Adaptation of the PESTonFARM Model to Support Decision-Making and Planning of Local Implementation of the Sterile Insect Technique in the Control of Ceratitis capitata Flies (Diptera: Tephritidae)

Abstract

The Sterile Insect Technique (SIT) is most effective at large regional scales when applied within an area-wide framework. However, there is a need to investigate its feasibility at smaller scales, e.g., for emergency responses to local fruit fly invasions or planning for preventive release of sterile males in local high-risk zones. Available decision support tools and SIT implementation models are effective for large-scale interventions but tend to ignore the influences of fine-grained terrain structures and therefore offer little guidance for small-scale SIT operations in locally diverse landscapes. This study addresses this issue by adapting a site-specific individual-based PESTonFARM model to simulate both the behaviour and fate of individual members of ultra-small invasive medfly propagules and the post-release dispersal and mating performance of sterile males in heterogeneous and mosaic landscapes. To illustrate model operation, several SIT implementation scenarios were simulated to reveal the influence of local landscape structure on the behaviour of wild and released sterile males and to quantitatively assess the effectiveness of different SIT scenarios. Our results demonstrate the sensitivity of the model and showed that the influence of the spatiotemporal structure of local resources should not be ignored when planning local SIT operations.

1. Introduction

The Sterile Insect Technique (SIT) is an environmentally friendly and technically advanced approach to pest control based on the use of mass-produced sterile live insects to control populations of wild individuals of the same species. Numerous regional and subregional pest suppression and eradication SIT operations have been and are being conducted against a range of insect pests such as screwworms, tsetse flies, tephritid fruit flies, moths, mosquitos and others [1,2]. Of these, medfly, the Mediterranean fruit fly, Ceratitis capitata (Wiedemann), is the target of the largest operations conducted in the US, Mexico, Guatemala, Argentina, Chile, Australia, Spain and other countries [3].

Medfly is recognized as one of the most destructive agricultural pests worldwide. Originally native to sub-Saharan Africa, the species is now cosmopolitan, present throughout the year in most tropical and subtropical regions, and seasonally established in temperate areas [4,5]. Medfly can infest the fruits of more than 300 plant species and lives across a wide variety of climatic zones [6,7]. C. capitata is listed as an A2 quarantine pest by the European and Mediterranean Plant Protection Organization (EPPO). Its economic impact is substantial, prompting some countries, such as the United States (mainland), Australia and Japan, to implement stringent phytosanitary regulations. These measures include strict import controls and quarantine procedures aimed at preventing the introduction and potential spread of this pest [4].

The effectiveness of SIT against medfly is significant when applied on a large, regional scale [8,9], following the recommended “area-wide” approach [10]. The spatial scale of SIT operations ranges from the largest programs in Mexico and Guatemala in Central America [11] to more limited eradication operations after local medfly detections in California, USA [12]. Generally, their ultimate goal is complete or near-complete elimination of medfly from the treated area in order to create low-prevalence or medfly-free areas or to eradicate new medfly incursions into areas declared medfly-free. Very detailed and comprehensive guidelines and protocols are available to guide and support timely medfly detection, SIT implementation, and post-treatment verification of eradication [12,13].

In Europe, however, in the highly fragmented, multi-owner setting typical of the Mediterranean landscape, the implementation of such approaches is difficult. SIT programs, such as in Valencia, Spain, or in the Neretva Valley, Croatia, have not only a more limited spatial scale (4000 hectares in the latter), but importantly, also a different approach, aiming at seasonal medfly suppression rather than its elimination [14,15,16]. Although SIT technology was originally designed for large scales, there is also interest in including it in the methodological repertoire of IPM methods, and trying it at a very local, even farm-like scale [17,18,19]. Recent detections of Bactrocera dorsalis (Hendel) in Europe [20,21,22] have also raised interest in the use of SIT in emergency response to very local invasions or preventive releases in spatially limited and isolated or semi-isolated zones at high risk of invasion, such as around fruit import ports or markets. Therefore, while fully recognizing the numerous limitations of such an approach, a need arose to investigate the feasibility of SIT on a smaller scale.

Consistent with the original concept of SIT implementation, available decision support tools and models focus on interventions at large spatial scales [23] and therefore tend to ignore the influence of fine-grained terrain structures and their impact on the behavior of both wild and released sterile medflies. Consequently, they offer little guidance for small-scale SIT operations in locally diverse landscapes. This study aims to address this problem by developing a species- and site-specific model that could simulate the post-release behavior, dispersal, and mating performance of released sterile males in a heterogeneous and mosaic landscape, and thus serve as a decision support system (DSS) to guide the planning of local SIT interventions.

The DSS was developed based on the existing PESTonFARM model [24,25], validated on fruit producer farms [26,27]. The model is species-specific and simulates behavior and fate of large cohorts of medfly females in a heterogeneous landscape and their response to different IPM scenarios. The model works with a spatial resolution of 10 × 10 m and daily temporal resolution. The individual development of each “virtual” female fly is stochastically simulated from the egg up to adult stage, and then its daily events such as oviposition and fruit infestation, local dispersal and shifts among fruits of different phenology, and finally mortality caused by natural factors or IPM treatments. Males are not simulated because they do not damage the fruit, but their presence in a 1:1 ratio and numbers sufficient to fertilize all mature females are assumed. Simulated IPM scenarios may consist of up to several different IPM treatments applied in different farm zones according to user-defined schedules. The model generates results characterizing effects of each IPM treatment, medfly seasonal densities and mortality profiles (eggs, larvae, pupae, and adults), fruit infestation at harvest and net costs and benefits of IPM [27]. The presented DSS, SIT-adapted PESTonFARM model, was developed by designing additional software modules emulating the specific mechanism of SIT treatment, the actions of sterile males released into the field and their interactions with wild females.

