Deconstructing the eradication of new world screwworm in North America: retrospective analysis and climate warming effects

Biology and migration of screwworm

Males have a promiscuous mating behaviour (polygyny), whereas females mate only once. This mating biology was a key factor in the eradication success because matings between sterile males and wild females yield non‐viable eggs (Knipling, 1955). Adult male flies feed at flowers and live for 2–3 weeks, whereas adult females live approximately 10 days on average, feeding on serous fluids at animal wounds and decomposing animals (Thomas & Mangan, 1989; OIE, World Organisation for Animal Health, 2013). Approximately 3–4 days after mating, female flies begin to seek wounds on vertebrates to lay large batches of eggs but, because the species is autogenous, females can complete two or three vitellogenic cycles without a protein meal (Crystal, 1966). Under field conditions in Central America, the half‐life of mated wild female is 3.7 days, with a mean age at wounds of 7.5 days and a maximum age of 21 days (Thomas & Chen, 1990).

Screwworm females are attracted to wounds that may be as small as those caused by the feeding of the invasive cattle tick Rhipicephalus (Boophilus) microplus (OIE, World Organisation for Animal Health, 2013), which has periodic outbreaks in Mexico and south Texas (Pérez de León et al., 2012). Feeding by screwworm larvae expands the wound (myiasis) attracting further oviposition and, if not treated, this may cause the death of the animal.

The effects of temperature

Screwworm is a cold intolerant species that has high lower and upper developmental thermal thresholds (14.5 and 43.5 °C, respectively) with the optimal temperature for survival and adult reproduction being approximately 27.5 °C (data from Adams, 1979; Berkebile et al., 2006). The egg and larval stages develop on/in the host at host body temperature and, at maturity, the larvae drop to the ground to pupate. Pupae and free‐living adults experience near ambient temperatures.

The daily mortality rate of pupae and adults [µ_adult(T)] per day at different temperatures (T) was captured by a symmetrical convex function fitted to laboratory data (Adams, 1979; Berkebile et al., 2006) (Eqn (1), see Gutierrez & Ponti, 2014).

In the model, we focus on the limiting effects of average ambient daily temperatures T < 27.2 °C that reduce reproduction and survival, especially during the critical autumn–winter period. Temperatures of T ≥ 27.2 °C in summer also affect the fly demographic rates, although such temperatures are generally not limiting (Gutierrez & Ponti, 2014). To capture the limiting effects of colder temperatures, the daily values of µ_adult(T < 27.2^°C) were summed over the year [i.e. µ_cold(y)] (Eqn (2) for each of the 20 335 lattice cells (25 km²) across the continental U.S.A. and Mexico during the period 1975–2005.

Gutierrez & Ponti (2014) estimated that an average value of ${\overline{\mu }}_{\text{cold}}\approx 10$ was a good metric defining the geographical limits for screwworm year round survival (endemicity) that accorded well with field observations. Using this metric, cold weather on average restricts the area of fly endemicity in North America to south Texas and much of Florida, a broad north‐south band of eastern Mexico (the states of Nuevo Leon, Tamaulipas and Veracruz), the Yucatan Peninsula, and tropical south Mexico (Area I), as well as to Baja California and western coastal Mexico (Area II) (Fig. 1A). Higher elevations of Mexico may also be outside of the favourable zone of endemicity (Fig. 1B). We designate the northern reaches of Areas I and II as transition zones of fly persistence.

Adult dispersal and migration

Mated adult females have high disperal capacity, whereas males exhibit an aggregation–wait station behaviour that greatly limits their dispersal (Krafsur, 1978). In the tropics, fly numbers are consistently higher in forest habitats (Phillips et al., 2004). In warm humid areas with a high density of animals, dispersal of females is approximately 3 km, although a dispersal of 10–20 km has been reported, with distances of 300 km being documented in arid environments with dispersal aided by prevailing winds that may enable flights across open water (Barrett, 1937; Deonier, 1942; Hightower et al., 1965; Mayer & Atzeni, 1993; Skoda et al., 2017).

Historically, the first cases of myiasis in the U.S.A. were reported in south Texas (and south Florida) during mid‐winter and, during some years, the infestations spread northward in Texas and neighbouring states but subsided in the autumn as temperatures cooled [Eqn (2) (Baumhover, 2002; Comis et al., 2012)]. Reinvasion of non‐endemic areas of the SW U.S.A. during the late spring to autumn period is assumed to be enhanced by NAMS winds (Fig. 2) (Gutierrez & Ponti, 2014). Arias et al. (2012) classified NAMS as being wet or dry based on weather in a rectangular area of Northwest (NW) Mexico (20°N 107°W, 22°N 101°W, 32°N 106°W, 30°N 112°W) (Fig. 1A). Although less impressive than the Asian monsoons, wet NAMS regimes (storms) produce northward surges of relatively cool, moist maritime air from the eastern tropical Pacific into the SW U.S.A. via the Gulf of California (Fig. 2A), although some wet NAMS may also draw atmospheric moisture from the Gulf of Mexico (Fig. 2B) (Carleton et al., 1990; Douglas et al., 1993; Stensrud et al., 1995; Adams & Comrie, 1997; Wright et al., 2001; Higgins et al., 2004; Vera et al., 2006). Wide variations of NAMS occur on intra‐seasonal and decadal time‐scales and are related to the frequency of gulf surges and the latitudinal position of the subtropical ridge during the late June to September period. The Julian dates (from 1 January) for the onsets and retreats of NAMS and the associated total amount of monsoon rainfall (mm) during 1948–2010 are shown in Fig. 3A (data from Arias et al., 2012). Wet NAMS have early seasonal onset and late retreat, and the prevailing winds are anticyclonic and generally flow in a northward direction. By contrast, dry NAMS have late seasonal onset and early retreats, low rainfall, and prevailing winds are often in a southward direction (Fig. 2C). We use total rainfall in the reference area as a surrogate metric for the strength of annual NAMS, as well as for the dominant direction of storm wind flow favourable for fly movement.

