Understanding biological control of greenhouse whitefly with the parasitoid Encarsia formosa : from individual behaviour to population dynamics

H.J.W. van Roermund

Research output: Thesisinternal PhD, WU


The greenhouse whitefly, Trialeurodes vaporariorum (Westwood) (Homoptera, Aleyrodidae), is a very common, highly polyphagous pest insect all over the world. Biological control of whiteflies with the parasitoid Encarsia formosa Gahan (Hymenoptera, Aphelinidae) was already applied in the 1920s in England, Australia, New Zealand and Canada. The use of the parasitoid was discontinued in the fourties and fifties when chemical pesticides were used extensively. In the seventies, when the first problems with pesticide resistance occurred in Western Europe, interest in using the parasitoid increased again. A reliable introduction scheme of the parasitoid was found by a 'trial and error' approach: natural enemies were released at different times and in different numbers, and their level of control was examined. In 20 of the 35 countries with a greenhouse industry, the parasitoid is used on about 5000 ha. Biological control with E. formosa is now used commercially in 90% of the tomato growing areas in the Netherlands. On several other important greenhouse crops such as cucumber and gerbera, biological control of whitefly is not so successful. This study aims at integrating existing knowledge on the major processes known to affect the whitefly-parasitoid interaction in a crop by means of an explanatory simulation model. The goal is to obtain quantitative understanding of the tritrophic system cropgreenhouse whitefly- E. formosa to explain failure or success of biological control. With the model we are able to (1) explain the ability of E. formosa to reduce whitefly populations in greenhouses on crops like tomato, (2) improve introduction schemes of parasitoids for crops where control is more difficult to obtain and (3) predict effects of changes in cropping practices (e.g. greenhouse climate, choice of cultivars) on the reliability of biological control. Direct observation experiments on foraging of E. Formosa. When the present research project started, the behaviour of E. formosa had been observed in various experiments. These experiments resulted in the following picture. E. formosa is a solitary larval parasitoid: females lay one egg per host during an oviposition. Like in other synovigenic parasitoids new eggs mature when the egg load of the parasitoid drops below the storage capacity, which is 8-10 mature eggs for E. formosa. About ten days after oviposition the immature parasitoid pupates in the host pupa, which (in case of greenhouse whitefly) then turns black and parasitism can easily be seen from the outward appearance of the whitefly. Female parasitoids produce daughters parthenogenetically. Males are rarely observed. The parasitoid searches for the sessile whitefly immatures by flying or hopping from leaf(let) to leaf(let) without distinguishing between infested and clean plants or leaves before landing. Once on the leaf she starts walking and drumming the leaf with her antennae. Hosts are encountered randomly and the walking pattern is not changed after an encounter with a host. After an encounter, four behaviours on that host can be distinguished: the parasitoid may reject the host after an inspection with the antennae (antennal rejection) or after insertion of the ovipositor (ovipositorial rejection), she may parasitize (oviposition) or she may use the host as a food source (host feeding). However, these earlier experiments did not lead to a complete picture of foraging behaviour. Quantitative data on some aspects were lacking, such as the parasitoids' searching or walking activity between host encounters, and the effect of temperature on the foraging processes. In many of the earlier experiments parasitoids were confined to an experimental arena, and therefore little was known about the time allocation of the parasitoid on leaves, such as the time until leaving, the time spent on upper and lower leaf side, and how these are affected by encounters with or ovipositions in hosts. In this thesis, these gaps in our knowledge are first identified and studied experimentally. Chapters 2, 3 and 4 describe experiments where individual parasitoids were observed continuously until they flew away, either on clean tomato leaflets, on<br/>leaflets with honeydew, or on leaflets with unparasitized and parasitized whitefly larvae. In Chapter 2 the residence times of the parasitoids on leaflets are discussed. In Chapter 3 the leaving tendency of the parasitoid from the leaflet and effects of several<br/>intra-patch experiences with hosts are quantified. In Chapter 4 other basic aspects of foraging are quantified, such as the parasitoids' walking speed and walking activity, the probability of each handling behaviour to occur after an encounter with a host and the host handling times. These data enable quantification of the foraging process of the parasitoid from landing on a leaf until departure. For an overview of the subsequent processes, see Figure 1 of Chapter 4. The work described in Chapters 2, 3 and 4 resulted in the following conclusions:<br/>- The parasitoid E. formosa searches at random without a preference for the edge or the middle of a leaf, or for the upper or lower leaf side, whereas whitefly immatures (the hosts) are present only on the lower side of a tomato leaflet.