The aim of this article is to present the main features of the developed DSS modules emulating SIT and to demonstrate its ability to capture and quantify the impact of relatively small changes in site topography on medfly development and on the effectiveness of different approaches to local SIT interventions. The overall performance of the model was verified by relating the results of simulated SIT application scenarios to general knowledge on SIT. Empirical validation of the developed DSS is ongoing within the framework of Horizon project REACT (grant agreement No 101059523), where a series of site-specific SIT scenarios were prepared for several ‘real’ fruit-growing locations in the Naoussa Valley, Greece. These scenarios will be implemented and evaluated during the 2025 and 2026 seasons and the results will be published separately.

2. Materials and Methods

2.1. SIT Expansion—Model Operation

The SIT-expansion DSS modules simulate multiple release events of sterile male cohorts with seasonally modulated sizes and released according to flexible user-defined schedules. In general, it is possible to simulate two different SIT regimes applied to different farm sectors with different schedules. In each of these two cases, one can choose one of two methods to release sterile males: uniform release over defined farm zones (e.g., using a low-flying drone) and/or “manual” release from the ground at fixed, user-defined points.

Only sterile males that survive in the field to the time of reproductive maturity, actively take part in leks, and effectively compete with wild males for females are of interest for IPM. Thus, for each released cohort, the model estimates the number of “effectively competitive” sterile males, based on user-declared SIT quality parameters (emergence rate, flight ability, survival to sexual maturity, propensity to call and join leks, and mating competitiveness). Then, for such an “effective cohort”, daily post-release dispersal and/or mortality events are randomly simulated for each individual sterile male.

The model simulates the daily presence of sterile and wild males, and mature, mating-ready wild females, in 10 × 10 m farm sectors. For each wild female, the day of sexual maturity is estimated based on assumed temperature patterns, and for that day, the effect of competition between the locally present sterile and wild males is simulated. For each female that has copulated with a sterile male, a possible re-mating with a wild male is stochastically simulated. The females that ultimately did not copulate with wild males and do not participate in further reproduction are therefore counted as “eliminated by SIT”.

2.2. Demonstration of the Model’s Potential for Application as a Decision Support Tool

2.2.1. General Approach

To demonstrate the performance of the model, a series of scenarios was simulated of the application of the SIT method for local control of the medfly at three hypothetical sites located in a fruit-growing region in Naoussa Valley, Central Macedonia, northern Greece, with a resident low-density medfly population. The simulated scenarios were chosen solely for the purpose of illustrating the model’s features and its ability to reveal and quantify performance differences between simulated sites and SIT implementation options.

2.2.2. Fruit Growing Sites

To illustrate the influence of site structure on medfly development and the effectiveness of SIT interventions, three simplified hypothetical fruit growing sites of 25 ha (500 × 500 m) were assumed (Figure 1). Each site contained the same assortment of seasonal fruits: very early (apricots), mid-season (peaches), and late (apples). In sites A and B, the area occupied by different fruits was the same (4.7, 11.6 and 7.0 ha for apricots, peaches and apples, respectively), but as shown in Figure 1, their spatial distribution was very different. Sites B and C had similar spatial distribution of seasonal fruits, but at site C, the share of apricots and apples was reduced and the share of peaches increased (3.2, 17.9 and 2.4 hectares of apricots, peaches and apples, respectively). Fruit species differ in their attractiveness to egg-laying medfly females and in their suitability for the development of eggs and larvae. The model can take these differences into account, but for the presented simulations, a simplifying assumption was adopted that all fruit categories present were equally attractive and suitable. It was also assumed that each fruit-growing site was surrounded by a 50 m buffer zone without medfly hosts and non-host tree canopies (Figure 1). Outside the buffer zone, a similar mosaic fruit-containing landscape was assumed.

2.2.3. Weather Data

Local historical weather datasets (air temperature, wind speed and direction, precipitation, solar radiation, soil temperature and water content) for Naoussa Valley were obtained from the EU Copernicus Era5-Land instrument (European Centre for Medium-Range Weather Forecasts, Reading, United Kingdom), covering the period from 1 January to 31 December 2023. The spatial and temporal resolution of the obtained data was about 9 × 9 km and 1 h, respectively.

2.2.4. Medfly “Starting” Population

To enable easy comparison between scenarios, each simulation started on 1 January and the same, arbitrarily determined initial cohort of 10 wild female medflies arriving on 1 April and randomly distributed across the 25-hectare hypothetical site (0.4/ha) was used.

2.2.5. Quality of Sterile Males

The model estimates the number of effective competitors that emerge from a batch of delivered irradiated male pupae and simulates their post-release survival, dispersal, and mating performance. The estimates are based on user-supplied quality parameters. In the simulations presented, quality parameters, reported in

Table 1. The sterile male quality parameters and their assumed values.

