8
A Diffusion Model for Glossina palpalis gambiensis in Burkina Faso J. BOUYER 1,2 , A. SIBERT 1,3 , M. DESQUESNES 1,2 , D. CUISANCE 4 and S. DE LA ROCQUE 1,3 1 Centre de Coopération Internationale en Recherche Agronomique pour le Développement, Département d'Elevage et de Médecine Vétérinaire, BP 5035, 34032 Montpellier, France 2 Centre International de Recherche-développement sur l'Elevage en Zone Subhumide, BP 454, Bobo Dioulasso, Burkina Faso 3 Institut Sénégalais de Recherches Agricoles, Laboratoire National d'Etudes et de Recherches Vétérinaires, BP 2057, Dakar-Hann, Senegal 4 Conseil Général Vétérinaire, 251 rue de Vaugirard, 75732 Paris Cedex 15, France ABSTRACT The dispersal of Glossina species is of interest to pest control personnel since these flies are the biological vectors of human and animal trypanosomes in Africa. The design of control and/or erad- ication programmes requires an accurate knowledge of the ecological characteristics of tsetse flies and the geographic structure of their populations. The present study attempts to model the dispersal process of a riverine tsetse species, i.e. Glossina palpalis gambiensis Vanderplank in Burkina Faso along an apparent homogeneous gallery forest. While for savannah species, dispersal is usually modelled as a two-dimension- al random walk (in time and space) or diffusion (its continuous analogue), for riverine species, dispersal can be viewed more simply as a one-dimensional random walk. The data reported here show that the topol- ogy of the habitat, which is a system of tributaries rather than a straight line, has a great impact on the dis- persal process. Moreover, since only a part of the river system can be observed in practice, the effect of partial observation when estimating dispersal parameters can be quantified. The results reported here were obtained using a data set from a mark-release-recapture experiment carried out with G. p. gambiensis on a tributary of the Mouhoun River in Burkina Faso. The model was fitted to field data and used to estimate the displacement of a fly during 10% of its lifespan (13 kilometres) and the probability of it dispersing more than 10 kilometres from its initial position (P > 0.1). The analysis was carried out by either taking into account, or ignoring, the fact that only part of the river system was observed during the mark-release- recapture protocol. KEY WORDS dispersal, tsetse fly, Glossina palpalis gambiensis, diffusion model, mark-release- recapture 1. Introduction In West Africa, climate and land use pressures result in fragmentation of tsetse fly habitats, especially at the northern end of their distribu- tion areas (Hendrickx et al. 2004). Residual subpopulations, surrounded by semi-perma- nent barriers (particularly cotton crops), are increasingly isolated and can become a prior- ity target for sustainable control. If the tsetse population is not isolated, barriers of traps or screens impregnated with insecticides can be deployed, to prevent reinvasion (Cuisance and Politzar 1983, Politzar and Cuisance 1983, 1984), at least until the adjacent tsetse popula- tions are suppressed. To establish these barri- ers, traps or screens are placed every 100 metres along five to ten kilometre river sec- , 221–228. 221 © 2007 IAEA. M.J.B. Vreysen, A.S. Robinson and J. Hendrichs (eds.), Area-Wide Control of Insect Pests

Area-Wide Control of Insect Pests || A Diffusion Model for Glossina palpalis gambiensis in Burkina Faso

  • Upload
    j

  • View
    214

  • Download
    0

Embed Size (px)

Citation preview

A Diffusion Model for Glossina palpalisgambiensis in Burkina Faso

J. BOUYER1,2, A. SIBERT1,3, M. DESQUESNES1,2,D. CUISANCE4 and S. DE LA ROCQUE1,3

1Centre de Coopération Internationale en Recherche Agronomique pourle Développement, Département d'Elevage et de Médecine Vétérinaire,

BP 5035, 34032 Montpellier, France2Centre International de Recherche-développement sur l'Elevage en

Zone Subhumide, BP 454, Bobo Dioulasso, Burkina Faso3Institut Sénégalais de Recherches Agricoles, Laboratoire National

d'Etudes et de Recherches Vétérinaires, BP 2057, Dakar-Hann, Senegal4Conseil Général Vétérinaire, 251 rue de Vaugirard, 75732 Paris Cedex

