- Email: [email protected]
Forecasting blowfly strike in Queensland sheep flocks
Veterinary Parasitology 92 (2000) 309–317
Forecasting blowfly strike in Queensland sheep flocks Michael P. Ward∗ Queensland Department of Primary Industries, Animal Research Institute, Locked Mail Bag 4, Moorooka QLD 4105, Australia Received 18 February 2000; received in revised form 12 May 2000; accepted 23 May 2000
Abstract The occurrence of blowfly strike between 1993 and 1999, derived from the reported use of pesticides for flystrike control, was investigated in 247 sheep flocks in Queensland, Australia using autoregressive techniques. Although there was a small increase (0.0016 per year) in flystrike incidence during the study period, this long-term linear trend was not significant (p=0.53). The occurrence of flystrike was best described by an autoregressive model that included flystrike in the previous 2 months: flystriket =0.0170+0.0392 flystriket−1 +0.3589 flystriket−2 . Flystrike was associated with the southern oscillation index (SOI). The SOI is based on barometric pressure readings and is associated with periods of below- (negative SOI) and above-average (positive SOI) rainfall in northern Australia. Flystrike incidence was significantly (p=0.03) greater in months in which the SOI was positive. The strongest correlation (r=0.33) was found between flystrike incidence and the SOI 2 months previously. Using the SOI, the best-fitting autoregressive model describing flystrike was flystriket =0.0238+0.3033 flystriket−2 +0.0009 SOIt−2 . The incidence of flystrike was significantly (p<0.05) correlated with average monthly radiation (r=0.26), but not with average monthly maximum and minimum temperature, total rainfall, evaporation and vapour pressure. The best-fitting autoregressive model describing flystrike occurrence based on these variables was flystriket =−0.0259+0.3610 flystriket−2 +0.0022 radiationt . Results suggest that a useful early-warning system could be developed based on the correlation between flystrike incidence and the SOI up to 2 months previously. Such attempts to forecast flystrike may assist decision-making by wool producers with respect to flystrike control options, leading to more efficient control of blowfly strike in their industry. © 2000 Elsevier Science B.V. All rights reserved. Keywords: Lucilia cuprina; Blowfly strike; Sheep-arthropoda; Climate; Time-series analysis; Southern oscillation index; Australia
∗ Present address: Department of Veterinary Pathobiology, School of Veterinary Medicine, Purdue University, West Lafayette IN 47907, USA. Tel.: +61-765-494-5796; fax: +61-765-494-9830. E-mail address: [email protected] (M.P. Ward).
0304-4017/00/$ – see front matter © 2000 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 4 0 1 7 ( 0 0 ) 0 0 2 8 3 - 1
310
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
1. Introduction The occurrence of blowfly strike in the Australian wool industry has major animal-welfare implications, and causes considerable economic loss (McLeod, 1995). Numerous strategies have been devised by woolgrowers to control flystrike, including husbandry procedures (e.g. crutching and the mules operation), management procedures (e.g. strategic shearing and breeding sheep for resistance to flystrike), the use of pesticides and blowfly baits and traps (Tellam and Bowles, 1997). Pesticides are a popular control method. In recent Queensland studies >80% of woolgrowers surveyed used pesticides to control flystrike in their flocks (Ward and Armstrong, 1998a), and the use of newer insect growth regulator pesticides appears to be replacing organophosphorous compounds for flystrike control (Ward and Armstrong, 1998b). However, such pesticide use has a number of disadvantages, including potential environmental damage caused by scouring wool that contains pesticide residues and occupational-health concerns about handling pesticide-treated sheep. Strategies aimed at reducing the use of pesticides are considered a priority in the Australian wool industry (Williams and Brightling, 1999). Most (>80 to 90%) strikes in Australian sheep flocks are caused by Lucilia cuprina. Its lifecycle is dependent on climatic factors: adult flies are active in the temperature range of 17–35◦ C (Norris, 1966), and moisture increases the survival of larvae and pupae in the soil, promotes plant growth (providing a food source for newly emerged adult flies) and predisposes sheep to fleece rot (Watts et al., 1981). This dependency can potentially be used to develop models to forecast the occurrence of flystrike, thus assisting woolgrowers to control flystrike more effectively using an integrated parasite management program with less reliance on pesticides (Karlsson, 1999). The aim of this study was to investigate the long-term associations between blowfly strike and the southern oscillation index (SOI) and a range of climatic parameters, and to identify associations that may prove useful in developing a forecasting system of flystrike in Queensland sheep flocks. 2. Materials and methods 2.1. Data source The occurrence of blowfly strike in Queensland sheep flocks was determined from information provided by woolgrowers in three postal surveys, conducted between 1995 and 1999, to assess the use of pesticides in the Queensland wool industry (Ward and Armstrong, 1998a; Ward and Armstrong, unpublished data). The sampling frame was all Queensland sheep flocks from which wool had been sold at auction during this time period. Flocks were randomly sampled from the sampling frame through a process of selection of wool lot numbers. Flock managers were sent a questionnaire seeking information on whether or not pesticides had been applied for the control of blowfly strike, when (months) application had occurred after shearing, and the reason for pesticide application (preventive or in response to the occurrence of flystrike). As part of the questionnaire surveys, information was also sought on when (month and year) shearing had most recently taken place and the months of wool growth at this shearing (thus enabling the month and year of the previous shearing to be calculated).