Sterile Male Quality Parameter	Assumed Value	Number of Individuals
PUPAE: base cohort of irradiated male pupae		1000
ADULT males: emergence rate	85%	850
Active fliers at the release time	85%	723
Time to maturity post-release	2 days
Daily mortality rate in the field (assumed 50% survival after 4 days)	20%
Effective fliers surviving in the field until maturity		462
Mature fliers joining wild male calling groups (leks)	90%	416
Sterile to wild mating competitiveness	30%
Effectively competitive sterile males from 1000 irradiated pupae		125
Female re-mating chance (mating with a sterile and then wild male)	5%

with their assumed values, were estimated based on a literature review [28,29,30,31,32,33,34,35,36,37] and discussions with experts experienced in managing large-scale anti-medfly SIT operations in different parts of the world. In all simulations, it was assumed that after emergence, males were kept in the laboratory for 1–3 days, fed sugar and provided with water.

2.2.6. Medfly Behavior and Dispersal

Medfly is most active around mid-day, optimally at air temperature close to 25 °C. Based on the literature [38] and our observations, the required conditions were estimated as follows: air temperature ranging from 15 to 35 °C, sunny conditions (solar radiation above approx. 150 W/m²), no or weak wind (below approx. 8 m/s) and no precipitation or very minimal (below approx. 0.5–0.8 mm/h). In contrast to most other models, the dispersal of individual medflies, although stochastic, is not completely random and is dependent on the topography of the site. The starting cohort of 10 wild females and their progeny that emerged within the site, as well as the released sterile males, was able to disperse outside the simulated area. It was assumed that 40% of these out-migrating individuals return to the site, while the rest of the cohort “vanishes” in the surrounding landscape.

2.2.7. Medfly Monitoring

It was assumed that a set of six traps (baited to attract females, e.g., a three-component PTA lure with putrescine, trimethylamine and ammonium acetate) for medfly detection and monitoring (equivalent to 24 traps per km²) was set over the site and maintained throughout the year. It was assumed that two traps were set in plots containing each phenological fruit category (very early, mid-season and late). The locations of the traps on each hypothetical site are given in Figure 1.

2.2.8. Simulated SIT Scenarios

All simulations covered one annual cycle, from 1 January until 31 December. The following scenarios were simulated for each of the three hypothetical sites (A, B, and C):

• NO IPM—this simulation was used to illustrate the potential of the medfly population in local conditions in the absence of any control measures. The scenario served as a reference (control). The simulated results were also used to determine SIT-opportunity windows—periods and places in which the next cohorts of new SIT-vulnerable generations of adult medflies emerge. The established SIT opportunity windows were then used to set the timing and duration of the simulated SIT releases in all scenarios presented below.
• Three SIT implementation scenarios with varying rates of releases of sterile males that emerged from a batch of 50,000, 100,000 or 200,000 of weekly supplied irradiated male pupae were used. All males released at once, in weekly intervals, were distributed uniformly by low-flying drones over all the plots that contain fruit trees.
• Three scenarios of bi-weekly release—the same numbers of weekly used batches of irradiated pupae (50,000, 100,000 or 200,000) were used, but the “base” weekly batch of pupae was divided into two parts, and the sterile males emerging from each part were released separately at set intervals of 3–4 days (twice a week).
• Spot ground release—three scenarios where the sterile males emerging from 50,000, 100,000 or 200,000 of weekly supplied irradiated male pupae were released once a week on the ground at 100 fixed release points arranged into a grid spaced at approximately 50 m.

3. Results

3.1. The Influence of Site Structure and Role of Seasonal Fruits in the Expansion of Medfly Population in the Absence of Any Control Measures

The contribution of seasonal hosts to population buildup is presented in

Table 2. The influence of site structure and role of seasonal fruits in the expansion of medfly population (number of medfly stages developing in the absence of any control measures). (VE, M and L denote very early, mid-season and late fruit, respectively).

	Site A	Site B	Site C
Eggs
VE apricot	543	371	471
M peach	29,049	20,603	20,262
L apple	945,083	120,421	95,796
TOTAL	974,675	141,395	116,529
Larvae
VE apricot	434	294	365
M peach	24,584	17,432	17,210
L apple	771,713	98,477	78,384
TOTAL	796,731	116,203	95,959
Pupae
VE apricot	192	141	172
M peach	9649	7022	6917
L apple	279,335	36,088	28,478
TOTAL	289,176	43,251	35,567
Adult females
VE apricot	133	75	115
M peach	4293	3045	3066
L apple	52,278	7976	5945
TOTAL	56,704	11,096	9126

, showing a cascade of rapidly increasing numbers of eggs laid, larvae, pupae and adult medfly females emerging in the plots containing subsequent seasonal fruits. As a result, even a very small cohort of a few females, if it arrives in an area with continuous and abundant seasonal availability of fruit and is left uncontrolled, can lead to a population “explosion” towards the end of the season (Figure 2).

Interestingly, our data revealed a significant effect of site structure on the potential for medfly development. The mere rearrangement of plots with existing fruits of different seasonality (change of site A to site B) resulted in a more than five-fold reduction in the peak annual population of flies, from 39,087 to 6,582 females in sites A and B, respectively (Figure 2). This effect can be largely attributed to the significantly fewer females that reached and laid eggs on late apple fruits at site B (Table 2). Further modification of the reduced area covered by very early and late fruits (change from site B to site C) resulted in additional reduction in medfly development (Table 2 and Figure 2). In all three sites, even without any control measures, the infestation of the very early and mid-term fruits was negligible, below 0.4%. However, the infestation of the late fruit varied substantially, from 15.2%, to 3.3% and 7.3% on sites A, B, and C, respectively. Interestingly, although the medfly population at site B was slightly larger, the degree of infestation of late fruit was more than two times lower than at site C, where the share of late fruit was smaller and therefore the medfly pressure at the end of the season was relatively higher.

