- Email: [email protected]
Models for the dispersal in Australia of the arbovirus vector, Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae)
Preventive Veterinary Medicine 47 (2000) 243±254
Models for the dispersal in Australia of the arbovirus vector, Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae) Alan L. Bishop*, Idris M. Barchia, Lorraine J. Spohr NSW Agriculture, Locked Bag 26, Gosford, NSW 2250, Australia Received 24 March 2000; accepted 27 August 2000
Abstract Culicoides brevitarsis is the main biting midge responsible for the transmission of bluetongue and Akabane viruses to livestock in Australia. Models are given for its dispersal after winter from endemic areas at the southern limit of its distribution in New South Wales (NSW); the models might also be applicable elsewhere. Model 1 shows that dispersal can be explained by distance from a key point just outside the endemic area in mid-northern/northern coastal NSW. The model provides probability data for times of first occurrence at sites within regions down the southern coastal plain or up the Hunter Valley towards (but rarely reaching) the western slopes and tablelands. Model 2 shows that the movement depends on temperature and wind speed from northerly and easterly directions. Preliminary data also are given to suggest a relationship between density in the endemic area and the maximum distance that C. brevitarsis can travel in a given year. The models can be linked to other information which in combination can provide probabilities for winter survival outside the endemic area, times of occurrence at sites where it cannot survive winter and times when activity ceases naturally at these sites at the end of the season. This information can be used to predict the potential for virus transmission and indicate zones of seasonal freedom from both vector and virus for the export of livestock. # 2000 Elsevier Science B.V. All rights reserved. Keywords: Culicoides brevitarsis; Dispersal; Vector; Arboviruses; Australia
1. Introduction Insects are endemic where temperature, moisture, habitat and feeding conditions are consistently suitable for their survival, growth and reproduction. Distributions beyond endemic areas can sometimes be accidental but often occur as a result of the insects' *
Corresponding author. Tel.: 61-2-4348-1928; fax: 61-2-4348-1910. E-mail address: [email protected] (A.L. Bishop). 0167-5877/00/$ ± see front matter # 2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 7 - 5 8 7 7 ( 0 0 ) 0 0 1 7 5 - 6
244
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
passive or active movements into areas that are favourable seasonally. Movement can be associated with wind and weather activity or be aided by mechanical means of transport (Fraval and El-Yousfi, 1989; Loxdale et al., 1993; Dillon et al., 1996). Several species of biting midge from the genus Culicoides (Diptera: Ceratopogonidae) transmit viruses to livestock and native fauna (Doherty et al., 1978; Muller et al., 1982; Standfast et al., 1984). Epidemiological studies suggest that the spread of many viruses may be related to wind-borne dispersal of the midge vectors (Sellers et al., 1977, 1992; Braverman, 1992). In Australia, Culicoides brevitarsis Kieffer is the main vector of the bluetongue and Akabane viruses that affect the health and export of livestock (Muller et al., 1982). Local dispersal of C. brevitarsis in the wind is an integral part of its biology. Flight, orientation and relationships with animals depend on temperature and the speed and direction of wind (Murray, 1987a). Retrospective evidence supports the occasional, long-distance and wind-borne dispersal of infected C. brevitarsis (Murray, 1987b; Murray and Kirkland, 1995). C. brevitarsis is endemic on the mid-northern/northern coastal plains of New South Wales (NSW). It disperses seasonally and normally retains a coastal distribution (Bishop et al., 1996a) Ð although occasionally it has been found inland (Muller et al., 1982). The GROWEST model of Nix (1976) was modified by Murray and Nix (1987) to explain and predict these occurrences. This model proposes areas suitable for summer multiplication and winter survival (based on temperature and rainfall) that can be enclosed within `brevitarsis lines'. Implicit in the model's concept is the existence of over-wintering foci from which both vector and virus develop. While biologically possible at some locations (Bishop et al., 1995a), development and movement from over-wintering foci have never been observed outside the endemic area in NSW between 1990 and 1999 (Bishop et al., 1995b, 1996a, unpublished data). This can be explained by low probabilities for survival based on temperatures and the need for C. brevitarsis to establish or re-establish populations after winter in many southern and western areas of NSW that theoretically can be enclosed within brevitarsis lines. The patterns of occurrence of C. brevitarsis and of virus transmissions in NSW have been consistent with failure to survive winter away from the endemic mid-north coast. This is followed by passive and gradual, seasonal movements that probably occur regularly down the southern coastal plains and up the coastal valleys (Bishop et al., 1995b, 1996a). Our aim was to model the observed dispersal of C. brevitarsis to the south and west of its endemic area in NSW. Movement was implicit in our concept in contrast to the Murray and Nix's (1987) reliance on development from existing foci. The Hunter Valley was the most-likely route of dispersal to the west because it has fewer geographical barriers and a greater availability of hosts than any other coastal valley. 2. Materials and methods C. brevitarsis exhibits crepuscular activity and can be caught in light traps, mainly in the first 2 h after sunset. It was sampled with single light traps located at selected coastal sites of NSW from Casino in the north to Moruya in the south and extended to the western slopes (Mudgee) and northern tablelands (Tamworth) for varying periods from
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
245
Fig. 1. Locations of light traps for C. brevitarsis in NSW, Australia for varying periods from 1990 to 1999. Sites 9±19 (excluding site 16) are within the Hunter and related valley system. (~) Sites used to develop the models; (*) sites within the species endemic area; ( ) sites where C. brevitarsis was never or was rarely recorded. Coastal and inland regions are divided by a line representing the escarpment of the Great Dividing Range.