<br/>- The median residence time of the parasitoid on uninfested tomato leaflets (or giving up time, GUT) is 18.6 min at 20, 25 and 30°C and equal to that on infested leaflets on which no hosts are encountered.<br/>- Parasitoids are arrested on the leaf by encounters with, and especially by ovipositions in, unparasitized hosts, by encounters with parasitized (unsuitable) hosts and by contact with honeydew. GUT since latest host encounter is again 18.6 min, also when the hosts were parasitized, but increases to 40 min after the first oviposition in an unparasitized host.<br/>- Parasitoids are arrested on the lower leaf side by encounters with hosts, and especially by ovipositions in unparasitized hosts. The median time since the arrival on a particular leaf side or, if it occurred, since the latest host encounter on that leaf side until changing to the other side (TUC) is initially 11.6 min and drops to 5.7 min after both leaf sides have been visited. After the first<br/>oviposition in an unparasitized host, TUC since latest host encounter on the lower leaf side (where hosts are present) becomes twice as long.<br/>- Parasitoids usually leave from the upper leaf side, where no hosts are present. The patch-leaving behaviour of the parasitoid can be described by a stochastic threshold mechanism, which is characterized by a certain tendency (probability per time) to leave. The parasitoid leaves after the host encounter rate falls below a certain threshold (encounters per time, which is the reciprocal of GUT). This threshold is not fixed, however, but shows a great variation and is expressed as a probability.<br/>- The parasitoids' walking speed increases linearly between 15 and 25-30°C.<br/>- The parasitoids' walking activity is very low at temperatures below 18°C and increases to about 75% of the total time on the leaf at 20, 25 and 30°C. The walking activity is not affected by host encounters, but decreases with decreasing egg load after 4 ovipositions.<br/>- The percentage of encounters resulting in an oviposition is about 75% for the most preferred stage (unparasitized L4 larva), but decreases with decreasing egg load.<br/>- Host handling behaviour and handling time is not influenced by the host plant.<br/>- The total handling time (including drumming etc.) for antennal rejection of an unparasitized hosts is about 20 s, for oviposition and for ovipositional rejection about 6 min, and for host feeding about 15 min. These handling times slightly differ when hosts are parasitized.<br/>- Self-superparasitism is not observed. Conspecific-superparasitism occurs in 14% of the contacted hosts containing a parasitoid egg, but is not observed anymore when the parasitoid egg had hatched.<br/>- No difference is observed in host handling behaviour between naive and experienced parasitoids.<br/>- Many inactive parasitoids are observed when the barometric pressure had decreased over a time span of at least 12 h.<p><strong>Simulation models of foraging behaviour of <em>E. formosa</em></strong><p>The information described above is used as input in the simulation models of <em>E</em> . <em>formosa's</em> foraging behaviour, which is described in <strong>Chapters 5, 6 and 7</strong> . Here, foraging behaviour is analyzed using Monte Carlo simulation at three spatial scales: in a small experimental arena, on a tomato leaflet and on a tomato plant. For an overview of the models, see the flow diagrams (Figure 1) of these chapters. Foraging behaviour is first studied at these small spatial scales, to better understand the quantitative effects of parasitoids on whitefly populations as observed in a crop.<p>The above simulation models are <em>mechanistic,</em> that is, they explain how parasitoids, in terms of searching efficiency, host handling and available eggs, realize the observed level of parasitism. Mechanistic explanations can help to understand failure or success of biological control in practice. The models do not explain why the parasitoids choose to behave in this way, in terms of the selection pressure acting on them. Thus, they do not provide a <em>functional</em> explanation <em></em> of the observed behaviour. This is subject of study in optimal foraging models.