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
311
2.2. Data analysis The reported application of pesticides by flock managers in response to blowfly strike (in contrast to the preventive use of pesticides for flystrike) was used as an indicator variable of blowfly strike. In flocks in which managers indicated that pesticides were not applied for flystrike control it was assumed that flystrike did not occur (since the occurrence of flystrike generally demands use of pesticides for both animal welfare and economic reasons). For all flocks in which flystrike did or did not occur, the period of risk was defined as the period extending from the time of shearing to either when flystrike first occurred (pesticides reported to have been applied in response to flystrike) or the subsequent shearing, respectively. Flystrike incidence was estimated using the number of flocks in which flystrike occurred (indicated by the use of pesticides in response to flystrike) in each month as the numerator and the total number of flocks at-risk of flystrike in each month as the denominator. Flystrike was assumed to occur within the month pesticide treatment was applied, since most strikes are initiated in the week preceding their report by woolgrowers (Wardhaugh and Morton, 1990). Thus, a time-series of flystrike incidence (risk) rate was created. Data describing climatic conditions within Queensland sheep-raising districts during the study period were obtained by interrogation of interpolated climate surfaces (Queensland Centre for Climate Applications, Department of Natural Resources, Brisbane; http://www. dnr.qld.gov.au/silo/datadril.html). Interpolated data were derived from historically recorded point data (Bureau of Meteorology recording stations), interpolated to 0.05◦ (≈5 km), using spline and krigging methods. To represent Queensland sheep-raising districts, the approximate centre (latitude, longitude) of 11 wool districts (Q6, 12, 14, 15, 18, 20, 21, 22, 23, 27 and 28) from which >90% of Queensland wool was produced in 1998/1999 (Australian Wool Exchange, 1999) were selected. At these sites, climatic data were interpolated for daily maximum temperature (◦ C), minimum temperature, rainfall (mm), evaporation (mm), radiation (MJ/m2 ) and vapour pressure (hPa) for each month of the flystrike time-series. For temperature, evaporation, radiation and vapour pressure variables, mean values were calculated from daily values for each month. For rainfall, a total value was calculated from daily values for each month. In addition, the SOI for all months of the time-series was identified. The SOI is based on the difference in barometric pressure readings measured at sites near Darwin (northern Australia) and Tahiti. It is associated with periods of belowand above-average rainfall in northern Australia. For example, when the SOI is strongly negative, tradewinds are weak and rainfall in northern Australia tends to be below-average with drought conditions often developing. This situation is popularly called an El Niño event (Voice, 1991). Troup’s SOI was used, (PA(Tahiti)−PA(Darwin))/s.d., where PA is the pressure anomaly (monthly mean−long-term mean) and s.d. is the standard deviation of the difference (Troup, 1965; http://www.dnr.qld.gov.au/longpdk/lpsoidat.htm#troup). The best autoregressive model (t, t−1, . . . , t−12) describing flystrike incidence was found based on minimum variance and number of model parameters, as determined by the goodness-of-fit statistic Akaike’s corrected information criterion, AICc (ASTSA for Windows, version 1, Division of Statistics, University of California, Davis). This autoregressive model identified flystrike in which previous months best described flystrike in the current month, t. The significance of model coefficients was tested using a t-statistic. Plots of the autocovariance function (ACF) and the partial autocovariance function (PACF) of model
312
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
residuals were examined to ensure stationarity and that the selected model adequately described the time-series. Statistical significance of the ACF and PACF was assessed using 95% confidence intervals. The fluctuation (trend) test was used to test for randomness of model residuals (ASTSA for Windows, version 1, Division of Statistics, University of California, Davis). The cross-correlation between flystrike incidence and the set of climatic parameters was estimated using lags (t) of up to 12 months (t, t−1, . . . , t−12). In addition, the association between flystrike incidence and positive or negative SOI was tested using Kruskal–Wallis nonparametric analysis-of-variance. Correlations between temperature, rainfall, evaporation, radiation and vapour pressure parameters were calculated. Evidence of highly correlated parameters (r>0.95) was used to delete redundant parameters from model building. Vector autoregression was used to identify the most appropriate model describing the association between flystrike incidence and the SOI, and between flystrike incidence and the other selected climatic parameters. Flystrike incidence, SOI and climatic parameter lags of up to t=12 were investigated. The significance of model coefficients was tested using a t-statistic. AICc was used as a goodness-of-fit statistic, and the ACF and PACF of residuals of selected models were used to assess stationarity. The fluctuation (trend) test was used assess series randomness (ASTSA for Windows, version 1, Division of Statistics, University of California, Davis).
3. Results Questionnaires were sent to 833 Queensland woolgrowers, representing 32% of Queensland sheep flocks. A total of 576 (69%) questionnaires were returned. Eighty-six woolgrowers reported applying pesticides to flocks when blowfly strike occurred (rather than applying pesticides preventively for flystrike) and 207 reported no application of pesticides (neither preventive nor control) for flystrike. This subset of 293 flocks was used to construct a time-series of flystrike incidence in Queensland sheep flocks. Month of pesticide application, time of shearing and months of wool growth was available for 73 (85%) of the flocks in which pesticides were used to control flystrike. Time of shearing and months of wool growth were available for 174 (84%) of the flocks in which pesticides were not used for flystrike. Therefore, the flystrike incidence time-series was constructed using data from 247 flocks. The period during which one or more flocks were at-risk of flystrike was June 1993 to April 1999, or 71 months. The mean number of flocks at-risk of flystrike in each study month was 38. The time-series of estimated flystrike incidence in Queensland sheep flocks is shown in Fig. 1. Estimated annual incidence for 1994–1998 (complete years covered by the study period) was 0.22, 0.29, 0.71, 0.19 and 0.26, respectively. Over the period included in the study there was an increase (0.0016 per year) in flystrike incidence, but this long-term linear trend was not significant (p=0.53). Accounting for the number of flocks at-risk of flystrike within each month of the study by using weighted linear regression, this trend was negative (0.0006 per year) but still non-significant (p=0.79). The incidence of flystrike was best (AICc =−5.60) described by an autoregressive model that included flystrike incidence in the previous 2 months: flystriket =0.0170+0.0392
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
313
Fig. 1. Incidence (risk) rate of flystrike in Queensland sheep flocks (open circles), estimated from woolgrower-reported pesticide use, during the period June 1993 to April 1999 and the southern oscillation index (full circles) for the corresponding period.
flystriket−1 +0.3589 flystriket−2 . No significant (p<0.05) trend or autocovariance was detected in model residuals, however partial autocovariance at lag t=14 was significant (p<0.05). Mean monthly flystrike incidence (0.036) was significantly (Kruskal–Wallis statistic =4.48, p=0.03) higher in months in which the SOI was positive than negative (0.021). Flystrike incidence was significantly (p<0.05) correlated with the SOI one, 2, 3, 4 and 6 months previously (Table 1). The strongest association (r=0.33) was found between flystrike incidence and the SOI 2 months previously, flystriket =0.0333+0.0011 SOIt−2 (AICc =−5.64), although residuals indicated non-stationarity and significant (p<0.05) autocovariance and partial autocovariance of residuals was detected at lags of t=2, 6 and 22 months. Using the SOI, the best-fitting (AICc =−5.69) autoregressive model Table 1 Cross-correlation between estimated flystrike incidence in Queensland sheep flocks between June 1993 and April 1999 and climatic parameters at lags of 0–12 months Lag
SOI
Maximum Minimum Rainfall temperature temperature
Evaporation
Radiation
Vapour pressure
0 1 2 3 4 5 6 7 8 9 10 11 12
0.15 0.27∗ 0.33** 0.28* 0.26* 0.22 0.26* 0.13 0.17 0.18 0.11 0.18 0.09
0.21 0.15 0.03 −0.07 −0.12 −0.14 −0.14 −0.10 −0.05 −0.01 0.11 0.15 0.17
0.19 0.16 0.04 −0.00 −0.07 −0.09 −0.16 −0.10 −0.08 −0.05 0.08 0.12 0.20
0.26* 0.20 0.06 −0.02 −0.11 −0.09 −0.15 −0.10 −0.10 −0.13 0.00 0.06 0.18
0.13 0.09 0.13 −0.08 −0.12 −0.21 −0.14 −0.12 −0.03 0.09 0.10 0.17 0.02
∗
p<0.05; **p<0.01.