1990 to 1999. Fifteen more light traps were located at more-westerly and southern sites after 1995±1996 (Fig. 1; Bishop et al., 1995a, 1996a). The traps were similar to those described by Dyce et al. (1971). They were modified by Standfast (personal communication) to include photoelectric cells to automatically trigger operation at night. Each trap had a 3.2-V globe and a small fan driven by three D-cell batteries. Collections were made into bottles containing 70% alcohol (plus 10±20% glycerol to prevent loss due to evaporation). The traps were suspended from trees about 2 m above the ground in areas where cattle were consistently nearby. Catches were made approximately weekly (3 times per month with the week of the full moon excluded), fortnightly or monthly depending on trap location, monitoring requirements and year. Culicoides species were separated from other insects under a binocular microscope. C. brevitarsis was identified by its wing pattern and its numbers recorded. Temperature, rainfall and wind data were obtained for the monitoring sites from the Bureau of Meteorology at Sydney and Melbourne. Winds from two cardinal directions were required to disperse C. brevitarsis. Winds blowing from the north (NNW, N, NNE, NE) were necessary to move C. brevitarsis down the coast from the north. Easterly (ENE, E, ESE, SE) winds were needed to move C. brevitarsis into and up the Hunter and related valley system represented by sites 9±19 (except 16) in Fig. 1. These winds represented the prevailing winds in the period of movement by C. brevitarsis. Wind speed was classified as low (<8 km/h) or high (>8 km/h) for modelling purposes because 8 km/h is the approximate speed at which C. brevitarsis ceases local flight (Murray, 1987a). Wind frequency was represented by the number of occasions in the month of the journey that
246
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
wind was observed blowing from the relevant direction (E or N) at low or high speed at 09:00 and 15:00 h. Weather data were expressed as weekly or monthly means (temperature, 8C) and total weekly or monthly rainfall (millimetre) and wind. 2.1. Compilation of data We recognised that factors such as moon phase, light intensity, temperature, wind speed, rainfall and the presence/absence of cattle could affect activity and numbers of C. brevitarsis caught in light traps (Bishop et al., 2000; Murray, 1987a). These factors were controlled as far as practical by avoiding sampling at the full phase of the moon and not locating traps in the field when rain or high winds were expected. However, comparisons of density between sites at different times were regarded as potentially being too variable for most modelling purposes. Data were therefore compiled based on the time that C. brevitarsis first occurred at a site so that their presence depended less on density. Density was included where possible relationships with dispersal were observed during compilation of the first occurrence data. The Hastings Valley (represented by site 1 in Fig. 1) is the approximate southern limit of the C. brevitarsis-endemic area. Survival is occasionally detected in the lower Hunter Valley, but this is not considered to contribute to initial movements (Bishop et al., 1996a). Site 3, about 50 km south of site 1, was chosen as a starting point outside the endemic area from which movements to other locations could be detected first. The subsequent pattern of dispersal of C. brevitarsis from site 3 was then thought to depend on both geographical and weather-related factors. Sites 3, 9, 12, 18, 23 and 25 (Fig. 1) were chosen as pivotal points through which C. brevitarsis would need to travel to bypass geographical barriers (i.e. sections and features of the Great Dividing Range or large urban areas like the Sydney basin) and to reach the potential limits to its seasonal movements. The direct distance (kilometre) from site 3 via the relevant pivotal sites to each site was measured. Weather events during the month that C. brevitarsis journeyed between its occurrence at a pivotal point and first occurrence at each site were compiled. Weather data included average temperature, total rainfall and wind information at pivotal points. 2.2. Statistical modelling The dispersal time (in days, denoted by t) was considered to be the time between the date of the first occurrence of C. brevitarsis at site 3 and the date of first occurrence at a particular site in corresponding years. The objective of the analysis was to model the dispersal data (t) as a linear function of site distance from site 3 (D), and of weather variables (TE: temperature, RN: rain, Wl8: frequency of wind with speed less than 8 km/h, Wg8: frequency of wind with speed greater than 8 km/h). Locally weighted regression smoothing (LOESS) (MathSoft, 1998) determined the relationship between distance and dispersal time. The LOESS graph (Fig. 2) indicated that there was a change point at a distance approximately 100 km from site 3.It also made biological and geographical sense to split and model the area of observation sites into the three regions as suggested in Fig. 2. A logarithmic transformation of distance best linearized the LOESS relation
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
247
Fig. 2. Locally weighted regression smoothing analysis (LOESS; MathSoft, 1998) of the relationship between time and loge(distance) of the dispersal of C. brevitarsis in NSW, Australia from 1990 to 1999. The data arbitrarily were divided into three regions depending on the direction of movement from site 3.
within each region. Region 1 covers those sites in a radius of 100 km loge 100 4:6 from site 3, Region 2 covers those sites extending west from site 9, and Region 3 covers those sites extending south from site 9. The models were proposed as follows: Model 1:
t aj bj loge D;
j Region 1; 2; 3
Model 2 : t b0 b1 loge D b2 TE b3 RN b4 Wl8 b5 Wg8 where aj, bj, b0, b1, b2, b3, b4 and b5 are the coefficients of regression. The hypothesis of parallelism (bj b0 ; where b0 is an unknown common slope) was tested in the analysis of Model 1. Because the observations of C. brevitarsis were not made daily, the actual day (t) of first occurrence was unknown. The dispersal data were instead represented by an interval variable C which contained an upper time (U), the time when C. brevitarsis was first observed at a site, and a lower time (L) which was the observation time immediately preceding that first occurrence, i.e. C L; U. At any site in any year, there were three possible outcomes for t: (1) when C. brevitarsis appears between one observation and the next, L < t < U (interval censoring); (2) when C. brevitarsis fails to appear, t > L (right censoring); (3) when C. brevitarsis was present before it was found in site 3, t < U (left censoring). A generalised failuretime analysis (as described by Turnbull, 1974, 1976) was used to test the above models. A hazard function that underlies the unobservable distribution of t was assumed to follow a
248
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
Gaussian distribution. This assumption was validated by a Gaussian probability plot of the standardised Cox±Snell censored residuals as described in Meeker and Escobar (1998), i.e. h t f z < Z where z N 0; 1, being the standardised Normal distribution. The relationship between time and distance was intrinsically written in the z variable as follows: z
t ÿ aj bj loge D s
z
t ÿ b0 b1 loge D b2 TE b3 RN b4 Wl8 b5 Wg8 s
for Model 1 for Model 2
where s2 is the variance of the unobservable, underlying distribution of t. The estimation of all parameters followed the procedure described in MathSoft (1998). C. brevitarsis densities at sites 1, 3 and 9 recorded in the month prior to first occurrence at an observation site were used as a covariate and added to Model 1 to examine whether density at a previous site had any effect on the speed of the insects' dispersal. The dispersal data and site distance for the trap sites south as well as west of site 9 were offset by the corresponding variables measured at site 9. The relationship between the longest distance from site 3 reached each year by C. brevitarsis and its average density at site 1 in November (when movements are first recorded) was examined using a correlation analysis on loge(density) and loge(distance) adjusted by region effects. 3. Results The time when C. brevitarsis was observed at the observation sites was related to distance from site 3 (Table 1). This suggested that the insects must have moved out of the endemic area and penetrated regions to the south and west. The significant interaction showed that the speed of infestation varied between regions. The slopes of the hazard functions below and tests of parallelism (Table 1) showed that movement was fastest in Region 1, slower within Region 3 and slowest in Region 2. The hazard functions for the three regions are given as follows: z
t ÿ ÿ66:7 15:6 13:6 loge D 5:66
z
t ÿ ÿ818 175 26:7 loge D 6:17
z
t ÿ ÿ290 69 17:3 loge D 5:18
for Region 1 for Region 2 for Region 3
where the values within round brackets are the standard errors.
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
249
Table 1 Effects of region and loge(distance) (Model 1) and weather variables (Model 2) on the dispersal time of C. brevitarsis in New South Wales, Australia between 1990 and 1999 Sources of variation
Regression coefficient b
d.f.
Likelihood ratio, w2
Pr(w2)
S.E.