<p>The simulated number of hosts encountered, parasitized and killed by host feeding, and the residence times on leaflets are validated with experimental data and the simulation results agree well with these observations. According to the model, E. formosa can parasitize 16 hosts per day on average at 25°C on a tomato leaflet if they start searching with a full batch of mature eggs and if host density is not limiting. Thus about 7 new eggs mature during the day (16 h) at that temperature. From the second day onwards, the parasitoid can parasitize I I hosts per day, due to egg limitation: if the parasitoid laid all eggs the preceding day, only 4 eggs mature during the night (of 8 h) at 25°C, so the parasitoids do not have a full batch of mature eggs the next morning. The model shows that at a density of I L3 larva per tomato leaflet, 15.7% of the parasitoids discover the larva before they leave. Also at higher host densities, not all hosts are encountered and patches (leaflets) are not depleted after one visit. Variation in number of encounters and ovipositions between parasitoids is considerable, mainly caused by the random encounter of hosts, the variation in handling behaviour of an encountered host and by the variation in GUT and TUC.<p>In greenhouses, whiteflies show a clustered distribution over plants and leaves and average numbers are usually very low. The models show that at such conditions the number of parasitizations on tomato leaflets or plants is strongly affected by the leaf area, the parasitoids' walking speed and walking activity, the probability of oviposition after encountering a host, the initial egg load (egg load at the beginning of the experiment) and the ratio of search times on both leaf sides. At extremely high host densities, the egg storage capacity and the initial egg load of the parasitoid are most important, and on plants with a clustered host distribution also the parasitoids' GUT.<p>At all spatial scales tested, the number of encounters, ovipositions and host feedings increase with host density with a decelerating rate until a maximum level is reached. This shape of the curves resemble a Holling Type II functional response, which is caused by the parasitoids' decreasing walking activity and probability of oviposition after encountering a host when egg load decreases. This is predominant at all levels, and even a change in GUT from 18.6 to 40 min after the first oviposition on the leaf does not result in an accelerating increase of the curve. The shape of the curves, describing the effect of host density on parasitism as a result of the basic processes, helps to understand the dynamics of the host-parasitoid interaction at the population level. In case of a Type II functional response, percentage parasitism declines with increasing host density and parasitism is inversily density dependent. A high host density thus reduces the per capita parasitization pressure caused by one parasitoid. According to theory, inversely density dependence tends to have a destabilizing effect on the dynamics of host and parasitoid. However, the functional response or the parasitization pressure caused by one parasitoid is only one factor in determining the dynamics at the population level. Another factor is the number of parasitoids on the leaf. For E. formosa, the effect on the population level depends on the balance between the parasitization pressure caused by one parasitoid and the arrestment and subsequent aggregation of parasitoids on leaves with high host density (see Chapter 10).<p><strong>Life-history parameters of greenhouse whitefly and <em>E. formosa</em></strong><p>In Chapters 8 and 9, life-history parameters of the greenhouse whitefly and <em>E. formosa</em> are reviewed. Data from literature were selected on development rate of each immature stage, percentage mortality of each immature stage, sex ratio, longevity, preoviposition period, period of increase of daily oviposition, fecundity and oviposition frequency. Most of these experiments have focused on the effect of temperature with little attention to other environmental factors such as humidity or light. With these data, the relationship between the life-history parameters and temperature are assessed by non-linear regression. Five mathematical equations were fitted, the best being selected on the basis of the coefficient of determination (r <sup>2</SUP>) and on visual comparison of the curves, which was necessary to check whether a curve was biologically realistic, particularly the tails. Coefficients to describe the mean of each life-history parameter as a function of temperature are summarized in these chapters. Coefficients of variation (cv: sd/mean) among individuals are also given. These coefficients are used as input in the submodels of population development of whitefly and parasitoid (see Chapter 10).