternate approach of modelling DMP after each feed event by least squares (using Solver), summing (for each animal) the contributions from each feed event by best-fitting a one-compartment model. Time since the preceding meal was of limited value in estimating DMP. Even when the meal sizes and time intervals between a spot measurement and all feeding events in the previous 72 h were assessed, only 16.9% of the variance in spot emission rate measured by GEM was explained by this feeding information. While using the preceding meal alone gave a biased (underestimate) of DMP, allowing for a longer feed history removed this bias. A power analysis taking into account the sources of variation in DMP indicated that to obtain an estimate of DMP with a 95% confidence interval within 5% of the observed 64 days mean of spot measures would require 40 animals measured over 45 days (two spot measurements per day) or 30 animals measured over 55 days. These numbers suggest that spot measurements could be made in association with feed efficiency tests made over 70 days. Spot measurements of enteric emissions can be used to define DMP but the number of animals and samples are larger than are needed when day-long measures are made.