Model 1 Region loge(distance) (km) Interaction Residual
2 1 2 106
68.0 61.1 22.9 33.5
<0.0001 <0.0001 <0.0001 ±
Parallelism tests Ri Region i comparing speeds of infestation R1 vs. R2 1 R1 vs. R3 1 R2 vs. R3 1
27.3 5.6 9.4
<0.0001 0.02 0.002
99.03 126.12 1.13 2.09 4.08 39.69
<0.0001 <0.0001 0.29 0.15 0.04 ±
Model 2 loge(distance) (km) Temperature (8C) Rainfall (mm) Wind frequency (<8 km/h) Wind frequency (>8 km/h) Residual
79.6 ÿ6.9 ÿ0.1 1.4 1.6 ±
8.5 2.1 0.1 0.9 0.8 ±
1 1 1 1 1 76
Using the functions above to consider occurrence near the recorded limits of dispersal, we can predict with 70% probability that C. brevitarsis could reach site 27 (377 km south of site 3) in 138 days [95% confidence interval (CI) from 121 to 154 days] after first occurrence at site 3 (Fig. 3). For example, when the first occurrence at site 3 was in November, the insect would reach site 27 in April the following year and could reach this site every year. Using the same probability, first occurrence at site 16 (314 km west of site 3) is predicted in 229 days (95% CI from 196 to 261 days) which is in mid-June. Based on the model, it was predicted that C. brevitarsis could only reach site 16 once in 10 years. The time of first occurrence was also related significantly to temperature and wind frequencies (Table 1). The relational equation in z transformed variable of time on weather data was obtained as follows: t ÿ ÿ190 79:6 8:52 loge D ÿ 6:93 2:05 TE ÿ 0:08 0:08 RN 1:43 0:91Wl8 1:63 0:81Wg8 z 6:3 Higher temperature hastened the first occurrence and higher frequencies of wind above 8 km/h slowed down the movement. When C. brevitarsis density at the pivotal points was included in Model 1, there was no obvious effect on the speed that the insect spread. However, the density recorded in November at site 1 appeared to have a significant influence on how far the insects travelled to the south or west of site 3 r 0:56; P < 0:05.
250
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
Fig. 3. Predicted times (in days) of first occurrence (70% probability with 95% CI) of C. brevitarsis in three regions [site 3 to site 9 (Region 1), site 9 to site 16 (Region 2) and site 9 to site 28 (Region 3)] dependent on the distance from site 3 in NSW, Australia from 1990 to 1999.
4. Discussion Infection of livestock with arboviruses such as bluetongue and Akabane are dependent on the presence or absence of C. brevitarsis. Other Culicoides species may also be involved across northern Australia, and Culicoides wadai Kitaoka may periodically enter northern coastal NSW (Muller, 1993; Bishop, unpublished data). Virus-free regions have been designated in all states/territories of Australia where monitoring has shown that the vector and viruses are usually absent. Livestock may be exported from these regions. The regions are currently static but could possibly be extended by defining and including areas that might be free seasonally. Some of these areas could fall within brevitarsis lines. However, the vectors do not exist because they are rarely present or they fail to survive because of low temperatures in the south and probably high temperatures and/or a lack of moisture at the northern and western limits. Temperature is the limiting factor on the coastal plains of NSW because rainfall is usually adequate. This was demonstrated by rainfall not being a significant factor in our weather model (Table 1). Failure to survive winter must be accompanied by infrequent invasions (western NSW) or seasonal reinvasions if the viruses are to be spread. It is logical to expect that delays expressed by
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
251
the model could be incurred relative to the distance the insects must travel from their source. Model 1 is the preferred functional model because it negates the need for unknown temperature and wind data. The model is based on distance from a starting point just outside the endemic area (in this case site 3). This model can be linked to other developed and partially developed information to make it more practical and improve its efficiency. For example: Before C. brevitarsis activity starts, survival probability can be assumed from historic temperature data (Bishop et al., 1995a) or calculated after winter each season. Probability that C. brevitarsis can survive increases markedly with only small (28C) rises in average winter temperatures at some southern coastal sites. There was an apparent relationship between average density in the endemic area in the month (November) that C. brevitarsis starts to disperse and the maximum distance that C. brevitarsis can move in a given year. At present, the reliability of the relationship is possibly confounded by variations when assessing C. brevitarsis density, limited time of sampling (7 years) and a limited number of sites at the outer limits of spread. If this relationship remains consistent for several more seasons and more sites are added, it could assist predictions by defining yearly limits to dispersal before movements occur. Links between the vector and the viruses are not clear. However, it is important to understand that presence of the vector does not imply that there will also be viral activity (P. Kirkland, EMAI Camden, personal communication). Limited data suggest that infections by viruses could be delayed for months after first occurrence of the vector (Bishop et al., 1995b, 1996a). While it may be some time before the mechanisms relating virus survival through to its subsequent transmission to livestock in the field are fully understood, it is clear that the viruses should also be subject to periods of seasonal freedom. These periods are of greater practical consequence and may be even longer than those of the vector. The month that C. brevitarsis should last occur at sites throughout NSW at the end of the season has been determined (Bishop et al., 1995a). This knowledge is necessary to override statistical predictions that are beyond the known biological limits of the species. It can be seen from the predictions for site 16 that although occurrence is predicted at 229 days (95% CI from 196 to 261 days) this would be in mid-June and be impossible because activity would usually cease naturally in April (because of declining temperatures). The absence of arboviruses at site 16 except in exceptional circumstances (Murray, 1987b) reflects the rare occurrence of C. brevitarsis. Model 2 indicates that viruses spread as the result of vectors moving under suitable weather conditions. This helps our understanding of C. brevitarsis ecology. Biologically, temperatures between 20 and 308C would maximise survival, development and behavioural activity (particularly, behaviour involving local flight which would place C. brevitarsis in a position to be moved passively by the wind) (Allingham, 1991; Bishop et al., 1996b; Murray, 1987a). Wind direction is obviously important for passive movements in any direction. Wind speed was also important with high frequencies of
252
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
wind above the 8 km/h threshold for local flight slowing the spread of C. brevitarsis. This was consistent with the observation that movement is passive and occurs gradually and probably regularly during the season. Rapid or long-distance movement would therefore be uncommon and depend on chance meteorological events Ð as described by Murray (1987b) and Murray and Kirkland (1995). Dispersal was fastest on the relatively flat coastal plain adjacent to the endemic area. This supports the observation that movements could be dependent on geographical features (plus events in, and distance from) of the endemic area. Slower movements in the region south of site 9 towards site 27 Ð while still on the coastal plain Ð could be related to the need for C. brevitarsis to traverse or bypass large urban areas (e.g. Sydney and Wollongong; Fig. 1). Even, slower movements in the region west of site 9 towards site 16 could be related to greater geographic complexity (e.g. hills, ranges and increasing altitude acting as barriers). Establishment of the vector and its ability to transmit viruses would be affected by weather conditions on arrival at each destination and the presence or absence of virus at the source. Both models could be applied to other situations and locations given an appropriate data set. They could apply in endemic areas where similar movements probably occur but are masked by C. brevitarsis that are already established at the site of destination. This movement would explain how viruses are dispersed from initial foci in the endemic area (P. Kirkland, EMAI Camden, personal communication). It is also possible that C. brevitarsis could invade northern and western NSW from Queensland (Murray and Nix, 1987) or move up other coastal valleys in north-eastern NSW. Applicability of the models could therefore require modifications to wind directions (possibly), greater emphasis on the effects of moisture, rainfall and pasture conditions in the drier west and the inclusion of obstructions by geographical features (such as the eastern escarpment of the Great Dividing Range acting as a barrier). Neither model precludes the possible importance and relevance of the Murray and Nix (1987) concept in regions where C. brevitarsis can survive over winter. Foci easily can develop in endemic areas where favourable sites can be limited by temperature or host and habitat availability. However, it appeared that the Murray and Nix (1987) model overestimates the lower-temperature survival potential of C. brevitarsis. 5. Conclusions The models for the dispersal of C. brevitarsis complete three key components considered necessary to understand and describe activity by C. brevitarsis at the southern limits of its distribution. These components define survival (Bishop et al., 1995a), dispersal (this paper) and the length of time C. brevitarsis can remain active outside of its endemic area (Bishop et al., 1995a). It is therefore possible to predict times when livestock are at risk of infection and times when areas are free of vectors throughout NSW. Seasonal freedom could apply to some currently restricted areas. Monitoring should be maintained to detect abnormal events and possible changes due to predicted world rises in temperature and to improve the accuracy of the models. Predictions of C. brevitarsis activity could be enhanced further by data clarifying questions relating to vector density, virus aetiology and other routes of dispersal.