<p>For greenhouse whitefly, the life-history parameters depend very much on the type of host plant. For <em>E. formosa</em> , data for several host plants were combined. The high r2 values indicate that host plant effects can be disregarded for the parasitoids' life-history parameters, except for oviposition at low host densities, which is caused by differences in the parasitoids' walking speed and walking activity on leaves with a different morphology. The host stage originally parasitized strongly affects the immature development rate and immature mortality of the parasitoid.<p>The development rate is calculated as the reciprocal of the stage duration. For all immature stages of whitefly and parasitoid the relationship between development rate and temperature is described by the Logan curve: just above the lower threshold temperature, the development rate increases exponentially to an optimum, whereafter it declines sharply until the upper lethal temperature has been reached. The relationships of longevity, fecundity, and oviposition frequency with temperature are described by the Weibull curve: they increase exponentially from the lower lethal temperature to an optimum, whereafter they decrease exponentially. Only for <em>E.</em> formosa, the longevity decreases exponentially with temperature and an optimum was not found at greenhouse conditions. No relationship with temperature is found for the immature mortality, the sex ratio and the cv values of the life-history parameters of whitefly and parasitoid.<p><strong>Simulation model of whitefly-parasitoid interaction in a crop</strong><p>The final model simulates the population dynamics of the pest insect-parasitoid interaction in a tomato crop and is described in Chapter 10. The model is based on the parasitoids' searching and parasitization behaviour and on developmental biology of the two insect species. This model comprises several submodels, such as the submodel for whitefly population development, for parasitoid population development, for the parasitoids' foraging behaviour on tomato leaflets (model of Chapter 6), for spatial distribution of whitefly and parasitoid in the canopy, for dispersion of adult whiteflies and parasitoids from leaf to leaf, for leaf production and a submodel for checking simulation errors. Life-history parameters of Chapters 8 and 9 are used as input in the submodels for population development of whitefly and parasitoid on tomato. For an overview of the model, see the relational diagrams (Figures 1, 2 and 3) of Chapter 10. The model is unique in that it is an individual-based model which simulates local searching and parasitization behaviour of a large number of individual parasitoids in a whitefly-infested crop. The model includes stochasticity and spatial structure which is based on location coordinates of plants and leaves.<p>The model is validated with population counts from experiments on tomato with and without introduction of <em>E.</em><em>formosa</em> in small greenhouse compartments and in a large commercial greenhouse. The simulated population increase of greenhouse whitefly in the <em>absence</em> of parasitoids agree well with the observations. This result can for an important part be explained by the accurate estimates of the life-history parameters, which are based on many experiments at a wide temperature range (see Chapters 8 and 9).<p>With these life-history parameters as input in the model, the intrinsic rate of increase (r <sub>m</sub> ) of both insect species is simulated. The (r <sub>m</sub> ) of <em>E. formosa is</em> much higher than that of the greenhouse whitefly above 14°C. The r <sub>m</sub> of a parasitoid however, plays a limited role in biological control, because it is only valid when all parasitoids can lay their daily egg load. This can only happen at extremely high host densities when the parasitoids do not have to spent much time searching for hosts. In greenhouses whitefly densities are usually much lower and the realized whitefly density depends on the parasitoids' searching efficiency. Therefore, to evaluate and understand success or failure of biological control, (r <sub>m</sub> ) values are inappropiate and it is essential to built models which include searching and parasitization behaviour of the natural enemy at very low host densities.