3.2. SIT Opportunity Windows—Timing and Patterns of Female Medfly Emergence in Successive Hosts

During summer and autumn, successive batches of progeny of the founding cohort of medfly females emerged on successive seasonal fruits. The periods of medfly emergence on the successive fruits follow fruit phenology with a degree of temporal overlap (Figure 3). In the three hypothetical sites, on very early apricot, the first adult female emerged on days 161–164 (mid-June), and the emergence continued until days 188–193 (mid-July). In mid-season peach, adult emergence lasted from days 181–184 (early July) to 262–266 (late September). In late apple, adult emergence lasted from day 241–244 (late August) to 313 (mid-November).

The time when new females emerge determines the temporal window of opportunity for SIT intervention. Newly emerged virgin females will reach their maturity within 5–15 days (depending on temperature) and become ready-to-mate and thus vulnerable to SIT. This means that in our hypothetical locations, the release of sterile males should begin around day 150 (beginning of June) and continue unabated until day 325 (late-November) or slightly longer depending on weather. Such timing was adopted in the simulation of all scenarios of SIT intervention, as discussed below.

3.3. Detection of Founding Cohort Members

Despite the fact that the 25-hectare orchard was supposed to have six traps continuously throughout the year (corresponding to 24 traps/km²), the first single fly was caught only on day 218 (early August). Such irregular catches of single flies continued until mid-September, becoming more regular later, although still very low and reaching a maximum of less than 0.5 females per trap per day in early November (day 308) (Figure 4). During the whole year, 86 females were caught on site A. On the other two sites (B and C), the trends were broadly similar, but the catches were highly erratic and very low, and only 31 and 42 flies were caught in total, respectively (Figure 4).

It should be noted that members of the founding cohort were not detected at any of the sites. Even at site A, with the largest medfly population, the traps detected only the first generation of offspring of the founding cohort. The first very irregular catches occurred almost two months after the new generation of females began to emerge in the apricot and peach plots. Regular catches began only after more than 2800 newly emerged females were already active in the field.

3.4. Outmigration

The total number of females that left sites A, B and C during the year was 3864, 1142 and 1237, respectively. The number of flies leaving a site was an indication of the population density on the different seasonal fruits (Figure 5), and the large emigration from site A was largely determined by the large number of flies (1943) leaving the plots with the more infested late apples.

3.5. The Impact of Site Structure on the Effectiveness of SIT Implementation

For each of the three hypothetical locations, the role of three variables typically considered when designing SIT implementation strategies was simulated: (a) cohort size of periodically released sterile males, (b) release frequency, and (c) release method—either ground releases from user-defined release points or evenly distributed aerial releases using low-flying drones. The results are presented in Table 3 and Table 4 and Figure 6, Figure 7 and Figure 8.

3.5.1. Cohort Size

In the case of uniform aerial release of sterile males (1/week), even in the scenarios with the lowest weekly release (50,000 irradiated pupae), compared to uncontrolled medfly development (NO IPM), the number of wild female medflies at peak was reduced from about 40,000 to 8200 (79% reduction), from about 6600 to 2221 (66% reduction) and from about 4900 to 970 females (80% reduction) for sites A, B and C, respectively (Figure 2, Figure 6, Figure 7 and Figure 8). Even in site A with the highest medfly population, the total number of emerging females during the whole year was reduced by more than fourfold. Similarly, there was a substantial reduction in the number of females surviving until the end of the year, which at a dose of 50,000 pupae per week ranged from 67% to 80% for all three sites (Table 3). Such suppression of the medfly population was sufficient to reduce the infestation of the latest fruit to a level generally acceptable to fruit growers, even in the site with the highest medfly population, where the infestation of late fruit dropped from over 15% to just over 5% (Table 4). Doubling the weekly cohort of sterile males from 50,000 to 100,000 pupae per week further improved medfly suppression at each site, reducing the seasonal maximum and the number of females surviving to the end of the year by between 88% and 96% for all three sites (Figure 8 and Table 3). However, this increase in the number of sterile males released did relatively little to reduce the already low infestations of late fruit (Table 4). The benefits of further increasing the number of sterile males (to 200,000 pupae per week) were virtually negligible.

3.5.2. Release Frequency

Changing the release frequency from once weekly to twice weekly had inconsistent effects on SIT efficacy, with the net effect depending on site structure and medfly population density and the overall weekly cohort size of sterile males released. At the lowest release frequency (50,000 pupae/week) across all sites, splitting the dose and releasing individual fractions twice weekly improved SIT efficacy, not only reducing the number of females emerging from sites during the year but, most importantly, the number of females surviving to the end of the year, from 4544 to 3707, from 1233 to 855, and from 506 to 401, for sites A, B, and C, respectively. However, at higher release rates, the effects of increased release frequency were variable and, although generally small, were in some cases downright negative (Table 3).

3.5.3. Release Method

In contrast, changing the method of releasing sterile males from a uniform aerial release to a ground-based weekly release from 100 fixed points in a 50 × 50 m grid had a clear negative effect, resulting in significantly reduced SIT efficacy and less suppression of the wild fly population. With the exception of the scenario with higher release rates at site C, where the fly population was the lowest, this effect was consistent across all other cases (Figure 7 and Figure 8). In general, to obtain an effect similar to using 50,000 pupae released evenly (aerial release), in the case of point release from the ground, it was necessary to use twice as many pupae, at least 100,000 per week.