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
253
Acknowledgements Mr. H. McKenzie is thanked for his technical expertise in sorting and identifying lighttrap catches. We thank Drs. P.D. Kirkland and B. Cullis for their comments during the preparation of this paper. The assistance and cooperation of the many light-trap operators was also appreciated. Funding was received from the Meat Research Council and the National Arbovirus Monitoring Program. References Allingham, P.G., 1991. Effect of temperature on late immature stages of Culicoides brevitarsis (Diptera: Ceratopogonidae). J. Med. Entomol. 28, 878±881. Bishop, A.L., Barchia, I.M., Harris, A.M., 1995a. Last occurrence and survival during winter of the arbovirus vector Culicoides brevitarsis at the southern limits of its distribution. Aust. Vet. J. 72, 53±55. Bishop, A.L., Kirkland, P.D., McKenzie, H.J., Spohr, L.J., Barchia, I.M., Muller, M.J., 1995b. Distribution and seasonal movements of Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae) at the southern limits and its distribution in New South Wales and their correlation with arboviruses affecting livestock. J. Aust. Ent. Soc. 34, 289±298. Bishop, A.L., Kirkland, P.D., McKenzie, H.J., Barchia, I.M., 1996a. The dispersal of Culicoides brevitarsis in eastern New South Wales and associations with the occurrences of arbovirus infections in cattle. Aust. Vet. J. 73, 174±178. Bishop, A.L., McKenzie, H.J., Barchia, I.M., Harris, A.M., 1996b. Effect of temperature regimes on the development, survival and emergence of Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae) in bovine dung. Aust. J. Entomol. 35, 361±368. Bishop, A.L., McKenzie, H.J., Barchia, I.M., Spohr, L.J., 2000. Moon phase and other factors affecting light-trap catches of Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae). Aust. J. Entomol. 39, 29±32. Braverman, Y., 1992. The possible introduction to Israel of Culicoides (Diptera, Ceratopogonidae) borne animal diseases by wind. In: Walton, T.E., Osburn, B.I. (Eds.), Bluetongue, African Horse Sickness, and Related Orbiviruses, Proceedings of the Second International Symposium.CRC Press Boca Raton, FL. Dillon, M.L., Fitt, G.P., Hamilton, J.G., Rochester, W.A., Henderson-Sellers, B., McAleer, M., Jakeman, A.J., 1996. A simulation model of wind-driven dispersal of Helicoverpa moths. Ecol. J. Model. 86, 145±150. Doherty, R.L., Standfast, H.A., Dyce, A.L., Carley, J.G., Gorman, B.M., Filippich, C., Kay, B.H., 1978. Mudjinbarry virus, an orbivirus related to Wallal virus isolated from midges from the Northern Territory of Australia. Aust. J. Biol. Sci. 31, 97±103. Dyce, A.L., Standfast, H.A., Kay, B.H., 1971. Collection and preparation of biting midges (Fam. Ceratopogonidae) and other small Diptera for virus isolation. J. Aust. Ent. Soc. 11, 91±96. Fraval, A., El-Yousfi, M., 1989. Active and passive dispersal of Lymantria dispar (L.) (Lep., Lymantriidae) in cork-oak stands on the Atlantic coast of Morocco. J. Appl. Entomol. 108, 335±346 (English summary). Loxdale, H.D., Hadie, J., Halbert, S., Foottit, R., Kidd, N.A.C., Carter, C.I., 1993. The relative importance of short- and long-range movement of flying aphids. Biol. Rev. Camb. Phil. Soc. 68, 291±311. MathSoft, 1998. S-PLUS 5 Guide to Statistics. MathSoft, Inc., Seattle, WA, 1014 pp. Meeker, W.M., Escobar, L.A., 1998. Statistical Methods for Reliability Data. Wiley, New York, 680 pp. Muller, M.J., 1993. Distribution of vectors of arboviruses affecting sheep in Australia. Final Report, WRDC Project CTI3. CSIRO, Long Pocket, Brisbane. Muller, M.J., Standfast, H.A., St. George, T.D., Cybinski, D.H., 1982. Culicoides brevitarsis (Diptera: Ceratopogonidae) as a vector of arboviruses in Australia. In: St. George, T.D., Kay, B.H. (Eds.), Proceedings of the Third Symposium on Arbovirus Research in Australia, Brisbane, Qld, pp. 43±49. Murray, M.D., 1987a. Local dispersal of the biting midge, Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae) in south-eastern Australia. Aust. J. Zool. 35, 559±573. Murray, M.D., 1987b. Akabane epizootics in New South Wales: evidence for long-distance dispersal of the biting midge Culicoides brevitarsis. Aust. Vet. J. 64, 305±308.
254
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
Murray, M.D., Kirkland, P.D., 1995. Bluetongue and Douglas virus activity in New South Wales in 1989: further evidence for long-distance dispersal of the biting midge Culicoides brevitarsis. Aust. Vet. J. 72, 56±57. Murray, M.D., Nix, H.A., 1987. Southern limits of distribution and abundance of the biting-midge Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae) in south-eastern Australia: an application of the GROWEST model. Aust. J. Zool. 35, 575±585. Nix, H.A., 1976. Environmental control of breeding, post-breeding dispersal and migration of birds in the Australian region. In: Frith, H.J., Calaby, J.H. (Eds.), Proceedings of the 16th International Ornithological Congress, Canberra, pp. 272±305. Sellers, R.F., Pedgley, D.E., Tucker, M.R., 1977. Possible spread of African horse sickness on the wind. J. Hyg. 79, 279±298. Sellers, R.F., Walton, T.E., Osburn, B.I., 1992. Weather, Culicoides, and the distribution and spread of bluetongue and African horse sickness viruses. In: Walton, T.E., Osburn, B.I. (Eds.), Bluetongue, African Horse Sickness, and Related Orbiviruses, Proceedings of the Second International Symposium. CRC Press, Boca Raton, FL, pp. 284±290. Standfast, H.A., Dyce, A.L., St. George, T.D., Muller, M.J., Doherty, R.L., Carley, J.G., Filippich, C., 1984. Isolation of arboviruses from insects collected at Beatrice Hill, Northern Territory of Australia, 1974±1976. Aust. J. Biol. Sci. 37, 351±366. Turnbull, B.W., 1974. Nonparametric estimation of a survivorship function with doubly censored data. J. Am. Stat. Assoc. 69, 169±173. Turnbull, B.W., 1976. The empirical distribution function with arbitrarily grouped, censored and truncated data. J. R. Stat. Soc. Ser. B 38, 290±295.