15, France

ABSTRACT The dispersal of Glossina species is of interest to pest control personnel since these fliesare the biological vectors of human and animal trypanosomes in Africa. The design of control and/or erad-ication programmes requires an accurate knowledge of the ecological characteristics of tsetse flies and thegeographic structure of their populations. The present study attempts to model the dispersal process of ariverine tsetse species, i.e. Glossina palpalis gambiensis Vanderplank in Burkina Faso along an apparenthomogeneous gallery forest. While for savannah species, dispersal is usually modelled as a two-dimension-al random walk (in time and space) or diffusion (its continuous analogue), for riverine species, dispersalcan be viewed more simply as a one-dimensional random walk. The data reported here show that the topol-ogy of the habitat, which is a system of tributaries rather than a straight line, has a great impact on the dis-persal process. Moreover, since only a part of the river system can be observed in practice, the effect ofpartial observation when estimating dispersal parameters can be quantified. The results reported here wereobtained using a data set from a mark-release-recapture experiment carried out with G. p. gambiensis on atributary of the Mouhoun River in Burkina Faso. The model was fitted to field data and used to estimatethe displacement of a fly during 10% of its lifespan (13 kilometres) and the probability of it dispersingmore than 10 kilometres from its initial position (P > 0.1). The analysis was carried out by either takinginto account, or ignoring, the fact that only part of the river system was observed during the mark-release-recapture protocol.

KEY WORDS dispersal, tsetse fly, Glossina palpalis gambiensis, diffusion model, mark-release-recapture

1. Introduction

In West Africa, climate and land use pressuresresult in fragmentation of tsetse fly habitats,especially at the northern end of their distribu-tion areas (Hendrickx et al. 2004). Residualsubpopulations, surrounded by semi-perma-nent barriers (particularly cotton crops), areincreasingly isolated and can become a prior-

ity target for sustainable control. If the tsetsepopulation is not isolated, barriers of traps orscreens impregnated with insecticides can bedeployed, to prevent reinvasion (Cuisance andPolitzar 1983, Politzar and Cuisance 1983,1984), at least until the adjacent tsetse popula-tions are suppressed. To establish these barri-ers, traps or screens are placed every 100metres along five to ten kilometre river sec-

, 221–228.221

© 2007 IAEA.M.J.B. Vreysen, A.S. Robinson and J. Hendrichs (eds.), Area-Wide Control of Insect Pests

tions. Control campaigns preferably targetareas where agricultural barriers already exist,and the methods used vary in their cost effec-tiveness depending on their reliability and theoverall objective, i.e. eradication or suppres-sion (IAEA/FAO 2001). Information aboutthe degree of isolation between populationscan be very useful for the design, implementa-tion and monitoring of area-wide integratedpest management programmes (AW-IPM),especially when resources are limited.

The cotton belt of Mali and Burkina Fasohas been identified as a priority area for tsetsecontrol because of the potential benefits asso-ciated with the removal of trypanosomosis(Hendrickx et al. 2004). In Burkina-Faso,riverine species (Glossina palpalis gambien-sis Vanderplank and Glossina tachinoidesWestwood) are the main vectors of Africananimal trypanosomosis, since the only savan-nah species (Glossina morsitans submorsitansNewstead) is now restricted to protected areasin the south of the country. Current studies onthe population dynamics of vector speciestherefore focus on the valleys and riverineforests, using remote sensing and ecologicaldata to evaluate the feasibility of an eradica-tion campaign based on area-wide principlesand in support of the Pan African Tsetse andTrypanosomiasis Eradication Campaign(PATTEC) recently initiated in Burkina Faso(IAEA/FAO 2001). In particular, researchersand programme managers are interested in thedegree of isolation of vector populations, orconversely their ability to disperse as estimat-ed by migration rates between favourablelandscapes. For this purpose, both mark-release field experiments and populationgenetic analyses have proven to be very effi-cient tools (Solano et al. 2000, Krafsur andEndsley 2002, Krafsur 2003).

The methodology for estimating diffusionparameters by mark-recapture methods is nowrobust for species dispersing in two dimen-sions (Okubo and Levin 2001). For the savan-nah species of tsetse, such as Glossina morsi-tans morsitans Westwood, diffusion modelshave been successfully applied to model thedispersal process (Bursell 1970, Hargrove

1981, Hargrove and Lange 1989, Hargrove2000). However, for tsetse flies dispersingalong gallery forests, these are less well devel-oped (Rogers 1977, Randolph and Rogers1984, Cuisance et al. 1985).