It should be emphasized, however, that in none of the locations, even in the best-case scenarios using 200,000 sterile pupae per week, did the local medfly population collapse and become exterminated, and in all cases, varying numbers of females, 499, 147, and 16 at sites A, B, and C, respectively, survived until the end of the year. Furthermore, despite the very low medfly population density at each site, a number of fertile females, i.e., 101, 100, and 111 from sites A, B, and C, respectively, left the treated area and disappeared into the external landscape.

Table 3. Number of females alive at the end of year.

Site	Release Rate	Release Method
		Aerial 1/Week	Aerial 2/Week	Ground 1/Week
Site A	0 (NO IPM)	21,999
	50 k	4544	3707	9654
	100 k	1245	1654	5057
	200 k	499	1039	1935
Site B	0 (NO IPM)	3688
	50 k	1223	855	2374
	100 k	427	143	684
	200 k	311	147	1151
Site C	0 (NO IPM)	2538
	50 k	506	401	817
	100 k	89	251	219
	200 k	44	112	16

Table 3. Number of females alive at the end of year.

Site	Release Rate	Release Method
		Aerial 1/Week	Aerial 2/Week	Ground 1/Week
Site A	0 (NO IPM)	21,999
	50 k	4544	3707	9654
	100 k	1245	1654	5057
	200 k	499	1039	1935
Site B	0 (NO IPM)	3688
	50 k	1223	855	2374
	100 k	427	143	684
	200 k	311	147	1151
Site C	0 (NO IPM)	2538
	50 k	506	401	817
	100 k	89	251	219
	200 k	44	112	16

Table 4. Infestation of the late fruit at harvest (%).

Site	Release Rate	Release Method
		Aerial 1/Week	Aerial 2/Week	Ground 1/Week
Site A	0 (NO IPM)	15.2
	50 k	5.3	4.7	9.8
	100 k	1.5	2.1	5.5
	200 k	0.9	1.8	2.8
Site B	0 (NO IPM)	3.3
	50 k	1.4	1.0	2.7
	100 k	0.6	0.2	0.9
	200 k	0.5	0.3	1.6
Site C	0 (NO IPM)	7.3
	50 k	1.8	1.5	3.1
	100 k	0.4	1.2	0.9
	200 k	0.1	0.4	0.0

Table 4. Infestation of the late fruit at harvest (%).

Site	Release Rate	Release Method
		Aerial 1/Week	Aerial 2/Week	Ground 1/Week
Site A	0 (NO IPM)	15.2
	50 k	5.3	4.7	9.8
	100 k	1.5	2.1	5.5
	200 k	0.9	1.8	2.8
Site B	0 (NO IPM)	3.3
	50 k	1.4	1.0	2.7
	100 k	0.6	0.2	0.9
	200 k	0.5	0.3	1.6
Site C	0 (NO IPM)	7.3
	50 k	1.8	1.5	3.1
	100 k	0.4	1.2	0.9
	200 k	0.1	0.4	0.0

Figure 6. The influence of the size of the released sterile male cohort on the annual population profile of medfly (daily numbers of females present on the site).

Figure 6. The influence of the size of the released sterile male cohort on the annual population profile of medfly (daily numbers of females present on the site).

Figure 7. The influence of the size of the sterile male cohort, the frequency of their release and the release method on the annual population profile of medfly in three hypothetical sites, A, B, and C with daily numbers of females present on the site, where 50 k, 100 k and 200 k denote 50,000, 100,000 and 200,000 pupae, respectively, delivered weekly.

Figure 7. The influence of the size of the sterile male cohort, the frequency of their release and the release method on the annual population profile of medfly in three hypothetical sites, A, B, and C with daily numbers of females present on the site, where 50 k, 100 k and 200 k denote 50,000, 100,000 and 200,000 pupae, respectively, delivered weekly.

Figure 8. The influence of the size of the sterile male cohort and the method of their release on the reduction in the annual medfly population peak.

Figure 8. The influence of the size of the sterile male cohort and the method of their release on the reduction in the annual medfly population peak.

4. Discussion

4.1. Study Area

Naoussa Valey in northern Greece, with its established and stable, very low-density medfly population, provides a convenient setting for both in silico simulations and empirical studies of the initial invasion processes and the possibilities of using SIT for local control of low-density medfly populations. The Valey constitutes the northern part of the medfly range in Greece. Although weather conditions are favorable in summer, cold winters significantly reduce the survival of overwintering individuals, reducing the spring cohort to only a few females per hectare, which keeps the annual population low compared to more climatically suitable sites in central or southern Greece [39]. The situation in which the medfly population regenerates from a few spring individuals resembles the early stage of invasion, where a small invasive propagule initiates the development and establishment of the medfly in a new area. The extremely low density of the seasonal medfly population (below 0.5 female per trap per day during the peak season) creates an opportunity to use SIT as one of the IPM options, or even consider using it as the sole method of medfly control.

To demonstrate the model’s performance and illustrate its sensitivity, we assumed only small variations in the topography of the three hypothetical locations, but maintained the same, near-optimal seasonal fruit assortment, ensuring an abundant and continuous supply of attractive fruit for egg-laying females and larval development. The assumed seasonal fruit composition, with 14–20% very early apricots, 50–76% mid-season peaches and 10–30% late apples, reflects the typical fruit structure in the Naoussa Valley. To mimic the seasonal patterns of medfly population in the Naoussa Valey, all simulations assumed that initial cohorts of only 10 female flies arrived in early April at each of the three hypothetical locations.