Models for the dispersal in Australia of the arbovirus vector, Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae) Alan L. Bishop*, Idris M. Barchia, Lorraine J. Spohr NSW Agriculture, Locked Bag 26, Gosford, NSW 2250, Australia Received 24 March 2000; accepted 27 August 2000
Abstract Culicoides brevitarsis is the main biting midge responsible for the transmission of bluetongue and Akabane viruses to livestock in Australia. Models are given for its dispersal after winter from endemic areas at the southern limit of its distribution in New South Wales (NSW); the models might also be applicable elsewhere. Model 1 shows that dispersal can be explained by distance from a key point just outside the endemic area in mid-northern/northern coastal NSW. The model provides probability data for times of first occurrence at sites within regions down the southern coastal plain or up the Hunter Valley towards (but rarely reaching) the western slopes and tablelands. Model 2 shows that the movement depends on temperature and wind speed from northerly and easterly directions. Preliminary data also are given to suggest a relationship between density in the endemic area and the maximum distance that C. brevitarsis can travel in a given year. The models can be linked to other information which in combination can provide probabilities for winter survival outside the endemic area, times of occurrence at sites where it cannot survive winter and times when activity ceases naturally at these sites at the end of the season. This information can be used to predict the potential for virus transmission and indicate zones of seasonal freedom from both vector and virus for the export of livestock. # 2000 Elsevier Science B.V. All rights reserved. Keywords: Culicoides brevitarsis; Dispersal; Vector; Arboviruses; Australia
1. Introduction Insects are endemic where temperature, moisture, habitat and feeding conditions are consistently suitable for their survival, growth and reproduction. Distributions beyond endemic areas can sometimes be accidental but often occur as a result of the insects' *
Corresponding author. Tel.: 61-2-4348-1928; fax: 61-2-4348-1910. E-mail address: [email protected] (A.L. Bishop). 0167-5877/00/$ ± see front matter # 2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 7 - 5 8 7 7 ( 0 0 ) 0 0 1 7 5 - 6
244
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
passive or active movements into areas that are favourable seasonally. Movement can be associated with wind and weather activity or be aided by mechanical means of transport (Fraval and El-Yousfi, 1989; Loxdale et al., 1993; Dillon et al., 1996). Several species of biting midge from the genus Culicoides (Diptera: Ceratopogonidae) transmit viruses to livestock and native fauna (Doherty et al., 1978; Muller et al., 1982; Standfast et al., 1984). Epidemiological studies suggest that the spread of many viruses may be related to wind-borne dispersal of the midge vectors (Sellers et al., 1977, 1992; Braverman, 1992). In Australia, Culicoides brevitarsis Kieffer is the main vector of the bluetongue and Akabane viruses that affect the health and export of livestock (Muller et al., 1982). Local dispersal of C. brevitarsis in the wind is an integral part of its biology. Flight, orientation and relationships with animals depend on temperature and the speed and direction of wind (Murray, 1987a). Retrospective evidence supports the occasional, long-distance and wind-borne dispersal of infected C. brevitarsis (Murray, 1987b; Murray and Kirkland, 1995). C. brevitarsis is endemic on the mid-northern/northern coastal plains of New South Wales (NSW). It disperses seasonally and normally retains a coastal distribution (Bishop et al., 1996a) Ð although occasionally it has been found inland (Muller et al., 1982). The GROWEST model of Nix (1976) was modified by Murray and Nix (1987) to explain and predict these occurrences. This model proposes areas suitable for summer multiplication and winter survival (based on temperature and rainfall) that can be enclosed within `brevitarsis lines'. Implicit in the model's concept is the existence of over-wintering foci from which both vector and virus develop. While biologically possible at some locations (Bishop et al., 1995a), development and movement from over-wintering foci have never been observed outside the endemic area in NSW between 1990 and 1999 (Bishop et al., 1995b, 1996a, unpublished data). This can be explained by low probabilities for survival based on temperatures and the need for C. brevitarsis to establish or re-establish populations after winter in many southern and western areas of NSW that theoretically can be enclosed within brevitarsis lines. The patterns of occurrence of C. brevitarsis and of virus transmissions in NSW have been consistent with failure to survive winter away from the endemic mid-north coast. This is followed by passive and gradual, seasonal movements that probably occur regularly down the southern coastal plains and up the coastal valleys (Bishop et al., 1995b, 1996a). Our aim was to model the observed dispersal of C. brevitarsis to the south and west of its endemic area in NSW. Movement was implicit in our concept in contrast to the Murray and Nix's (1987) reliance on development from existing foci. The Hunter Valley was the most-likely route of dispersal to the west because it has fewer geographical barriers and a greater availability of hosts than any other coastal valley. 2. Materials and methods C. brevitarsis exhibits crepuscular activity and can be caught in light traps, mainly in the first 2 h after sunset. It was sampled with single light traps located at selected coastal sites of NSW from Casino in the north to Moruya in the south and extended to the western slopes (Mudgee) and northern tablelands (Tamworth) for varying periods from
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
245
Fig. 1. Locations of light traps for C. brevitarsis in NSW, Australia for varying periods from 1990 to 1999. Sites 9±19 (excluding site 16) are within the Hunter and related valley system. (~) Sites used to develop the models; (*) sites within the species endemic area; ( ) sites where C. brevitarsis was never or was rarely recorded. Coastal and inland regions are divided by a line representing the escarpment of the Great Dividing Range.