0.17 0.14 0.07 −0.05 −0.10 −0.15 −0.14 −0.12 −0.05 0.04 0.12 0.17 0.13
0.11 0.12 0.19 0.12 −0.10 −0.05 −0.12 −0.01 −0.10 0.09 −0.03 0.21 −0.14
314
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
Fig. 2. Observed (open circles) versus predicted (full circles) flystrike incidence (risk) rate in Queensland sheep flocks during the period August 1993 to April 1999 using the model flystriket =0.0238+0.3033 flystriket−2 +0.0009 SOIt−2 .
describing flystrike incidence was flystriket =0.0238+0.3033 flystriket−2 +0.0009 SOIt−2 (AICc =−5.694). ACF and PACF of model residuals indicated stationarity and no significant (z=0.1, p=0.92) trend was detected. Observed versus predicted monthly flystrike incidence using this model during the period August 1993 to April 1999 is shown in Fig. 2. The correlation between observed and predicted flystrike incidence was 0.46. Cross-correlations between flystrike incidence and maximum and minimum temperature, rainfall, evaporation, radiation and vapour pressure are shown in Table 1. The monthly incidence of flystrike was significantly (p<0.05) correlated only with radiation (r=0.262). Moderate (r>0.2) but non-significant (p>0.05) correlation was found between flystrike incidence and average monthly maximum temperature (r=0.21), rainfallt−11 (r=0.21), evaporationt−12 (r=0.21), radiationt−1 (r=0.20) and vapour pressuret−5 (r=−0.20). The correlation matrix of average monthly maximum and minimum temperature, total rainfall, evaporation, radiation and vapour pressure is shown in Table 2. Considering these correlations, maximum temperature, rainfall, radiation and vapour pressure were chosen as independent variables to model monthly flystrike incidence. The only parameter significantly (p=0.03) associated with flystrike incidence was radiation: flystriket =−0.0176+0.0023 radiationt (AICc =−5.60). Residuals of this model were non-stationary, significant (p<0.05) Table 2 Correlation matrix for climatic parameters used in an investigation of long-term trend in the incidence of flystrike in Queensland sheep flocks
Minimum temperature Rainfall Evaporation Radiation Vapour pressure
Maximum temperature
Minimum temperature
Rainfall
Evaporation
Radiation
0.96 0.47 0.95 0.91 0.78
– 0.65 0.88 0.78 0.91
– – 0.37 0.33 0.80
– – – 0.97 0.63
– – – – 0.55
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
315
autocovariance being detected at t=2 and t=6 and significant partial autocovariance being detected at t=6. Stationarity was improved by fitting the model flystriket =−0.0259+0.3610 flystriket−2 +0.0022 radiationt (AICc =−5.67), although partial autocovariance of residual was significant (p<0.05) at t=14. However, no significant (z=0.1, p=0.92) trend was detected in model residual.
4. Discussion Results of this study suggest that a correlation between flystrike incidence and the SOI up to 2 months previously could be useful in forecasting the incidence of flystrike in Queensland sheep flocks. Other climatic parameters investigated did not appear useful for forecasting the incidence of flystrike, because of a lack of sufficient time interval (lag) or because of poor correlation. In addition, the SOI is widely available and is applicable to all areas of Queensland, whereas interpolated climatic data requires specification of point locations and interrogation of large databases. Reported application of pesticides to control blowfly strike was used as an indicator variable for flystrike occurrence. Although Ward and Armstrong (2000) have found the reported application of pesticides to be an appropriate proxy variable for the occurrence of flystrike, the possibility of misclassification bias still exists. Flystrike is an important disease that demands treatment on both welfare and economic grounds. Thus, failure by managers to treat sheep with pesticides when blowfly strike occurs is unlikely. Application of pesticides to control flystrike (in contrast to preventive applications) if flystrike has not occurred is also unlikely. Also, there is no reason to suspect that misclassification of flocks (with respect to the reason for pesticide application for flystrike) was associated with the risk factors examined in this study, namely climatic variables. The potential affect of response bias in the survey from which data was used to construct the model also needs to be considered. The overall response rate (69%) was considered high for epidemiologic surveys of this size (Vaillancourt et al., 1991). The questionnaire used was titled ’Sheep jetting and dipping questionnaire’. Wool producers were requested to provide information on a wide range of variables, including flock demographics, shearing practices, sheep lice and blowfly infestation and the use of pesticides. Given that questions on the control of flystrike were only a part of this survey, response bias with respect to the occurrence of flystrike is unlikely to be substantial and probably would not influence the model selected in this study. Flocks selected for study were not necessarily representative of the Queensland wool industry, and results should not be applied to flocks in which preventive pesticide application for blowfly strike is practised. No useful relationships between interpolated climatic data and flystrike incidence were identified in this study. Flystrike in Australian sheep flocks is influenced by climatic factors (in particular, rainfall and temperature) and by the presence of susceptible sheep (Wardhaugh and Morton, 1990). Failure to identify strong correlations between flystrike incidence and variables such as temperature and rainfall may indicate that these factors operate at a local, flock-level and that the broad climatic conditions existing in Queensland in any given month is not an accurate predictor of flystrike within individual flocks. The process of interpolation may also introduce some error into climatic data used.