Before addressing the problem of dispersalin fragmented landscapes, the present studyanalyses data obtained in a homogeneousriverine forest to understand the impact ofriver structure on the dispersal characteristicsof riverine tsetse (Cuisance et al. 1985). Inparticular, account is taken of the fact that ariver cannot be considered as a straight linebut rather as a river system. As underlined byBuxton (1955):

The fly belt occupied by G. palpalis is nearlyalways along the waterside. […] It is known thatthe insect moves very freely up or down streamor up a tributary. […] Evidently then the widthof the zones varies, but spontaneous movementof the insect is so closely confined to the vicini-ty of water that it is almost linear.

2. Materials and Methods

2.1. Field Data

The mark-release-recapture data were collect-ed in 1981 (Cuisance et al. 1985) on theDienkoa River, a tributary of the MouhounRiver in Burkina Faso. Forty-three biconicaltraps were deployed at distances of approxi-mately 500 metres over a distance of 20 kilo-metres along the main stream. The river sec-tion under study was bordered by a homoge-nous closed Guinean riparian forest. Traplocations (Fig. 1) were digitized from aLandsat Thematic Mapper image (pixel 30metres x 30 metres) from December 2000(cool dry season) using the original and verydetailed field map (Cuisance et al. 1985).

For this study, the data obtained from therelease of 8683 three-day-old male G. p. gam-biensis were analysed. The flies were from theCentre de Recherche sur les TrypanosomosesAnimales (CRTA) insectary and were fedonce on a rabbit on the day before release.They were not irradiated before release. Tenweekly releases, which were not analysed by

J. BOUYER ET AL.222

Cuisance et al. (1983), were performedbetween May and August 1981. During thesereleases, the initial release point (Fig. 1) waslocated in the middle of the trapped section.Flies were marked on their pronotum with adot of acrylic paint, the colour of which indi-cated the week of their initial release.Captures were performed three times perweek with traps set from 08.00 hours to 16.30hours each trapping day (On three days withheavy rain, traps could not be checked andonly releases were carried out). Marked tsetseflies were released again at the location wherethey were captured.

Trapping was terminated at the same datefor all cohorts so that data represent varioustimes from the release of the cohorts (15 andfour weeks for the first and last cohortsrespectively). The data available are of twotypes: total trap catches by time elapsed fromrelease cumulated over all ten cohorts(Cuisance et al. 1985), and a set of dataobtained from Cuisance’s diary correspondingto daily data cumulated for all traps.

2.2. Models

Adiscrete isotrope random walk (same proba-bility of going up- or downstream one dis-tance unit at each time unit), was used tomodel tsetse dispersal. It can be regarded asthe simplest individual-based model of disper-sal in one dimension. The model’s assump-tions are that, during one time unit (τ) a singlefly will travel exactly one unit length (λ) eitherto the right or left with equal probability.

In this case, if the number of time steps nis large, and denoting:

x = mλ, t = nτit can be shown that the probability density ofthe position at time t converges towards thedensity of a centred Gaussian random variablewith variance 2Dt, where D is the so-calleddiffusion coefficient and is equal to λ2/2τ(Williams et al. 1992).

Since mortality appears as the main causeof the end of the diffusion process, zero diffu-sion was used to model the progress of fly dis-persal. Such a model amounts to stopping an

A DIFFUSION MODEL FOR GLOSSINA PALPALIS GAMBIENSIS 223

Figure 1. Location of the biconical traps of the 1981 study along the Dienkoa river.

individual’s trajectory after a random timerepresenting the individual’s lifespan. Sincethe mortality rate between the first release andfirst recapture cannot be disentangled fromthe overall recapture rate, and since the mor-tality rate appeared almost constant during thelater stages of the releases, a constant mortal-ity rate, µ was used throughout (the model canhowever be easily extended to more accuratedata on fly mortality). The parameter µ wasroughly estimated as the log-regression coef-ficient of the number of recaptured fliesagainst time. Based on the homogeneity of thegallery and trapping conditions, two strongassumptions had to be made, i.e. the time andspace independence of capture and mortalityrates.

The model was first applied on a theoreti-cal system of tributaries, comprising an infi-nite main river where diffusion originates, andone or several tributaries connected to themain stream at definite locations (Fig. 2).Boundary conditions were isotropic at bifur-cations (when a fly came to a junctionbetween a tributary and the main river it hadequal probability of going up the tributary, orup or down the main stream), and reflexive atthe end of tributaries (when it reached the endof a tributary it always came back down at thenext step). In order to assess the effects ofgeometry on the diffusion process, configura-tions with n = 2 and n = 10 tributaries werestudied. To obtain a symmetrical process, therelease was simulated in the centre of two trib-utaries distant of 2l (Fig. 2): the distancebetween the origin of the process and the nexttributary was thus set to l = 1, and thediffusion coefficient was such that 2D = 1.