4.2. Performance of Sterile Males After Their Release

The success of the SIT operation depends crucially on the behaviour and quality of the released sterile males [3]. Their pre-release quality is routinely tested as part of the monitoring of the mass production process. However, these tests, conducted in laboratory conditions or in field cages, although valuable, do not fully reflect the performance of sterile males after their release under more challenging field conditions. Surprisingly, empirical data on the performance of males under realistic field conditions are scarce. Estimates of the fraction of active flyers emerging from pupae batches range from about 50% [30] to over 70% reported in our discussions with operators of large mass-rearing facilities. Post-release survival and dispersal distances can be inferred from the results of numerous recapture studies [28,29,30,31,32,33,34]. In general, the analysis of available data indicates that at least 50% of males would survive three, and possibly four, days after release. The range explored by most released sterile flies does not exceed 100–200 m, and more than 90% do not venture further than 400–500 m from the release point [32]. Released males actively follow the distribution of locally available resources and even when released from evenly spaced points, their aggregation areas tend to mirror those of wild males [19]. The least studied aspect concerns the interactions and mating competition between sterile and wild males under field conditions. It is known that in field cages or semi-field conditions with a rather basic canopy of potted or pruned fruit trees, sterile males tend to function quite well and, if present in equal numbers to wild males, can achieve 20–25 or even 35% of matings [36,37]. However, in a field study of strain Maui-03, Shelly and Whittier [35] reported that wild males had a mating rate of 91%, although in leks formed by sterile and wild males, the latter contributed only 19%. Overall, for strains like Maui-97, it can be estimated that the percentage of successful males from batches of irradiated pupae is about 12–13% in field cages and about 4% in the field [36]. For the improved and now widely used Vienna-8 strain, these numbers have been estimated to be about two to three times higher. In most cases, copulation with sterile males causes the female not to mate again [40], although cases of re-mating with another male, wild or sterile, do occasionally occur [41]. The frequency of such events for the Vienna-8 strain used in our simulations has been estimated at about 5%.

The parameters that were used in the presented simulations were estimated based on the analysis of available and published empirical data and discussions with experts involved in large-scale SIT operations. However, the performance parameters of sterile males are not fixed in the model and can be set by the user to reflect the quality of the simulated medfly strain.

4.3. Individual-Based Modelling Approach and Outline of Specific Features of the SIT-Enhanced PESTonFARM Model

Consistent with the original concept of SIT implementation, available decision support tools and models focus on interventions at large spatial scales on the order of several or thousands of square kilometers that cover the area of most or all of the targeted medfly population [42]. Typically, such models are based on deterministic mathematical equations or algorithms that estimate the effects of various aspects of SIT intervention on the generalized growth of the target pest population [23]. Although they answer general questions about SIT and are useful in planning large-scale operations, they tend to ignore the influence of fine-grained terrain structures and their impact on the behavior of both wild and released sterile medflies [23]. Therefore, for small-scale SIT operations in locally diverse landscapes, such an approach is not detailed and site-specific enough. It should be emphasized that a very local spatial scale, of the order of several or several dozen hectares, corresponds approximately to the area of average lifetime activity and dispersal of individual medflies [25,32]. The above seems to suggest that in such a situation, it may be more appropriate to approach the problem from the perspective of the individual insects. Indeed, in recent years, several agent- or individual-based models simulating the behaviour of individual medflies have been developed [24,25,27,43,44]. Although all of these models focus on reproducing the behavior and development of individual insects, they differ significantly in the extent to which they take into account local landscape features and generalize fly behavior.

Among the distinctive features of the presented SIT-enhanced PESTonFARM model is its ability to reflect the details of medfly-relevant terrain features, either by generating hypothetical landscapes of varying complexity, or by converting terrain orthophotomaps into digital mosaics showing the structure (size, density and distribution) of tree crowns and the distribution of plots with fruits of different phenology. Moreover, in the simulation of the exploratory behavior and dispersal of individual insects, the model does not use simplifications commonly used on other models, such as particle diffusion and random or correlated random walks algorithms [42,44,45,46]. Instead, the simulation of the daily exploration and movement of each individual fly is stochastically dependent on the current, daily changing pattern of features of the currently occupied and neighboring niches (10 × 10 m sectors), as well as on the individual’s age and the current ambient weather conditions [24,25]. Such an approach greatly increases the realism of the simulation of the movements of individual flies operating in a heterogeneous environment and their shifts between plots with fruit of various phenology.

For local SIT implementation, the presented model allows for the simultaneous simulation of daily events, distribution and interactions of individual wild females and the released sterile males. Importantly for the end user, the model estimates the seasonal distribution of medfly development stages (eggs, larvae, pupae and adults), their mortality caused by applied IPM measures (including SIT), their costs, and fruit infestation at harvest, separately for each IPM measure and for each fruit type.

Our goal was to present only the SIT-related model components that were added to the earlier PESTonFARM model, which already had the ability to simulate complex IPM schemes consisting of several different methods [26]. The SIT-enhanced PESTonFARM model “inherited” these capabilities, with the additional ability to include SIT in the repertoire of IPM methods. The model can be used, among other things, to develop more complex and realistic locally adapted scenarios for mixed implementation of SIT and other IPM methods, and indeed, it is currently being used for this purpose within the REACT project. The results of such applications will be published separately.