1990 to 1999. Fifteen more light traps were located at more-westerly and southern sites after 1995±1996 (Fig. 1; Bishop et al., 1995a, 1996a). The traps were similar to those described by Dyce et al. (1971). They were modified by Standfast (personal communication) to include photoelectric cells to automatically trigger operation at night. Each trap had a 3.2-V globe and a small fan driven by three D-cell batteries. Collections were made into bottles containing 70% alcohol (plus 10±20% glycerol to prevent loss due to evaporation). The traps were suspended from trees about 2 m above the ground in areas where cattle were consistently nearby. Catches were made approximately weekly (3 times per month with the week of the full moon excluded), fortnightly or monthly depending on trap location, monitoring requirements and year. Culicoides species were separated from other insects under a binocular microscope. C. brevitarsis was identified by its wing pattern and its numbers recorded. Temperature, rainfall and wind data were obtained for the monitoring sites from the Bureau of Meteorology at Sydney and Melbourne. Winds from two cardinal directions were required to disperse C. brevitarsis. Winds blowing from the north (NNW, N, NNE, NE) were necessary to move C. brevitarsis down the coast from the north. Easterly (ENE, E, ESE, SE) winds were needed to move C. brevitarsis into and up the Hunter and related valley system represented by sites 9±19 (except 16) in Fig. 1. These winds represented the prevailing winds in the period of movement by C. brevitarsis. Wind speed was classified as low (<8 km/h) or high (>8 km/h) for modelling purposes because 8 km/h is the approximate speed at which C. brevitarsis ceases local flight (Murray, 1987a). Wind frequency was represented by the number of occasions in the month of the journey that
246
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
wind was observed blowing from the relevant direction (E or N) at low or high speed at 09:00 and 15:00 h. Weather data were expressed as weekly or monthly means (temperature, 8C) and total weekly or monthly rainfall (millimetre) and wind. 2.1. Compilation of data We recognised that factors such as moon phase, light intensity, temperature, wind speed, rainfall and the presence/absence of cattle could affect activity and numbers of C. brevitarsis caught in light traps (Bishop et al., 2000; Murray, 1987a). These factors were controlled as far as practical by avoiding sampling at the full phase of the moon and not locating traps in the field when rain or high winds were expected. However, comparisons of density between sites at different times were regarded as potentially being too variable for most modelling purposes. Data were therefore compiled based on the time that C. brevitarsis first occurred at a site so that their presence depended less on density. Density was included where possible relationships with dispersal were observed during compilation of the first occurrence data. The Hastings Valley (represented by site 1 in Fig. 1) is the approximate southern limit of the C. brevitarsis-endemic area. Survival is occasionally detected in the lower Hunter Valley, but this is not considered to contribute to initial movements (Bishop et al., 1996a). Site 3, about 50 km south of site 1, was chosen as a starting point outside the endemic area from which movements to other locations could be detected first. The subsequent pattern of dispersal of C. brevitarsis from site 3 was then thought to depend on both geographical and weather-related factors. Sites 3, 9, 12, 18, 23 and 25 (Fig. 1) were chosen as pivotal points through which C. brevitarsis would need to travel to bypass geographical barriers (i.e. sections and features of the Great Dividing Range or large urban areas like the Sydney basin) and to reach the potential limits to its seasonal movements. The direct distance (kilometre) from site 3 via the relevant pivotal sites to each site was measured. Weather events during the month that C. brevitarsis journeyed between its occurrence at a pivotal point and first occurrence at each site were compiled. Weather data included average temperature, total rainfall and wind information at pivotal points. 2.2. Statistical modelling The dispersal time (in days, denoted by t) was considered to be the time between the date of the first occurrence of C. brevitarsis at site 3 and the date of first occurrence at a particular site in corresponding years. The objective of the analysis was to model the dispersal data (t) as a linear function of site distance from site 3 (D), and of weather variables (TE: temperature, RN: rain, Wl8: frequency of wind with speed less than 8 km/h, Wg8: frequency of wind with speed greater than 8 km/h). Locally weighted regression smoothing (LOESS) (MathSoft, 1998) determined the relationship between distance and dispersal time. The LOESS graph (Fig. 2) indicated that there was a change point at a distance approximately 100 km from site 3.It also made biological and geographical sense to split and model the area of observation sites into the three regions as suggested in Fig. 2. A logarithmic transformation of distance best linearized the LOESS relation
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
247
Fig. 2. Locally weighted regression smoothing analysis (LOESS; MathSoft, 1998) of the relationship between time and loge(distance) of the dispersal of C. brevitarsis in NSW, Australia from 1990 to 1999. The data arbitrarily were divided into three regions depending on the direction of movement from site 3.
within each region. Region 1 covers those sites in a radius of 100 km loge 100 4:6 from site 3, Region 2 covers those sites extending west from site 9, and Region 3 covers those sites extending south from site 9. The models were proposed as follows: Model 1:
t aj bj loge D;
j Region 1; 2; 3
Model 2 : t b0 b1 loge D b2 TE b3 RN b4 Wl8 b5 Wg8 where aj, bj, b0, b1, b2, b3, b4 and b5 are the coefficients of regression. The hypothesis of parallelism (bj b0 ; where b0 is an unknown common slope) was tested in the analysis of Model 1. Because the observations of C. brevitarsis were not made daily, the actual day (t) of first occurrence was unknown. The dispersal data were instead represented by an interval variable C which contained an upper time (U), the time when C. brevitarsis was first observed at a site, and a lower time (L) which was the observation time immediately preceding that first occurrence, i.e. C L; U. At any site in any year, there were three possible outcomes for t: (1) when C. brevitarsis appears between one observation and the next, L < t < U (interval censoring); (2) when C. brevitarsis fails to appear, t > L (right censoring); (3) when C. brevitarsis was present before it was found in site 3, t < U (left censoring). A generalised failuretime analysis (as described by Turnbull, 1974, 1976) was used to test the above models. A hazard function that underlies the unobservable distribution of t was assumed to follow a
248
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
Gaussian distribution. This assumption was validated by a Gaussian probability plot of the standardised Cox±Snell censored residuals as described in Meeker and Escobar (1998), i.e. h t f z < Z where z N 0; 1, being the standardised Normal distribution. The relationship between time and distance was intrinsically written in the z variable as follows: z
t ÿ aj bj loge D s
z
t ÿ b0 b1 loge D b2 TE b3 RN b4 Wl8 b5 Wg8 s
for Model 1 for Model 2
where s2 is the variance of the unobservable, underlying distribution of t. The estimation of all parameters followed the procedure described in MathSoft (1998). C. brevitarsis densities at sites 1, 3 and 9 recorded in the month prior to first occurrence at an observation site were used as a covariate and added to Model 1 to examine whether density at a previous site had any effect on the speed of the insects' dispersal. The dispersal data and site distance for the trap sites south as well as west of site 9 were offset by the corresponding variables measured at site 9. The relationship between the longest distance from site 3 reached each year by C. brevitarsis and its average density at site 1 in November (when movements are first recorded) was examined using a correlation analysis on loge(density) and loge(distance) adjusted by region effects. 3. Results The time when C. brevitarsis was observed at the observation sites was related to distance from site 3 (Table 1). This suggested that the insects must have moved out of the endemic area and penetrated regions to the south and west. The significant interaction showed that the speed of infestation varied between regions. The slopes of the hazard functions below and tests of parallelism (Table 1) showed that movement was fastest in Region 1, slower within Region 3 and slowest in Region 2. The hazard functions for the three regions are given as follows: z
t ÿ ÿ66:7 15:6 13:6 loge D 5:66
z
t ÿ ÿ818 175 26:7 loge D 6:17
z
t ÿ ÿ290 69 17:3 loge D 5:18
for Region 1 for Region 2 for Region 3
where the values within round brackets are the standard errors.
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
249
Table 1 Effects of region and loge(distance) (Model 1) and weather variables (Model 2) on the dispersal time of C. brevitarsis in New South Wales, Australia between 1990 and 1999 Sources of variation
Regression coefficient b
d.f.
Likelihood ratio, w2
Pr(w2)
S.E.