316
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
Although the association between climate and blowfly strike has been investigated previously in Australia (and predictive models have been developed), these models have not been adopted by disease control authorities (Tellam and Bowles, 1997). For example, Wardhaugh and Morton (1990) and Wardhaugh et al. (1994) used variables such as soil moisture and soil temperature to predict the occurrence of blowfly strike. Such data may be collected at research stations, but it is not routinely available across regions of western Queensland where sheep are grazed. Fenton et al. (1998) developed a model of flystrike by L. sericata on British sheep farms, based on simulated blowfly abundance and sheep susceptibility. Input parameters included daily rainfall and maximum and minimum temperature data, pasture worm counts and fleece length. Although the resulting model appeared realistic and was useful for investigating the interaction of parameters on flystrike, it could not be used for predictive purposes. Another reason for lack of model adoption is insufficient predictive ability over a period of more than 1 month (Tellam and Bowles, 1997). Thus, the adoption of forecasting systems may be improved by the use of variables for which data are available at any flock location, and by identifying associations which will allow flystrike to be predicted 2 months or more in advance. Study results suggest that the SOI is such a variable, since a single value is available for all locations and its value up to 6 months previously is moderately (r>0.2) correlated with flystrike incidence. Such a forecasting system could be easily operated and forecasts would be simple and easy to interpret by woolgrowers. Associations between the SOI and arboviral diseases, such as Murray Valley encephalitis, dengue fever and Ross River fever, have been described previously and considered useful for providing simple, objective early-warning systems (Nicholls, 1986; Hales et al., 1999; Maelzer et al., 1999). Investigation of the SOI to predict parasitic diseases does not appear to have been reported. Forecasting systems have been used previously to predict the occurrence of parasitic infestation in sheep flocks. For example, Flasse et al. (1998) used remotely-sensed temperature data to predict outbreaks of the nasal fly (Oestrus ovis) in Namibia, Gardiner and Gettinby (1983) used climatic parameters to predict activity of the sheep tick (Ixodes ricinus) in Ireland, and predictive models of gastrointestinal parasites in sheep have been developed (for example, those of Dobson et al., 1990). The development of such forecasting systems is motivated by a need to increase efficiency of control measures, particular the use of parasiticides. Since most parasiticides have a limited duration of effectiveness, timing their application to coincide with a predicted increased risk of infestation may reduce the amount of parasiticide used. Consequently, residues in food and fibre, pesticide resistance in the target parasite species, and occupational health and safety risks can be minimised (Karlsson, 1999). Such systems need to be accurate, but forecasts need to be simple and timely. The challenge is to identify associations that can potentially be incorporated into a forecasting system that provides woolgrowers with sufficient warning of increasing risk of flystrike. Study results suggest that the SOI should be considered for inclusion in forecasting systems of flystrike in Queensland sheep flocks. The model including the SOI described in this study is likely to be easier to apply than previous models of blowfly strike and would provide a greater period of warning for wool producers. If used as an extension tool, it could lead to more effective control of flystrike in Queensland sheep flocks, when combined with other information delivered to woolgrowers in an integrated parasite management package.
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
317
Acknowledgements Alan Beswick and Keith Moodie, Queensland Department of Natural Resources, Brisbane are thanked for providing interpolated climate data. Robert Armstrong, Queensland Department of Primary Industries is thanked for making survey data available for use in this study. References Australian Wool Exchange, 1999. Australian wool statistics yearbook: 1998–1999 season. Australian Wool Exchange Market Reporting Service, Melbourne, pp. 42–43. Dobson, R.J., Donald, A.D., Barnes, E.H., Waller, P.J., 1990. Population dynamics of Trichostrongylus colubriformis in sheep: model to predict the worm population over time as a function of infection rate and host age. Int. J. Parasitol. 20, 365–373. Fenton, A., Wall, R., French, N., 1998. The incidence of sheep strike by Lucilia sericata on sheep farms in Britain: a simulation model. Vet. Parasitol. 76, 211–228. Flasse, S., Walker, C., Biggs, H., Stephenson, P., Hutchinson, P., 1998. Using remote sensing to predict outbreaks of Oestrus ovis in Namibia. Prev. Vet. Med. 33, 31–38. Gardiner, W.P., Gettinby, G., 1983. A weather based prediction model for the life-cycle of the sheep tick, Ixodes ricinus L.. Vet. Parasitol. 13, 77–84. Hales, S., Weinstein, P., Souares, Y., Woodward, A., 1999. El Niño and the dynamics of vectorborne disease transmission. Environ. Health Perspect. 107, 99–102. Karlsson, L.J.E., 1999. Sustainable sheep ectoparasitic control using IPM. In: Proceedings of the Australian Sheep Veterinarian Society Conference, Hobart 1999, pp. 128–133. McLeod, R.S., 1995. Costs of major parasites to the Australian livestock industries. Int. J. Parasitol. 25, 1363–1367. Maelzer, D., Hales, S., Weinstein, P., Zalucki, M., Woodward, A., 1999. El Niño and arboviral disease prediction. Environ. Health Perspect. 107, 817–818. Nicholls, N., 1986. A method for predicting Murray Valley encephalitis in southeast Australia using the southern oscillation. Aust. J. Exp. Biol. Med. Sci. 64, 587–594. Norris, K.R., 1966. Daily patterns of flight activity of blowflies (Calliphoridae: Diptera) in the Canberra district as indicated by trap catches. Aust. J. Zool. 14, 835–853. Tellam, R.L., Bowles, V.M., 1997. Control of blowfly strike in sheep: current strategies and future prospects. Int. J. Parasitol. 27, 261–273. Troup, A.J., 1965. The southern oscillation. Quart. J. R. Meteorol. Soc. 91, 490–506. Vaillancourt, J-P., Martineau, G., Morrow, M., Marsh, W., Robinson, A., 1991. Construction of questionnaires and their use in veterinary medicine. In: Thrusfield, M.V. (Ed.), Proceedings of the Society for Veterinary Epidemiology and Preventive medicine Annual Conference. London, 1991, pp. 94–106. Voice, M.E., 1991. The southern oscillation. In: Partridge, I. (Ed.), Will it Rain? El Niño and the Southern Oscillation. Information Series QI91028, Queensland Department of Primary Industries, Brisbane, pp. 12–19. Ward, M.P., Armstrong, R.T.F., 1998a. Pesticide use and residues on Queensland wool. Aust. Vet. J. 76, 739–742. Ward, M.P., Armstrong, R.T.F., 1998b. Trends in the use of pesticides and pesticide residues on Queensland wool. Aust. Vet. J. 76, 694–697. Ward, M.P., Armstrong, R.T.F., 2000. Time-space clustering of reported blowfly strike in Queensland sheep flocks. Prev. Vet. Med. 43, 195–202. Wardhaugh, K.G., Morton, R., 1990. The incidence of flystrike in sheep in relation to weather conditions, sheep husbandry, and the abundance of the Australian sheep blowfly, Lucilia cuprina (Wiedemann) (Diptera: Calliphoridae). Aust. J. Agric. Res. 41, 1155–1167. Wardhaugh, K.G., Bedo, D., Vogt, W.G., 1994. Management model for the control of flystrike on sheep. In: Proceedings of the Australian Sheep Veterinarian Society Conference, Canberra 1994, pp. 70–73. Watts, J.E., Merritt, G.C., Goodrich, B.S., 1981. The ovipositional response of the Australian sheep blowfly, Lucilia cuprina, to fleece-rot odours. Aust. Vet. J. 57, 450–454. Williams, S.H., Brightling, A., 1999. The Australian wool industry’s response to the issue of pesticides residues. In: Proceedings of the Australian Sheep Veterinarian Society Conference, Hobart 1999, pp. 77–84.