4.4. Detection of Founding Cohort Members

SIT emergency response operations are initiated upon the detection of a new invasion. Our simulations indicate that even when assuming a relatively dense network of surveillance traps (6 per 25 ha, which corresponds to 24 traps per km² or 54 traps per mile²), the chances of detecting members of the original founding cohort are extremely low or negligible. In practice, it is the first or even second generation of the offspring that can only be detected several months after ‘invasion’, in late in summer or early autumn, when medfly numbers have already increased.

Such a result is not unusual and remains consistent with empirical studies on medfly detection [47,48], as well as with the extensive experience of medfly monitoring in the northern parts of their Mediterranean range, such as northern Greece or Croatia. Despite the year-round presence of established medfly populations in the area, it is normal to detect the first flies in late summer, with usually no catches during the first half of the year [49,50,51].

4.5. Determining the Spatial and Temporal SIT Opportunity Window

It is important to bear in mind that at a local scale, the way the SIT method works is quite different from typical IPM methods such as cover spray pesticide application, mass trapping or bait sprays. The operation of the latter is spatially limited to the areas of their application, and medflies stay susceptible to them throughout their lifetime, regardless of age and mating status. In contrast, the area of activity of the released sterile males is less defined because they disperse freely after release and, like wild individuals of the target population, follow the seasonally changing spatial distribution of available resources [19]. Typically, large SIT operations are guided by medfly monitoring, which shows the presence and density of flies in the treated area [10]. However, it is not the number of flies present that determines the optimal timing of SIT intervention, but their physiological state. Practically, only newly emerged wild female flies are briefly susceptible to SIT. When planning SIT operations, it is important to remember that SIT releases will have no effect either before or after the time at which new females emerge and reach maturity.

Understanding the relationship between windows of susceptibility to SIT at the individual insect and population levels and its relationship to the phenology of locally available fruits is crucial for the implementation of SIT. Newly emerged, wild medfly females become sexually mature and ready to mate several days after emerging from the soil [52]. In general, most females mate only once, although infrequent re-mating occurs [40,41]. Thus, apart from these rare cases, for individual females, their temporal “window of vulnerability” to SIT is limited to just one day, at which point the female chooses to mate with either a wild or sterile male. For the population as a whole, consisting of many cohorts of females of different ages, the length of the temporal window for SIT operations is much longer. Its length is determined by the emergence time of new cohorts of virgin females, which in turn depends on the seasonal continuity of the availability of fruits suitable for larval development [51]. Although fruits are usually most attractive near ripeness, they can be infested at or even earlier than their “green maturity stage” and in some cases, at the very early, fruitlet stage [53]. Although not at their optimal suitability, such unripe fruits can still support some larval development. This means that for each fruit phenological category, the time window in which medfly can develop is much wider than the fruit ripening period preceding harvest.

In the highly fragmented mosaic landscapes typical of European fruit growing [54,55], the long lifespan and considerable mobility of wild medflies allows individuals to move between different plots during the season according to the phenology of local fruits [51,56,57,58,59,60]. Consequently, the exact locations (specific plots) where cohorts of successive generations emerge, where individual females disperse and mate, and where they lay eggs and cause fruit damage, are often different and seasonally variable. Under such circumstances, ensuring temporal and spatial convergence between wild and briefly SIT-vulnerable ready-to-mate females and short-lived sterile males, which is crucial for SIT success, is challenging and requires precise planning.

The model reflects these processes by simulating the development of both medfly and fruit development according to local weather conditions, as well as seasonal movement of medfly between plots containing fruits with different phenology. This allows the model to generate locally relevant information on the optimal timing and spatial targeting of local SIT activities.

4.6. Model Sensitivity—Ability to Capture and Quantify the Impact of Even Minor Differences in Terrain Topography on the Development of Medfly and Performance of SIT

To highlight the role of the spatial structure of the local terrain on the development of medfly population, and to demonstrate the model’s ability to reflect and quantify such interactions, all three sites were assumed to have a similar structure and identical seasonal fruit assortment. They differed only in the spatial arrangement of fruit plots or in a relatively small change in the ratio occupied by different fruit phenological categories. Our simulations showed that even such small structural differences between sites severely affected the development and ultimately the size of the seasonal medfly population. Simple spatial rearrangement and consolidation of existing fruit plots into phenological blocks, in which mid-season and late fruits were separated by the zone with the earliest fruits, severely hampered medfly development. During the critical period of early spring, this reduced the chances of the few randomly scattered medfly founders to locate the earliest fruits and forced their progeny that emerged there to move and disperse far within bigger (consolidated) mid-season fruit block. Most importantly, later on, the fruitless and less attractive block of the earliest fruiting trees served as a buffer and hindered the medfly movement from the mid-season fruit to the less accessible and more distant late fruits. As a result, late fruits, which typically develop the largest populations in the annual cycle and are the main target of medfly attack, were less infested. Further modification of the reduced area occupied by very early and late fruits, although small, also had a measurable effect on medfly development.

In addition to estimating target pest densities and determining temporal and spatial focal zones, designing a local SIT also entails decisions about the size of the sterile male cohorts to be released, and the frequency and method of release (e.g., uniform aerial vs. ground points). Although it seems intuitively obvious that increasing the number, frequency, and uniformity of male releases should improve the efficiency of the operation, quantifying the location-specific benefits in relation to the incurred effort and costs is not.