Several lengths of tributaries (L = 1, 3, 10)were used in the simulations.

The model’s parameters (D, µ) were thenestimated using the field data set and follow-ing two assumptions: (1) the trapping systemallows a complete observation of the dispersalprocess which occurs only along the mainriver, and (2) the trapping system allows onlya partial observation of the dispersal processwhich occurs throughout the whole river sys-tem (river basin with all its tributaries).

Finally, two “summary statistics” were cal-culated using the two-parameter estimationsin a hypothetical case where two populationsare ten kilometres apart on a straight rivercourse without tributaries: (1) the distance tothe origin exceeded during 10% of the fly’saverage lifespan, and (2) the probability of afly travelling beyond a given distance to theorigin. Details of the mathematical processwill be described elsewhere (Sibert et al., inpreparation).

3. Results

3.1. Theoretical Systems

The effects of considering only the main riverfor estimating D and µ while dispersal occursthroughout a river system was investigatedusing a variety of symmetric and asymmetricriver systems. The results of the simulationsdepend on the complexity of the river systemand the value of the diffusion coefficient. Theresults from the nine theoretical symmetricriver systems (Fig. 3) are only illustrative butthey use a realistic D, namely 0.5 square kilo-metres per day and a realistic river system,

J. BOUYER ET AL.224

Figure 2. Symmetric systems used for the simulations.

with ten tributaries. To appreciate the poten-tial impact of partial observation on the esti-mation of D and µ, three variables are illus-trated: (1) the proportion of flies observed onthe main river (to illustrate the effect of partialobservation on the estimation of mortalityrates), (2) the average position of the popula-tion (to show that an isotropic random walkcan lead to an apparent displacement of thepopulation on the observed section), and (3)the ratio between observed and actual vari-ance (to illustrate the impact of partial obser-vation on the estimation of the diffusion coef-ficient (Fig. 3).

From the simulations, it is clear that spatialsampling, i.e. observing only the main riverwith traps while dispersal occurs on the wholeriver system, can lead to overestimation ofmortality rates by up to 30% at day 40 (Fig. 3,first column, first line). The average positionof the population on the main course canevolve up to one kilometre from the releaseposition (Fig. 3, third column, second line),towards the centre of gravity of the system,leading to an apparent drift of the population,which would be impossible if the diffusionprocess occurred in an infinite symmetricalsystem. Finally, the variance in position (pro-

A DIFFUSION MODEL FOR GLOSSINA PALPALIS GAMBIENSIS 225

Figure 3. (upper graphs) Proportion of flies observed on the main river, (middle graphs) aver-age position in kilometres, and (lower graphs) ratio between observed and real variance infunction of time (in days) in nine hypothetical river systems (n=10, L=1, 2 or 3 and l=1, 3 or10).

J. BOUYER ET AL.226

portional to the diffusion coefficient) can beunderestimated by 15% in the case of com-plex systems (Fig. 3, first column, third line).

3.2. Real Data Set

With the real data set presented above, D andµ could be estimated depending on two differ-ent assumptions: (1) dispersal occurs only onthe main river and the trapping system pro-vides complete information, and (2) dispersaloccurs on the whole river system and the dataset is thus spatially sampled.

If dispersal occurs only on the main river,a complete observation of a random walk on astraight line can be implied. Daily mortalityrate estimated over the first month is then6.5%, corresponding to a mean lifespan of15.5 days. The corresponding diffusion coeffi-cient is 0.29 km2/day.

In the alternative hypothesis (partial obser-vation of a random walk on a tree), the esti-mated mortality rate is only 4.4%, correspon-ding to a mean lifespan of 22.7 days. The dif-fusion coefficient is 0.46 km2/day.

The two estimation errors due to spatialsampling demonstrate the same tendency, i.e.an underestimation of the probability of long-distance movement in the case of partialobservation. When the model coefficients arecorrected for the partial observation bias, theaverage distance covered by a fly during 10%of its lifespan increases from four to 13 kilo-metres. At the same time, the probability of afly reaching a population located ten kilome-tres away increases from less than 0.01 tomore than 0.1. In other words, the (apparent)displacement – and the (apparent) probabilityof reaching another population – is muchlower than it is if only the data from the mainriver are included.