Forecasting blowfly strike in Queensland sheep flocks Michael P. Ward∗ Queensland Department of Primary Industries, Animal Research Institute, Locked Mail Bag 4, Moorooka QLD 4105, Australia Received 18 February 2000; received in revised form 12 May 2000; accepted 23 May 2000
Abstract The occurrence of blowfly strike between 1993 and 1999, derived from the reported use of pesticides for flystrike control, was investigated in 247 sheep flocks in Queensland, Australia using autoregressive techniques. Although there was a small increase (0.0016 per year) in flystrike incidence during the study period, this long-term linear trend was not significant (p=0.53). The occurrence of flystrike was best described by an autoregressive model that included flystrike in the previous 2 months: flystriket =0.0170+0.0392 flystriket−1 +0.3589 flystriket−2 . Flystrike was associated with the southern oscillation index (SOI). The SOI is based on barometric pressure readings and is associated with periods of below- (negative SOI) and above-average (positive SOI) rainfall in northern Australia. Flystrike incidence was significantly (p=0.03) greater in months in which the SOI was positive. The strongest correlation (r=0.33) was found between flystrike incidence and the SOI 2 months previously. Using the SOI, the best-fitting autoregressive model describing flystrike was flystriket =0.0238+0.3033 flystriket−2 +0.0009 SOIt−2 . The incidence of flystrike was significantly (p<0.05) correlated with average monthly radiation (r=0.26), but not with average monthly maximum and minimum temperature, total rainfall, evaporation and vapour pressure. The best-fitting autoregressive model describing flystrike occurrence based on these variables was flystriket =−0.0259+0.3610 flystriket−2 +0.0022 radiationt . Results suggest that a useful early-warning system could be developed based on the correlation between flystrike incidence and the SOI up to 2 months previously. Such attempts to forecast flystrike may assist decision-making by wool producers with respect to flystrike control options, leading to more efficient control of blowfly strike in their industry. © 2000 Elsevier Science B.V. All rights reserved. Keywords: Lucilia cuprina; Blowfly strike; Sheep-arthropoda; Climate; Time-series analysis; Southern oscillation index; Australia
∗ Present address: Department of Veterinary Pathobiology, School of Veterinary Medicine, Purdue University, West Lafayette IN 47907, USA. Tel.: +61-765-494-5796; fax: +61-765-494-9830. E-mail address: [email protected] (M.P. Ward).
0304-4017/00/$ – see front matter © 2000 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 4 0 1 7 ( 0 0 ) 0 0 2 8 3 - 1
310
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
1. Introduction The occurrence of blowfly strike in the Australian wool industry has major animal-welfare implications, and causes considerable economic loss (McLeod, 1995). Numerous strategies have been devised by woolgrowers to control flystrike, including husbandry procedures (e.g. crutching and the mules operation), management procedures (e.g. strategic shearing and breeding sheep for resistance to flystrike), the use of pesticides and blowfly baits and traps (Tellam and Bowles, 1997). Pesticides are a popular control method. In recent Queensland studies >80% of woolgrowers surveyed used pesticides to control flystrike in their flocks (Ward and Armstrong, 1998a), and the use of newer insect growth regulator pesticides appears to be replacing organophosphorous compounds for flystrike control (Ward and Armstrong, 1998b). However, such pesticide use has a number of disadvantages, including potential environmental damage caused by scouring wool that contains pesticide residues and occupational-health concerns about handling pesticide-treated sheep. Strategies aimed at reducing the use of pesticides are considered a priority in the Australian wool industry (Williams and Brightling, 1999). Most (>80 to 90%) strikes in Australian sheep flocks are caused by Lucilia cuprina. Its lifecycle is dependent on climatic factors: adult flies are active in the temperature range of 17–35◦ C (Norris, 1966), and moisture increases the survival of larvae and pupae in the soil, promotes plant growth (providing a food source for newly emerged adult flies) and predisposes sheep to fleece rot (Watts et al., 1981). This dependency can potentially be used to develop models to forecast the occurrence of flystrike, thus assisting woolgrowers to control flystrike more effectively using an integrated parasite management program with less reliance on pesticides (Karlsson, 1999). The aim of this study was to investigate the long-term associations between blowfly strike and the southern oscillation index (SOI) and a range of climatic parameters, and to identify associations that may prove useful in developing a forecasting system of flystrike in Queensland sheep flocks. 2. Materials and methods 2.1. Data source The occurrence of blowfly strike in Queensland sheep flocks was determined from information provided by woolgrowers in three postal surveys, conducted between 1995 and 1999, to assess the use of pesticides in the Queensland wool industry (Ward and Armstrong, 1998a; Ward and Armstrong, unpublished data). The sampling frame was all Queensland sheep flocks from which wool had been sold at auction during this time period. Flocks were randomly sampled from the sampling frame through a process of selection of wool lot numbers. Flock managers were sent a questionnaire seeking information on whether or not pesticides had been applied for the control of blowfly strike, when (months) application had occurred after shearing, and the reason for pesticide application (preventive or in response to the occurrence of flystrike). As part of the questionnaire surveys, information was also sought on when (month and year) shearing had most recently taken place and the months of wool growth at this shearing (thus enabling the month and year of the previous shearing to be calculated).