The presented model reflects the influence of the local landscape structure on the behaviour of both wild females and released sterile males and allows for a quantitative estimation of the performance of different SIT scenarios. Our results confirmed the sensitivity of the model and its ability to detect, discriminate and quantify the impacts of relatively minor changes in local topography. Our simulations revealed that even seemingly small alterations in the spatial organization of the landscape and fruit arrangement can significantly affect medfly development and cause the performance of different SIT scenarios to change in a way that is not always consistent with intuitive expectations. It also showed that when planning local SIT activities, challenges should not be assessed solely on the basis of the general presence of host fruits, and that the influence of the spatial arrangement of local resources can be significant and cannot be ignored. The results generated by the model allow for informed decisions to be made, selecting of the most promising scenario and optimally tuning it to the local fruit and landscape structure, and site-specific weather conditions.

4.7. Stochasticity of Very Small Invasive Propagules

Finally, it should be clearly emphasized that in the case of an extremely small founding cohorts (overwintering flies or invasive propagules), stochastic effects have a very significant impact on both the fate of its individual members and on the final effects of its activity. The fundamental uncertainty of the fate of a small cohort of founders is well known from ecological studies [61] and is also reflected in the variability of the results generated by successive model runs. The PESTonFARM model stochastically simulates daily events such as dispersal, egg laying, mating with wild or sterile males, mortality due to various natural and IPM-related factors independently for each individual virtual insect. The final outcomes such as annual population size, annual pattern and survival, fruit infestation, etc., are not programmed but actively generated by the daily stochastic activities of the virtual insects during successive model runs. This means that repeated runs of the model with unchanged assumptions and parameters generate a stochastic equivalent (not identical), but only one of the possible instances of the simulated scenario. For larger populations, typical of IPM applications, there is very little difference between successive model runs. However, for ultra-small established populations or in early invasion cases, when the founder propagule is extremely small (in our case, only 10 females), the final results may differ from run to run. In such a situation, repeated simulation runs reveal a spectrum of possible outcomes, while the most likely final effect will be close to the intermediate results. Such cases were used to illustrate the relationships and trends presented in this article.

4.8. Different Perspectives and Measures of Success of SIT Operation

Remarkably, from a farmer’s perspective, nearly all simulated SIT-only scenarios were successful. Even in location A with the relatively highest medfly population and with the lowest sterile male release rate, significant reductions in medfly population were achieved, which translated into reduced fruit infestation to levels acceptable to farmers. Such results are consistent with published cases of small- to medium-scale SIT application in 2.5 ha experimental plots in citrus plantations in Morocco [18], 18 ha apple plantations in Uruguay [17] and 1000 ha citrus (mandarin) areas in Croatia [14]. It is important to emphasize that the 50 m buffer used in our simulations was intended to provide minimal separation of the treated site and was not expected to serve as an effective barrier to prevent the movement of flies into or out of the SIT-treated area. However, our simulations indicate that in areas with very low fly density, even this minimal buffer may be sufficient for IPM applications where the goal is simply to seasonally suppress (but not eliminate entirely) medfly population to reduce fruit infestation.

For more demanding applications, such as eradication or maintaining fly-free zones, much wider buffers, on the order of 2 km or more, are needed [42]. Indeed, our simulations have shown that achieving goals such as eliminating or containing local invasions of medfly has proven more problematic. Although the starting cohort and seasonal medfly population in each site were very small, even assuming the release of a relatively large number of males (200,000 sterile pupae per week in an area of 25 ha) did not lead to the complete eradication of the medfly in any of the hypothetical sites. Moreover, the simulated SIT interventions, although they caused significant population suppression, were not able to prevent the spread of descendant medfly generations outside the area of the treated location.

Such a result is not unusual for short (seasonal, approximately 6 months) releases of sterile males, without support of other IPM methods. In an eradication program conducted in Australia on a relatively small area of 25 km², the release of 7.5 million sterile flies per week (which corresponds to 75,000 males per 25 ha site) for 15 months, without the support of chemical methods, was insufficient to affect and suppress the medfly population. Eradication was achieved only after releasing 12 million sterile flies per week for 34 months, combined with intensive use of other control measures, such as spraying of hydrolysed protein and trichlorfon baits, pesticide cover sprays and fruit stripping [62].

Our results do not exclude the possibility of obtaining positive results in specific cases but strongly suggest the need for very careful and pragmatic selection of locations where the specific and medfly suboptimal spatiotemporal configuration of host fruits, terrain topography and weather conditions might offer a chance for success in the application of SIT alone, without the usually needed heavy support of pesticide sprays and other methods. The use of the developed modelling DSS tool enables rational decisions about site selection and optimisation of the response to local conditions.

5. Conclusions

• The SIT-enhanced PESTonFARM model reflects the influence of landscape structure on the behaviour of both wild female flies and released sterile males, and enables quantitative assessment of the effectiveness of different SIT scenarios.
• The model is sensitive enough to distinguish and quantify the effects of rather small spatial changes in local topography and fruit structure on medfly development and the effects of different SIT implementation approaches.
• The model simulates the development of both flies and fruits according to local annual weather patterns and can therefore generate locally relevant information on the optimal timing and spatial focus of local SIT operations.
• The use of the model enables informed decision-making and design of SIT implementation scenarios according to local conditions.