4. Discussion

The estimates of mortality rate obtained fromthe raw data are 30% lower when a partial sys-tem of tributaries is assumed (0.040 versus0.065). However, mortality remains very highfor this highly favourable environment

(closed and conserved Guinean gallery for-est). Density-dependent factors may induce anincrease of mortality rates next to the releasepoints due to increased local density (Rogersand Randolph 1984). The assumption of aconstant mortality rate in space is then notvalid as the fly density will be higher at therelease point, at least during the releases andmaybe during the following days. In all cases,taking into account the spatial complexityseems very important when estimating mor-tality rates, in the case of sterile insect releas-es for example. Moreover, dispersal over theentire available system (including the tributar-ies) may be considered as a way to reducemortality density-dependent factors as muchas possible.

The structure of a river system is not staticin time, but evolves with macro-climatic vari-ations through seasons. In the hot dry season,the small tributaries are drained and becomeunsuitable for tsetse survival, leading to theirconcentration in the moister section, i.e. in themain river. While dispersing with the samediffusion coefficient on either the entire sys-tem or the main course only, the probability oflong-distance movements on the main riverwill increase, thereby enhancing long-dis-tance dispersal during the dry season. Thus,on the one hand, tributaries can act as brakesto longitudinal movements particularly duringthe rainy season while, on the other hand, themortality rate will decrease as relative humid-ity increases during the rainy season. Thesetwo parameters have thus opposing effects onlong-distance dispersal.

In the model used, the two-dimensionalstructure of the system was not taken intoaccount, and the possibility cannot be exclud-ed that some flies escaped from the gallery(main stream or tributaries) and re-entered thesystem at some other point. However, duringthe release period (beginning of the rainy sea-son), 145 traps were set in two circles locatedat one and two kilometres from the releasepoint in the neighbouring savannah (Cuisanceet al. 1983), and only 1.1% (36 from 3228) ofthe captured flies were caught outside thegallery. Moreover, as movement in the savan-

nah is a two-dimensional diffusion, the latterenvironment can be considered less diffusivethan the gallery itself and its contribution tolong displacements along the main course canbe neglected. During the rainy season howev-er, riverine flies migrate into nearby savannahareas, especially while following their hosts,and their movement, perpendicular to the riversystems, should be taken into consideration ifattempting to estimate the probability of ariver basin being “isolated” from another.

In the present work, the experimental sitewas chosen for its homogeneous, closedgallery. Provided that the spatial complexity istaken into account, a diffusion process wasappropriate to describe tsetse fly dispersal.However, in fragmented landscapes, galleryforests are in fact very heterogeneous (Morel1983). In such situations, a diffusion processmay be unsuitable to describe fly dispersal, asthe decision to disperse becomes a trade-offbetween a reduction in mortality-dependentfactors and an increase in mortality-independ-ent factors while moving from a favourableecosystem (natural riverine forest) to a lessfavourable one (disturbed riverine forest)(Blondel 1995). Experimental releases inte-grating the lessons from the present work arepresently being conducted in fragmentedlandscapes previously characterized by phyto-sociological analysis (Bouyer et al. 2005), tointegrate spatial heterogeneity into dispersalmodels. In association with population genet-ics, these should be very useful in understand-ing the population structure within theMouhoun river basin and may allow the iden-tification of “natural” barriers in the planningof an AW-IPM programme in Burkina Faso.

5. Conclusions

A normal diffusion model can fit the data onriverine tsetse dispersal in homogeneous land-scapes provided that the spatial complexity ofthe river system is taken into account. Dailymortality rate has an important effect on dis-persal and especially on long-distance move-ments. Since they depend mainly on the sea-son and the conservation status of the gallery

forest, these elements may have a great impacton the tsetse dispersal process. Mark-releasefield experiments, supported by new tech-niques like dispersal models, remote sensingand geographic information systems (GIS)can be very useful in designing the sequentialprocess of an AW-IPM project. They couldalso be used to optimize the releases of sterileflies during an eradication campaign.

6. References

Blondel, J. 1995. Biogéographie. Approcheécologique et évolutive. Masson, Paris,France.

Bouyer, J., L. Guerrini, J. César, S. de laRocque, and D. Cuisance. 2005. A phyto-sociological analysis of the distribution ofriverine tsetse flies in Burkina Faso. Medicaland Veterinary Entomology 19: 372-378.

Bursell, E. 1970. Dispersal and concentrationof Glossina, pp. 382-394. In Mulligan, H.W.(ed.), The African trypanosomiases. GeorgeAllen and Unwin, London, UK.