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
311
2.2. Data analysis The reported application of pesticides by flock managers in response to blowfly strike (in contrast to the preventive use of pesticides for flystrike) was used as an indicator variable of blowfly strike. In flocks in which managers indicated that pesticides were not applied for flystrike control it was assumed that flystrike did not occur (since the occurrence of flystrike generally demands use of pesticides for both animal welfare and economic reasons). For all flocks in which flystrike did or did not occur, the period of risk was defined as the period extending from the time of shearing to either when flystrike first occurred (pesticides reported to have been applied in response to flystrike) or the subsequent shearing, respectively. Flystrike incidence was estimated using the number of flocks in which flystrike occurred (indicated by the use of pesticides in response to flystrike) in each month as the numerator and the total number of flocks at-risk of flystrike in each month as the denominator. Flystrike was assumed to occur within the month pesticide treatment was applied, since most strikes are initiated in the week preceding their report by woolgrowers (Wardhaugh and Morton, 1990). Thus, a time-series of flystrike incidence (risk) rate was created. Data describing climatic conditions within Queensland sheep-raising districts during the study period were obtained by interrogation of interpolated climate surfaces (Queensland Centre for Climate Applications, Department of Natural Resources, Brisbane; http://www. dnr.qld.gov.au/silo/datadril.html). Interpolated data were derived from historically recorded point data (Bureau of Meteorology recording stations), interpolated to 0.05◦ (≈5 km), using spline and krigging methods. To represent Queensland sheep-raising districts, the approximate centre (latitude, longitude) of 11 wool districts (Q6, 12, 14, 15, 18, 20, 21, 22, 23, 27 and 28) from which >90% of Queensland wool was produced in 1998/1999 (Australian Wool Exchange, 1999) were selected. At these sites, climatic data were interpolated for daily maximum temperature (◦ C), minimum temperature, rainfall (mm), evaporation (mm), radiation (MJ/m2 ) and vapour pressure (hPa) for each month of the flystrike time-series. For temperature, evaporation, radiation and vapour pressure variables, mean values were calculated from daily values for each month. For rainfall, a total value was calculated from daily values for each month. In addition, the SOI for all months of the time-series was identified. The SOI is based on the difference in barometric pressure readings measured at sites near Darwin (northern Australia) and Tahiti. It is associated with periods of belowand above-average rainfall in northern Australia. For example, when the SOI is strongly negative, tradewinds are weak and rainfall in northern Australia tends to be below-average with drought conditions often developing. This situation is popularly called an El Niño event (Voice, 1991). Troup’s SOI was used, (PA(Tahiti)−PA(Darwin))/s.d., where PA is the pressure anomaly (monthly mean−long-term mean) and s.d. is the standard deviation of the difference (Troup, 1965; http://www.dnr.qld.gov.au/longpdk/lpsoidat.htm#troup). The best autoregressive model (t, t−1, . . . , t−12) describing flystrike incidence was found based on minimum variance and number of model parameters, as determined by the goodness-of-fit statistic Akaike’s corrected information criterion, AICc (ASTSA for Windows, version 1, Division of Statistics, University of California, Davis). This autoregressive model identified flystrike in which previous months best described flystrike in the current month, t. The significance of model coefficients was tested using a t-statistic. Plots of the autocovariance function (ACF) and the partial autocovariance function (PACF) of model
312
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
residuals were examined to ensure stationarity and that the selected model adequately described the time-series. Statistical significance of the ACF and PACF was assessed using 95% confidence intervals. The fluctuation (trend) test was used to test for randomness of model residuals (ASTSA for Windows, version 1, Division of Statistics, University of California, Davis). The cross-correlation between flystrike incidence and the set of climatic parameters was estimated using lags (t) of up to 12 months (t, t−1, . . . , t−12). In addition, the association between flystrike incidence and positive or negative SOI was tested using Kruskal–Wallis nonparametric analysis-of-variance. Correlations between temperature, rainfall, evaporation, radiation and vapour pressure parameters were calculated. Evidence of highly correlated parameters (r>0.95) was used to delete redundant parameters from model building. Vector autoregression was used to identify the most appropriate model describing the association between flystrike incidence and the SOI, and between flystrike incidence and the other selected climatic parameters. Flystrike incidence, SOI and climatic parameter lags of up to t=12 were investigated. The significance of model coefficients was tested using a t-statistic. AICc was used as a goodness-of-fit statistic, and the ACF and PACF of residuals of selected models were used to assess stationarity. The fluctuation (trend) test was used assess series randomness (ASTSA for Windows, version 1, Division of Statistics, University of California, Davis).
3. Results Questionnaires were sent to 833 Queensland woolgrowers, representing 32% of Queensland sheep flocks. A total of 576 (69%) questionnaires were returned. Eighty-six woolgrowers reported applying pesticides to flocks when blowfly strike occurred (rather than applying pesticides preventively for flystrike) and 207 reported no application of pesticides (neither preventive nor control) for flystrike. This subset of 293 flocks was used to construct a time-series of flystrike incidence in Queensland sheep flocks. Month of pesticide application, time of shearing and months of wool growth was available for 73 (85%) of the flocks in which pesticides were used to control flystrike. Time of shearing and months of wool growth were available for 174 (84%) of the flocks in which pesticides were not used for flystrike. Therefore, the flystrike incidence time-series was constructed using data from 247 flocks. The period during which one or more flocks were at-risk of flystrike was June 1993 to April 1999, or 71 months. The mean number of flocks at-risk of flystrike in each study month was 38. The time-series of estimated flystrike incidence in Queensland sheep flocks is shown in Fig. 1. Estimated annual incidence for 1994–1998 (complete years covered by the study period) was 0.22, 0.29, 0.71, 0.19 and 0.26, respectively. Over the period included in the study there was an increase (0.0016 per year) in flystrike incidence, but this long-term linear trend was not significant (p=0.53). Accounting for the number of flocks at-risk of flystrike within each month of the study by using weighted linear regression, this trend was negative (0.0006 per year) but still non-significant (p=0.79). The incidence of flystrike was best (AICc =−5.60) described by an autoregressive model that included flystrike incidence in the previous 2 months: flystriket =0.0170+0.0392
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
313
Fig. 1. Incidence (risk) rate of flystrike in Queensland sheep flocks (open circles), estimated from woolgrower-reported pesticide use, during the period June 1993 to April 1999 and the southern oscillation index (full circles) for the corresponding period.
flystriket−1 +0.3589 flystriket−2 . No significant (p<0.05) trend or autocovariance was detected in model residuals, however partial autocovariance at lag t=14 was significant (p<0.05). Mean monthly flystrike incidence (0.036) was significantly (Kruskal–Wallis statistic =4.48, p=0.03) higher in months in which the SOI was positive than negative (0.021). Flystrike incidence was significantly (p<0.05) correlated with the SOI one, 2, 3, 4 and 6 months previously (Table 1). The strongest association (r=0.33) was found between flystrike incidence and the SOI 2 months previously, flystriket =0.0333+0.0011 SOIt−2 (AICc =−5.64), although residuals indicated non-stationarity and significant (p<0.05) autocovariance and partial autocovariance of residuals was detected at lags of t=2, 6 and 22 months. Using the SOI, the best-fitting (AICc =−5.69) autoregressive model Table 1 Cross-correlation between estimated flystrike incidence in Queensland sheep flocks between June 1993 and April 1999 and climatic parameters at lags of 0–12 months Lag
SOI
Maximum Minimum Rainfall temperature temperature
Evaporation
Radiation
Vapour pressure
0 1 2 3 4 5 6 7 8 9 10 11 12
0.15 0.27∗ 0.33** 0.28* 0.26* 0.22 0.26* 0.13 0.17 0.18 0.11 0.18 0.09
0.21 0.15 0.03 −0.07 −0.12 −0.14 −0.14 −0.10 −0.05 −0.01 0.11 0.15 0.17
0.19 0.16 0.04 −0.00 −0.07 −0.09 −0.16 −0.10 −0.08 −0.05 0.08 0.12 0.20
0.26* 0.20 0.06 −0.02 −0.11 −0.09 −0.15 −0.10 −0.10 −0.13 0.00 0.06 0.18
0.13 0.09 0.13 −0.08 −0.12 −0.21 −0.14 −0.12 −0.03 0.09 0.10 0.17 0.02
∗
p<0.05; **p<0.01.