<p>Also in the <em>presence</em> of parasitoids, the simulation results agree well with greenhouse observations on tomato. Apparently, the hypothesized random host encounter of <em>E. formosa</em> in a tomato crop is reliable. In the model, the parasitoid does not distinguish between uninfested and infested leaflets before landing, the parasitoid searches randomly for hosts once on the leaflet, and shows a strong arrestment effect: it stays longer on the leaflet once a host is encountered. Simulations show that the adult parasitoid-whitefly ratio is very high and can even reach 250:1. As a result, whiteflies are suppressed rather than regulated by the parasitoids at extremely low host densities (&lt;0.3 unparasitized pupae per plant), but never become extinct. These whitefly densities are much lower than the economic damage threshold for greenhouse whitefly. Percentage black pupae fluctuates between 40 and 70%. According to the model, the parasitoid adults reach high densities of 7.4 per plant, but due to the low whitefly density not more than 1% of the parasitoids is searching on infested leaflets.<p>The giving up times (GUT) of <em>E. formosa</em> vary to a large extend. The degree of whitefly control is very sensitive to those GUT's lower than 800 s of the parasitoids. The whiteflies are suppressed at much lower densities when the parasitoids stay at least five minutes on each leaflet (infested or uninfested) and after each host encounter. This minimum time increases the arrestment effect and the resulting percentage of parasitoids on infested leaflets, thereby reducing the chance that clustered hosts escape from parasitism. When variation in GUT is excluded in the model, the whitefly population becomes less stable and nearly goes extinct. Variation in GUT on leaflets induces host refuges from parasitoid attack. Also from more theoretical studies, host refuges are known to stabilize populations.<p>Whitefly adults migrate to young leaves in the top of the plant. A slower leaf production results in a longer stay and more ovipositions of whitefly adults on a particular leaflet. Thus, the same number of hosts are distributed over fewer leaflets, resulting in a more aggregated host distribution. Whiteflies are then suppressed by <em>E. formos</em> a to much lower numbers, according to the model. Parasitism of one <em>E. formosa</em> female on a tomato leaflet is inversely density-dependent, which is caused by a decreasing walking activity and probability of oviposition after encountering a host (Chapters 5, 6 and 7). Host aggregation thus 'dilutes' the per capita parasitization pressure caused by one parasitoid on the leaflet. The effect on the population level however, depends on the balance between this 'dilution' effect and the strength of the arrestment and aggregation of <em>E. formosa</em> . Therefore, the stronger whitefly reduction when whiteflies are more aggregated is caused by a stronger parasitoid arrestment and subsequent increase in the relative number of parasitoids searching on infested leaflets.<p>This shows that differences in whitefly distribution among crops are one factor in causing differences in success of biological control. Other factors are the size, number and surface (hairiness) of leaves in the canopy. Leaf size and total leaf area have a strong effect on whitefly control according to the model, caused by their direct effect on host density. Furthermore, leaf size and leaf surface strongly affect the efficiency of <em>E. formosa</em> by changing the parasitoids' arrestment effect (GUT) and the walking speed and activity, respectively.<p>Another important factor is the whitefly development duration on the crop. The model shows that plant resistance breeding aimed at an increase in egg-to-adult duration of the whiteflies is very efficient in causing a severe reduction of whitefly numbers, when biological control is applied. Observed development times of whitefly differ very little between tomato genotypes, and a much larger difference is found for whitefly longevity, oviposition rate and immature mortality. These parameters have a smaller effect on whitefly population development.<p>The important factors or crop properties affecting the success of biological control cannot be compared independently. For instance crops with large leaves usually have lower number of leaves which are produced at a lower rate than crops with small leaves or leaflets. It is particularly the combined effect of these important factors that can be tested with this model for different crops or plant varieties.