Buxton, P. A. 1955. The natural history oftsetse flies. An account of the biology of thegenus Glossina (Diptera). Lewis H. K. andCo. Ltd., London, UK.

Cuisance, D., and H. Politzar. 1983. Etude del’efficacité contre Glossina palpalis gambi-ensis et Glossina tachinoides de barrièresconstituées d’écrans ou de pièges biconiquesimprégnés de DDT, de deltaméthrine ou dedieldrine. Revue d’Elevage et de Médecinevétérinaire des Pays tropicaux 36: 159-168.

Cuisance, D., J. Février, J. Filledier, and J.Dejardin. 1983. Etude sur le pouvoir dedispersion des glossines. CRTA/IEMVT/OMS-83, Bobo Dioulasso, Burkina Faso.

Cuisance, D., J. Février, J. Dejardin, and J.Filledier. 1985. Dispersion linéaire deGlossina palpalis gambiensis et G. tachi-noides dans une galerie forestière en zonesoudano-guinéenne (Burkina Faso). Revued’Elevage et de Médecine vétérinaire desPays tropicaux 38: 153-172.

Hargrove, J. W. 1981. Tsetse dispersal recon-sidered. Journal of Animal Ecology 50: 351-373.

A DIFFUSION MODEL FOR GLOSSINA PALPALIS GAMBIENSIS 227

Hargrove, J. W. 2000. A theoretical study ofthe invasion of cleared areas by tsetse flies(Diptera: Glossinidae). Bulletin ofEntomological Research 90: 201-209.

Hargrove, J. W., and K. Lange. 1989. Tsetsedispersal viewed as a diffusion process.Transactions of the Zimbabwe ScientificAssociation 64: 1-8.

Hendrickx, G., S. de la Rocque, and R. C.Mattioli. 2004. Long-term tsetse and try-panosomiasis management options in WestAfrica. FAO, Rome, Italy.

(IAEA/FAO) International Atomic EnergyAgency/Food and Agriculture Organiza-tion of the United Nations. 2001.Workshop on strategic planning of area-widetsetse and trypanosomiasis control in WestAfrica. IAEA-63, Ouagadougou, BurkinaFaso.

Krafsur, E. S. 2003. Tsetse fly populationgenetics: an indirect approach to dispersal.Trends in Parasitology 19: 162-166.

Krafsur, E. S., and M. A. Endsley. 2002.Microsatellite diversities and gene flow inthe tsetse fly, Glossina morsitans s.l. Medicaland Veterinary Entomology 16: 292-300.

Morel, P. C. 1983. Guide pour la déterminationdes arbres et des arbustes dans les savanesOuest-Africaines. IEMVT, Maisons-Alfort,France.

Okubo, A., and S. A. Levin. 2001. Diffusionand ecological problems: modern perspec-tives. Springer-Verlag, New York, USA.

Politzar, H., and D. Cuisance. 1983. A trap-barrier to block reinvasion of a river system

by riverine tsetse species. Revue d'Elevage etde Médecine vétérinaire des Pays tropicaux36: 364-370.

Politzar, H., and D. Cuisance. 1984. An inte-grated campaign against riverine tsetse fliesGlossina palpalis gambiensis and Glossinatachinoides by trapping and the release ofsterile males. Insect Science and itsApplication 5: 439-442.

Randolph, S. E., and D. J. Rogers. 1984.Movement patterns of the tsetse fly Glossinapalpalis palpalis (Robineau-Desvoidy)(Diptera: Glossinidae) around villages in thepre-forested zone of Ivory Coast. Bulletin ofEntomological Research 74: 689-705.

Rogers, D. J. 1977. Study of a natural popula-tion of Glossina fuscipes fuscipes Newsteadand a model of fly movement. Journal ofAnimal Ecology 46: 309-330.

Rogers, D. J., and S. E. Randolph. 1984. Areview of density-dependent processes intsetse population. Insect Science and itsApplication 5: 397-402.

Solano, P., S. de la Rocque, T. de Méeus, G.Cuny, G. Duvallet, and D. Cuisance. 2000.Microsatellite DNA markers reveal geneticdifferentiation among populations ofGlossina palpalis gambiensis in the agropas-toral zone of Sideradougou, Burkina Faso.Insect Molecular Biology 9: 433-439.

Williams, B., R. Dransfield, and R.Brightwell. 1992. The control of tsetse fliesin relation to fly movement and trapping effi-ciency. Journal of Applied Ecology 29: 163-179.

J. BOUYER ET AL.228