0.17 0.14 0.07 −0.05 −0.10 −0.15 −0.14 −0.12 −0.05 0.04 0.12 0.17 0.13
0.11 0.12 0.19 0.12 −0.10 −0.05 −0.12 −0.01 −0.10 0.09 −0.03 0.21 −0.14
314
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
Fig. 2. Observed (open circles) versus predicted (full circles) flystrike incidence (risk) rate in Queensland sheep flocks during the period August 1993 to April 1999 using the model flystriket =0.0238+0.3033 flystriket−2 +0.0009 SOIt−2 .
describing flystrike incidence was flystriket =0.0238+0.3033 flystriket−2 +0.0009 SOIt−2 (AICc =−5.694). ACF and PACF of model residuals indicated stationarity and no significant (z=0.1, p=0.92) trend was detected. Observed versus predicted monthly flystrike incidence using this model during the period August 1993 to April 1999 is shown in Fig. 2. The correlation between observed and predicted flystrike incidence was 0.46. Cross-correlations between flystrike incidence and maximum and minimum temperature, rainfall, evaporation, radiation and vapour pressure are shown in Table 1. The monthly incidence of flystrike was significantly (p<0.05) correlated only with radiation (r=0.262). Moderate (r>0.2) but non-significant (p>0.05) correlation was found between flystrike incidence and average monthly maximum temperature (r=0.21), rainfallt−11 (r=0.21), evaporationt−12 (r=0.21), radiationt−1 (r=0.20) and vapour pressuret−5 (r=−0.20). The correlation matrix of average monthly maximum and minimum temperature, total rainfall, evaporation, radiation and vapour pressure is shown in Table 2. Considering these correlations, maximum temperature, rainfall, radiation and vapour pressure were chosen as independent variables to model monthly flystrike incidence. The only parameter significantly (p=0.03) associated with flystrike incidence was radiation: flystriket =−0.0176+0.0023 radiationt (AICc =−5.60). Residuals of this model were non-stationary, significant (p<0.05) Table 2 Correlation matrix for climatic parameters used in an investigation of long-term trend in the incidence of flystrike in Queensland sheep flocks
Minimum temperature Rainfall Evaporation Radiation Vapour pressure
Maximum temperature
Minimum temperature
Rainfall
Evaporation
Radiation
0.96 0.47 0.95 0.91 0.78
– 0.65 0.88 0.78 0.91
– – 0.37 0.33 0.80
– – – 0.97 0.63
– – – – 0.55
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
315
autocovariance being detected at t=2 and t=6 and significant partial autocovariance being detected at t=6. Stationarity was improved by fitting the model flystriket =−0.0259+0.3610 flystriket−2 +0.0022 radiationt (AICc =−5.67), although partial autocovariance of residual was significant (p<0.05) at t=14. However, no significant (z=0.1, p=0.92) trend was detected in model residual.
4. Discussion Results of this study suggest that a correlation between flystrike incidence and the SOI up to 2 months previously could be useful in forecasting the incidence of flystrike in Queensland sheep flocks. Other climatic parameters investigated did not appear useful for forecasting the incidence of flystrike, because of a lack of sufficient time interval (lag) or because of poor correlation. In addition, the SOI is widely available and is applicable to all areas of Queensland, whereas interpolated climatic data requires specification of point locations and interrogation of large databases. Reported application of pesticides to control blowfly strike was used as an indicator variable for flystrike occurrence. Although Ward and Armstrong (2000) have found the reported application of pesticides to be an appropriate proxy variable for the occurrence of flystrike, the possibility of misclassification bias still exists. Flystrike is an important disease that demands treatment on both welfare and economic grounds. Thus, failure by managers to treat sheep with pesticides when blowfly strike occurs is unlikely. Application of pesticides to control flystrike (in contrast to preventive applications) if flystrike has not occurred is also unlikely. Also, there is no reason to suspect that misclassification of flocks (with respect to the reason for pesticide application for flystrike) was associated with the risk factors examined in this study, namely climatic variables. The potential affect of response bias in the survey from which data was used to construct the model also needs to be considered. The overall response rate (69%) was considered high for epidemiologic surveys of this size (Vaillancourt et al., 1991). The questionnaire used was titled ’Sheep jetting and dipping questionnaire’. Wool producers were requested to provide information on a wide range of variables, including flock demographics, shearing practices, sheep lice and blowfly infestation and the use of pesticides. Given that questions on the control of flystrike were only a part of this survey, response bias with respect to the occurrence of flystrike is unlikely to be substantial and probably would not influence the model selected in this study. Flocks selected for study were not necessarily representative of the Queensland wool industry, and results should not be applied to flocks in which preventive pesticide application for blowfly strike is practised. No useful relationships between interpolated climatic data and flystrike incidence were identified in this study. Flystrike in Australian sheep flocks is influenced by climatic factors (in particular, rainfall and temperature) and by the presence of susceptible sheep (Wardhaugh and Morton, 1990). Failure to identify strong correlations between flystrike incidence and variables such as temperature and rainfall may indicate that these factors operate at a local, flock-level and that the broad climatic conditions existing in Queensland in any given month is not an accurate predictor of flystrike within individual flocks. The process of interpolation may also introduce some error into climatic data used.