Model 1 Region loge(distance) (km) Interaction Residual
2 1 2 106
68.0 61.1 22.9 33.5
<0.0001 <0.0001 <0.0001 ±
Parallelism tests Ri Region i comparing speeds of infestation R1 vs. R2 1 R1 vs. R3 1 R2 vs. R3 1
27.3 5.6 9.4
<0.0001 0.02 0.002
99.03 126.12 1.13 2.09 4.08 39.69
<0.0001 <0.0001 0.29 0.15 0.04 ±
Model 2 loge(distance) (km) Temperature (8C) Rainfall (mm) Wind frequency (<8 km/h) Wind frequency (>8 km/h) Residual
79.6 ÿ6.9 ÿ0.1 1.4 1.6 ±
8.5 2.1 0.1 0.9 0.8 ±
1 1 1 1 1 76
Using the functions above to consider occurrence near the recorded limits of dispersal, we can predict with 70% probability that C. brevitarsis could reach site 27 (377 km south of site 3) in 138 days [95% confidence interval (CI) from 121 to 154 days] after first occurrence at site 3 (Fig. 3). For example, when the first occurrence at site 3 was in November, the insect would reach site 27 in April the following year and could reach this site every year. Using the same probability, first occurrence at site 16 (314 km west of site 3) is predicted in 229 days (95% CI from 196 to 261 days) which is in mid-June. Based on the model, it was predicted that C. brevitarsis could only reach site 16 once in 10 years. The time of first occurrence was also related significantly to temperature and wind frequencies (Table 1). The relational equation in z transformed variable of time on weather data was obtained as follows: t ÿ ÿ190 79:6 8:52 loge D ÿ 6:93 2:05 TE ÿ 0:08 0:08 RN 1:43 0:91Wl8 1:63 0:81Wg8 z 6:3 Higher temperature hastened the first occurrence and higher frequencies of wind above 8 km/h slowed down the movement. When C. brevitarsis density at the pivotal points was included in Model 1, there was no obvious effect on the speed that the insect spread. However, the density recorded in November at site 1 appeared to have a significant influence on how far the insects travelled to the south or west of site 3 r 0:56; P < 0:05.
250
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
Fig. 3. Predicted times (in days) of first occurrence (70% probability with 95% CI) of C. brevitarsis in three regions [site 3 to site 9 (Region 1), site 9 to site 16 (Region 2) and site 9 to site 28 (Region 3)] dependent on the distance from site 3 in NSW, Australia from 1990 to 1999.
4. Discussion Infection of livestock with arboviruses such as bluetongue and Akabane are dependent on the presence or absence of C. brevitarsis. Other Culicoides species may also be involved across northern Australia, and Culicoides wadai Kitaoka may periodically enter northern coastal NSW (Muller, 1993; Bishop, unpublished data). Virus-free regions have been designated in all states/territories of Australia where monitoring has shown that the vector and viruses are usually absent. Livestock may be exported from these regions. The regions are currently static but could possibly be extended by defining and including areas that might be free seasonally. Some of these areas could fall within brevitarsis lines. However, the vectors do not exist because they are rarely present or they fail to survive because of low temperatures in the south and probably high temperatures and/or a lack of moisture at the northern and western limits. Temperature is the limiting factor on the coastal plains of NSW because rainfall is usually adequate. This was demonstrated by rainfall not being a significant factor in our weather model (Table 1). Failure to survive winter must be accompanied by infrequent invasions (western NSW) or seasonal reinvasions if the viruses are to be spread. It is logical to expect that delays expressed by
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
251
the model could be incurred relative to the distance the insects must travel from their source. Model 1 is the preferred functional model because it negates the need for unknown temperature and wind data. The model is based on distance from a starting point just outside the endemic area (in this case site 3). This model can be linked to other developed and partially developed information to make it more practical and improve its efficiency. For example: Before C. brevitarsis activity starts, survival probability can be assumed from historic temperature data (Bishop et al., 1995a) or calculated after winter each season. Probability that C. brevitarsis can survive increases markedly with only small (28C) rises in average winter temperatures at some southern coastal sites. There was an apparent relationship between average density in the endemic area in the month (November) that C. brevitarsis starts to disperse and the maximum distance that C. brevitarsis can move in a given year. At present, the reliability of the relationship is possibly confounded by variations when assessing C. brevitarsis density, limited time of sampling (7 years) and a limited number of sites at the outer limits of spread. If this relationship remains consistent for several more seasons and more sites are added, it could assist predictions by defining yearly limits to dispersal before movements occur. Links between the vector and the viruses are not clear. However, it is important to understand that presence of the vector does not imply that there will also be viral activity (P. Kirkland, EMAI Camden, personal communication). Limited data suggest that infections by viruses could be delayed for months after first occurrence of the vector (Bishop et al., 1995b, 1996a). While it may be some time before the mechanisms relating virus survival through to its subsequent transmission to livestock in the field are fully understood, it is clear that the viruses should also be subject to periods of seasonal freedom. These periods are of greater practical consequence and may be even longer than those of the vector. The month that C. brevitarsis should last occur at sites throughout NSW at the end of the season has been determined (Bishop et al., 1995a). This knowledge is necessary to override statistical predictions that are beyond the known biological limits of the species. It can be seen from the predictions for site 16 that although occurrence is predicted at 229 days (95% CI from 196 to 261 days) this would be in mid-June and be impossible because activity would usually cease naturally in April (because of declining temperatures). The absence of arboviruses at site 16 except in exceptional circumstances (Murray, 1987b) reflects the rare occurrence of C. brevitarsis. Model 2 indicates that viruses spread as the result of vectors moving under suitable weather conditions. This helps our understanding of C. brevitarsis ecology. Biologically, temperatures between 20 and 308C would maximise survival, development and behavioural activity (particularly, behaviour involving local flight which would place C. brevitarsis in a position to be moved passively by the wind) (Allingham, 1991; Bishop et al., 1996b; Murray, 1987a). Wind direction is obviously important for passive movements in any direction. Wind speed was also important with high frequencies of
252
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
wind above the 8 km/h threshold for local flight slowing the spread of C. brevitarsis. This was consistent with the observation that movement is passive and occurs gradually and probably regularly during the season. Rapid or long-distance movement would therefore be uncommon and depend on chance meteorological events Ð as described by Murray (1987b) and Murray and Kirkland (1995). Dispersal was fastest on the relatively flat coastal plain adjacent to the endemic area. This supports the observation that movements could be dependent on geographical features (plus events in, and distance from) of the endemic area. Slower movements in the region south of site 9 towards site 27 Ð while still on the coastal plain Ð could be related to the need for C. brevitarsis to traverse or bypass large urban areas (e.g. Sydney and Wollongong; Fig. 1). Even, slower movements in the region west of site 9 towards site 16 could be related to greater geographic complexity (e.g. hills, ranges and increasing altitude acting as barriers). Establishment of the vector and its ability to transmit viruses would be affected by weather conditions on arrival at each destination and the presence or absence of virus at the source. Both models could be applied to other situations and locations given an appropriate data set. They could apply in endemic areas where similar movements probably occur but are masked by C. brevitarsis that are already established at the site of destination. This movement would explain how viruses are dispersed from initial foci in the endemic area (P. Kirkland, EMAI Camden, personal communication). It is also possible that C. brevitarsis could invade northern and western NSW from Queensland (Murray and Nix, 1987) or move up other coastal valleys in north-eastern NSW. Applicability of the models could therefore require modifications to wind directions (possibly), greater emphasis on the effects of moisture, rainfall and pasture conditions in the drier west and the inclusion of obstructions by geographical features (such as the eastern escarpment of the Great Dividing Range acting as a barrier). Neither model precludes the possible importance and relevance of the Murray and Nix (1987) concept in regions where C. brevitarsis can survive over winter. Foci easily can develop in endemic areas where favourable sites can be limited by temperature or host and habitat availability. However, it appeared that the Murray and Nix (1987) model overestimates the lower-temperature survival potential of C. brevitarsis. 5. Conclusions The models for the dispersal of C. brevitarsis complete three key components considered necessary to understand and describe activity by C. brevitarsis at the southern limits of its distribution. These components define survival (Bishop et al., 1995a), dispersal (this paper) and the length of time C. brevitarsis can remain active outside of its endemic area (Bishop et al., 1995a). It is therefore possible to predict times when livestock are at risk of infection and times when areas are free of vectors throughout NSW. Seasonal freedom could apply to some currently restricted areas. Monitoring should be maintained to detect abnormal events and possible changes due to predicted world rises in temperature and to improve the accuracy of the models. Predictions of C. brevitarsis activity could be enhanced further by data clarifying questions relating to vector density, virus aetiology and other routes of dispersal.