<p>In biological control programs, parasitoids are usually tested in small-scale experiments at high host densities before introduction in the field. As a result, maximum daily oviposition of parasitoids is measured, whereas this study shows that egg storage capacity and egg maturation rate of <em>E. formosa</em> is not important for the level of whitefly control. In commercial greenhouses, whitefly densities have to be very low for biological control to be judged successful, therefore effective host searching is the most essential process. When selecting parasitoids for biological control, attention should be focused on the parasitoids' arrestment effect (minimum GUT), walking speed, walking activity, the probability of oviposition after encountering a host, the ratio of search times on both leaf sides and on longevity, when comparing different synovigenic and solitary parasitoid species with random search. These characteristic attributes of parasitoids are easily measured in laboratory studies. Again, they cannot be compared independently, however, because the attributes of natural enemies are often found in particular combinations. The combined effect of these important attributes of a parasitoid can be tested with this model.<p><strong>Epilogue</strong><br/>The present study aimed at integrating existing knowledge on the major processes known to affect the whitefly-parasitoid interaction in a crop. Because of the multitude of relationships between the three trophic levels (crop-pest-parasitoid), it was decided to follow a combined experimental-simulation approach. The goal was to obtain quantitative understanding of the tritrophic system to explain failure or success of biological control. With the model we now unravelled the ability of <em>E. formosa</em> to reduce whitefly populations on greenhouse tomato. The study resulted in increased understanding of the relative importance of basic processes that affect the population interaction of whitefly and natural enemy. The life-history of parasitoids, often summarized in a (r <sub>m</sub> ) value, are less important than the parasitoids' searching capacity. This shows that in addition to the traditional selection criteria, a criterion based on searching efficiency is essential. The study has further generated knowledge on foraging behaviour of <em>E. formosa</em> . The tremendous effect of variation in patch times of individual parasitoids on the whitefly population in the greenhouse shows that individual-based models which include stochasticity and local searching behaviour are a necessity when developing models of host-parasitoid interaction at extremely low host densities and aggregated host distributions.<p>One of the main questions was to identify the main causal factors for differences among crops in success of biological control of whitefly. The parasitoid is more successful on tomato than on cucumber or gerbera. The present study showed that attention should be focused on differences in the parasitoids' arrestment effect (GUT), the parasitoids' walking speed and activity, the whitefly development duration and the number, size and production of leaves in the canopy. These parameters have also been quantified for cucumber and gerbera and some are very different from those of tomato (see Chapters 2, 3 and 8). The next step in the research is to use the simulation model presented in Chapter 10 for the other two crops and evaluate the main causal factors for success or failure of biological control.<p>When adapting the parameters in the model for gerbera and cucumber we are able to (1) explain the lower ability of the parasitoid to reduce whitefly populations on these crops, (2) improve introduction schemes of parasitoids for these crops, and (3) predict effects of changes in cropping practices (e.g. greenhouse climate, choice of cultivars) on the reliability of biological control. Furthermore, with the model we can identify the characteristics which compose an efficient natural enemy. These characteristics can later be used as evaluation criteria in natural enemy selection programs. In fact ideotypes of natural enemies may be designed, tailored to crop, whitefly and environmental conditions. In that way, a new field of ecological engineering may be explored. The present study already pointed at important selection criteria when comparing different synovigenic, solitary parasitoids showing random search. The model can be adapted for other parasitoids with different foraging strategies or for other natural enemies of whitefly.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • van Lenteren, Joop, Promotor
  • Rabbinge, R., Promotor, External person
Award date18 Oct 1995
Place of PublicationS.l.
Print ISBNs9789054854371
Publication statusPublished - 1995


  • insects
  • plant pests
  • aleyrodidae
  • biological control
  • beneficial insects
  • chalcididae
  • eulophidae
  • trichogrammatidae
  • trialeurodes vaporariorum


Dive into the research topics of 'Understanding biological control of greenhouse whitefly with the parasitoid Encarsia formosa : from individual behaviour to population dynamics'. Together they form a unique fingerprint.

Cite this