316
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
Although the association between climate and blowfly strike has been investigated previously in Australia (and predictive models have been developed), these models have not been adopted by disease control authorities (Tellam and Bowles, 1997). For example, Wardhaugh and Morton (1990) and Wardhaugh et al. (1994) used variables such as soil moisture and soil temperature to predict the occurrence of blowfly strike. Such data may be collected at research stations, but it is not routinely available across regions of western Queensland where sheep are grazed. Fenton et al. (1998) developed a model of flystrike by L. sericata on British sheep farms, based on simulated blowfly abundance and sheep susceptibility. Input parameters included daily rainfall and maximum and minimum temperature data, pasture worm counts and fleece length. Although the resulting model appeared realistic and was useful for investigating the interaction of parameters on flystrike, it could not be used for predictive purposes. Another reason for lack of model adoption is insufficient predictive ability over a period of more than 1 month (Tellam and Bowles, 1997). Thus, the adoption of forecasting systems may be improved by the use of variables for which data are available at any flock location, and by identifying associations which will allow flystrike to be predicted 2 months or more in advance. Study results suggest that the SOI is such a variable, since a single value is available for all locations and its value up to 6 months previously is moderately (r>0.2) correlated with flystrike incidence. Such a forecasting system could be easily operated and forecasts would be simple and easy to interpret by woolgrowers. Associations between the SOI and arboviral diseases, such as Murray Valley encephalitis, dengue fever and Ross River fever, have been described previously and considered useful for providing simple, objective early-warning systems (Nicholls, 1986; Hales et al., 1999; Maelzer et al., 1999). Investigation of the SOI to predict parasitic diseases does not appear to have been reported. Forecasting systems have been used previously to predict the occurrence of parasitic infestation in sheep flocks. For example, Flasse et al. (1998) used remotely-sensed temperature data to predict outbreaks of the nasal fly (Oestrus ovis) in Namibia, Gardiner and Gettinby (1983) used climatic parameters to predict activity of the sheep tick (Ixodes ricinus) in Ireland, and predictive models of gastrointestinal parasites in sheep have been developed (for example, those of Dobson et al., 1990). The development of such forecasting systems is motivated by a need to increase efficiency of control measures, particular the use of parasiticides. Since most parasiticides have a limited duration of effectiveness, timing their application to coincide with a predicted increased risk of infestation may reduce the amount of parasiticide used. Consequently, residues in food and fibre, pesticide resistance in the target parasite species, and occupational health and safety risks can be minimised (Karlsson, 1999). Such systems need to be accurate, but forecasts need to be simple and timely. The challenge is to identify associations that can potentially be incorporated into a forecasting system that provides woolgrowers with sufficient warning of increasing risk of flystrike. Study results suggest that the SOI should be considered for inclusion in forecasting systems of flystrike in Queensland sheep flocks. The model including the SOI described in this study is likely to be easier to apply than previous models of blowfly strike and would provide a greater period of warning for wool producers. If used as an extension tool, it could lead to more effective control of flystrike in Queensland sheep flocks, when combined with other information delivered to woolgrowers in an integrated parasite management package.
M.P. Ward / Veterinary Parasitology 92 (2000) 309–317
317
Acknowledgements Alan Beswick and Keith Moodie, Queensland Department of Natural Resources, Brisbane are thanked for providing interpolated climate data. Robert Armstrong, Queensland Department of Primary Industries is thanked for making survey data available for use in this study. References Australian Wool Exchange, 1999. Australian wool statistics yearbook: 1998–1999 season. Australian Wool Exchange Market Reporting Service, Melbourne, pp. 42–43. Dobson, R.J., Donald, A.D., Barnes, E.H., Waller, P.J., 1990. Population dynamics of Trichostrongylus colubriformis in sheep: model to predict the worm population over time as a function of infection rate and host age. Int. J. Parasitol. 20, 365–373. Fenton, A., Wall, R., French, N., 1998. The incidence of sheep strike by Lucilia sericata on sheep farms in Britain: a simulation model. Vet. Parasitol. 76, 211–228. Flasse, S., Walker, C., Biggs, H., Stephenson, P., Hutchinson, P., 1998. Using remote sensing to predict outbreaks of Oestrus ovis in Namibia. Prev. Vet. Med. 33, 31–38. Gardiner, W.P., Gettinby, G., 1983. A weather based prediction model for the life-cycle of the sheep tick, Ixodes ricinus L.. Vet. Parasitol. 13, 77–84. Hales, S., Weinstein, P., Souares, Y., Woodward, A., 1999. El Niño and the dynamics of vectorborne disease transmission. Environ. Health Perspect. 107, 99–102. Karlsson, L.J.E., 1999. Sustainable sheep ectoparasitic control using IPM. In: Proceedings of the Australian Sheep Veterinarian Society Conference, Hobart 1999, pp. 128–133. McLeod, R.S., 1995. Costs of major parasites to the Australian livestock industries. Int. J. Parasitol. 25, 1363–1367. Maelzer, D., Hales, S., Weinstein, P., Zalucki, M., Woodward, A., 1999. El Niño and arboviral disease prediction. Environ. Health Perspect. 107, 817–818. Nicholls, N., 1986. A method for predicting Murray Valley encephalitis in southeast Australia using the southern oscillation. Aust. J. Exp. Biol. Med. Sci. 64, 587–594. Norris, K.R., 1966. Daily patterns of flight activity of blowflies (Calliphoridae: Diptera) in the Canberra district as indicated by trap catches. Aust. J. Zool. 14, 835–853. Tellam, R.L., Bowles, V.M., 1997. Control of blowfly strike in sheep: current strategies and future prospects. Int. J. Parasitol. 27, 261–273. Troup, A.J., 1965. The southern oscillation. Quart. J. R. Meteorol. Soc. 91, 490–506. Vaillancourt, J-P., Martineau, G., Morrow, M., Marsh, W., Robinson, A., 1991. Construction of questionnaires and their use in veterinary medicine. In: Thrusfield, M.V. (Ed.), Proceedings of the Society for Veterinary Epidemiology and Preventive medicine Annual Conference. London, 1991, pp. 94–106. Voice, M.E., 1991. The southern oscillation. In: Partridge, I. (Ed.), Will it Rain? El Niño and the Southern Oscillation. Information Series QI91028, Queensland Department of Primary Industries, Brisbane, pp. 12–19. Ward, M.P., Armstrong, R.T.F., 1998a. Pesticide use and residues on Queensland wool. Aust. Vet. J. 76, 739–742. Ward, M.P., Armstrong, R.T.F., 1998b. Trends in the use of pesticides and pesticide residues on Queensland wool. Aust. Vet. J. 76, 694–697. Ward, M.P., Armstrong, R.T.F., 2000. Time-space clustering of reported blowfly strike in Queensland sheep flocks. Prev. Vet. Med. 43, 195–202. Wardhaugh, K.G., Morton, R., 1990. The incidence of flystrike in sheep in relation to weather conditions, sheep husbandry, and the abundance of the Australian sheep blowfly, Lucilia cuprina (Wiedemann) (Diptera: Calliphoridae). Aust. J. Agric. Res. 41, 1155–1167. Wardhaugh, K.G., Bedo, D., Vogt, W.G., 1994. Management model for the control of flystrike on sheep. In: Proceedings of the Australian Sheep Veterinarian Society Conference, Canberra 1994, pp. 70–73. Watts, J.E., Merritt, G.C., Goodrich, B.S., 1981. The ovipositional response of the Australian sheep blowfly, Lucilia cuprina, to fleece-rot odours. Aust. Vet. J. 57, 450–454. Williams, S.H., Brightling, A., 1999. The Australian wool industry’s response to the issue of pesticides residues. In: Proceedings of the Australian Sheep Veterinarian Society Conference, Hobart 1999, pp. 77–84.