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
253
Acknowledgements Mr. H. McKenzie is thanked for his technical expertise in sorting and identifying lighttrap catches. We thank Drs. P.D. Kirkland and B. Cullis for their comments during the preparation of this paper. The assistance and cooperation of the many light-trap operators was also appreciated. Funding was received from the Meat Research Council and the National Arbovirus Monitoring Program. References Allingham, P.G., 1991. Effect of temperature on late immature stages of Culicoides brevitarsis (Diptera: Ceratopogonidae). J. Med. Entomol. 28, 878±881. Bishop, A.L., Barchia, I.M., Harris, A.M., 1995a. Last occurrence and survival during winter of the arbovirus vector Culicoides brevitarsis at the southern limits of its distribution. Aust. Vet. J. 72, 53±55. Bishop, A.L., Kirkland, P.D., McKenzie, H.J., Spohr, L.J., Barchia, I.M., Muller, M.J., 1995b. Distribution and seasonal movements of Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae) at the southern limits and its distribution in New South Wales and their correlation with arboviruses affecting livestock. J. Aust. Ent. Soc. 34, 289±298. Bishop, A.L., Kirkland, P.D., McKenzie, H.J., Barchia, I.M., 1996a. The dispersal of Culicoides brevitarsis in eastern New South Wales and associations with the occurrences of arbovirus infections in cattle. Aust. Vet. J. 73, 174±178. Bishop, A.L., McKenzie, H.J., Barchia, I.M., Harris, A.M., 1996b. Effect of temperature regimes on the development, survival and emergence of Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae) in bovine dung. Aust. J. Entomol. 35, 361±368. Bishop, A.L., McKenzie, H.J., Barchia, I.M., Spohr, L.J., 2000. Moon phase and other factors affecting light-trap catches of Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae). Aust. J. Entomol. 39, 29±32. Braverman, Y., 1992. The possible introduction to Israel of Culicoides (Diptera, Ceratopogonidae) borne animal diseases by wind. In: Walton, T.E., Osburn, B.I. (Eds.), Bluetongue, African Horse Sickness, and Related Orbiviruses, Proceedings of the Second International Symposium.CRC Press Boca Raton, FL. Dillon, M.L., Fitt, G.P., Hamilton, J.G., Rochester, W.A., Henderson-Sellers, B., McAleer, M., Jakeman, A.J., 1996. A simulation model of wind-driven dispersal of Helicoverpa moths. Ecol. J. Model. 86, 145±150. Doherty, R.L., Standfast, H.A., Dyce, A.L., Carley, J.G., Gorman, B.M., Filippich, C., Kay, B.H., 1978. Mudjinbarry virus, an orbivirus related to Wallal virus isolated from midges from the Northern Territory of Australia. Aust. J. Biol. Sci. 31, 97±103. Dyce, A.L., Standfast, H.A., Kay, B.H., 1971. Collection and preparation of biting midges (Fam. Ceratopogonidae) and other small Diptera for virus isolation. J. Aust. Ent. Soc. 11, 91±96. Fraval, A., El-Yousfi, M., 1989. Active and passive dispersal of Lymantria dispar (L.) (Lep., Lymantriidae) in cork-oak stands on the Atlantic coast of Morocco. J. Appl. Entomol. 108, 335±346 (English summary). Loxdale, H.D., Hadie, J., Halbert, S., Foottit, R., Kidd, N.A.C., Carter, C.I., 1993. The relative importance of short- and long-range movement of flying aphids. Biol. Rev. Camb. Phil. Soc. 68, 291±311. MathSoft, 1998. S-PLUS 5 Guide to Statistics. MathSoft, Inc., Seattle, WA, 1014 pp. Meeker, W.M., Escobar, L.A., 1998. Statistical Methods for Reliability Data. Wiley, New York, 680 pp. Muller, M.J., 1993. Distribution of vectors of arboviruses affecting sheep in Australia. Final Report, WRDC Project CTI3. CSIRO, Long Pocket, Brisbane. Muller, M.J., Standfast, H.A., St. George, T.D., Cybinski, D.H., 1982. Culicoides brevitarsis (Diptera: Ceratopogonidae) as a vector of arboviruses in Australia. In: St. George, T.D., Kay, B.H. (Eds.), Proceedings of the Third Symposium on Arbovirus Research in Australia, Brisbane, Qld, pp. 43±49. Murray, M.D., 1987a. Local dispersal of the biting midge, Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae) in south-eastern Australia. Aust. J. Zool. 35, 559±573. Murray, M.D., 1987b. Akabane epizootics in New South Wales: evidence for long-distance dispersal of the biting midge Culicoides brevitarsis. Aust. Vet. J. 64, 305±308.
254
A.L. Bishop et al. / Preventive Veterinary Medicine 47 (2000) 243±254
Murray, M.D., Kirkland, P.D., 1995. Bluetongue and Douglas virus activity in New South Wales in 1989: further evidence for long-distance dispersal of the biting midge Culicoides brevitarsis. Aust. Vet. J. 72, 56±57. Murray, M.D., Nix, H.A., 1987. Southern limits of distribution and abundance of the biting-midge Culicoides brevitarsis Kieffer (Diptera: Ceratopogonidae) in south-eastern Australia: an application of the GROWEST model. Aust. J. Zool. 35, 575±585. Nix, H.A., 1976. Environmental control of breeding, post-breeding dispersal and migration of birds in the Australian region. In: Frith, H.J., Calaby, J.H. (Eds.), Proceedings of the 16th International Ornithological Congress, Canberra, pp. 272±305. Sellers, R.F., Pedgley, D.E., Tucker, M.R., 1977. Possible spread of African horse sickness on the wind. J. Hyg. 79, 279±298. Sellers, R.F., Walton, T.E., Osburn, B.I., 1992. Weather, Culicoides, and the distribution and spread of bluetongue and African horse sickness viruses. In: Walton, T.E., Osburn, B.I. (Eds.), Bluetongue, African Horse Sickness, and Related Orbiviruses, Proceedings of the Second International Symposium. CRC Press, Boca Raton, FL, pp. 284±290. Standfast, H.A., Dyce, A.L., St. George, T.D., Muller, M.J., Doherty, R.L., Carley, J.G., Filippich, C., 1984. Isolation of arboviruses from insects collected at Beatrice Hill, Northern Territory of Australia, 1974±1976. Aust. J. Biol. Sci. 37, 351±366. Turnbull, B.W., 1974. Nonparametric estimation of a survivorship function with doubly censored data. J. Am. Stat. Assoc. 69, 169±173. Turnbull, B.W., 1976. The empirical distribution function with arbitrarily grouped, censored and truncated data. J. R. Stat. Soc. Ser. B 38, 290±295.