Myiasis data

During the SIT programme, larval samples (cases of myiasis) submitted by stockmen during the eradication period (1962–1983) were used to document the geographical distribution and relative levels of myiasis, as well as to target sterile fly releases (E. S. Krafsur, personal communication). Totals of myiasis cases reported in the SW U.S.A. during the 1962–1980 eradication period and the total annual number of sterile insects of both sexes released are shown in Fig. 3B (Novy, 1991; Gutierrez & Ponti, 2014) (see also the ratio data in Fig. SM1 in the Supporting information, File S1). Approximately 96% of the cases were reported from Texas, and hence this area is a major focus of the present study. The myiasis data implicitly include weather and migration effects, stockmen reporting bias, and the effects of SIT releases of hundreds of millions of adult sterile males and females. The myiasis data are monthly county level totals, georeferenced to the county seat, and are mapped as yearly log₁₀(total cases of myiasis + 1). Data for the 1962 and 1972 outbreaks were further summarized on a weekly and monthly basis respectively, and were mapped to illustrate the time development of the infestations across Texas. The myiasis data are available on request in excel format (Microsoft Corp, Redmond, WA, USA).

Weather data

Observed daily weather data (maximum–minimum temperature and rainfall) for McAllen, Texas for the 1 January 1942 to 31 June 2017 were obtained from the Global Historical Climatology Network – Daily (GHCN‐Daily, Version 3) (Gutzler et al., 2005; Menne et al., 2012a, 2012b; Arias et al., 2012; see supplemental materials, File S1). Daily maximum–minimum temperature and rainfall data for the historical period 1975–2005 and for the future period 2045–2075 at each of 20 355 lattice cells of approximately 25 km² (spatial resolution 0.25°) across the U.S.A. and Mexico were obtained from the National Aeronautics and Space Administration (NASA) Earth Exchange Global Daily Downscaled Projections (NEX‐GDDP) dataset (Thrasher et al., 2012; https://nex.nasa.gov/nex/projects/1356/). The NEX‐GDDP dataset includes global downscaled climate scenarios that are derived from the General Circulation Model (GCM) runs conducted under the Coupled Model Intercomparison Project Phase 5 (CMIP5) (Taylor et al., 2012). Specifically, we use climate model data from the Max Planck Institute Earth System Model low resolution (MPI‐ESM‐LR) GCM, forced by the Representative Concentration Pathway 8.5 (RCP 8.5), a scenario of high greenhouse gas emissions relative to other RCPs (Riahi et al., 2011), as downscaled in the NEX‐GDDP dataset. Sheffield et al. (2013) evaluated historical simulations of North American climate in CMIP5 experiments using continental metrics of bias relative to weather observations and showed that MPI‐ESM‐LR is the top ranked among the core set of 17 GCMs considered, with a particularly good performance in terms of temperature. The NEX‐GDDP dataset also provides a set of global, high resolution, bias‐corrected climate change projections for use when assessing climate change impacts on processes that are sensitive to finer‐scale climate and local topography, including biological processes of poikilotherm organisms such as screwworm. Downscaling (i.e. increasing the spatial resolution) of GCM output addresses two primary limitations: the relatively coarse spatial resolution of most GCMs (e.g. hundreds of km) and their statistical bias compared with observations (Thrasher et al., 2012).

Analysis of myiasis outbreaks in Texas

Despite massive releases of sterile flies, five peaks of myiasis of varying intensity occurred during the 1962–1982 eradication period: 1962, 1968, 1972, 1975–1976 and 1978 (Fig. 3B). A small outbreak in 1978 was mostly a result of cases from Arizona and New Mexico. Prior analysis of the same myiasis data (Gutierrez & Ponti, 2014) suggested that cold weather and total rainfall at McAllen, Texas during the previous year (y − 1) could be used as predictors of screwworm myiasis outbreaks in Texas during the current year (y).

In the present study, log₁₀ cases of myiasis(y) in year (y) were regressed on (a) the dates in five‐day intervals (pentads) of the onset and retreat of NAMS; (b) their duration; (c) total rainfall (mm) in the NW Mexico reference area (data from Arias et al., 2012) (Fig. 2A); (d) yearly estimates of µ_cold at McAllen, Texas, as a measure of temperature conditions in the north transition zone (Area I); and (e) yearly log₁₀ sterile insects released. We note that total rainfall in NW Mexico during the eradication period is correlated to the duration of NAMS in pentads (r ² = 0.71) (Fig. 4A), although the relationship for the 1948–2010 period is weaker (r ² = 0.56).

Multiple linear regression analysis (Eqn 3) found that only lagged values of µ_cold (y − 1) at McAllen in the transition zone, and total rainfall (y − 1) in the NW Mexico reference area were significant predictors of log₁₀ cases of myiasis(y). None of the interaction terms were significant, and the correlation between µ_cold (y) and total rain (y) was weak (r ² = 0.053). This regression model has the added advantage (Gutierrez & Ponti, 2014) in that rainfall values suggest the strength of the NAMS and the direction of the prevailing monsoon winds (Arias et al., 2012).

log₁₀ myiasis(y) declines with increasing µ_cold(y − 1) (P < 0.01) and increases with total rainfall (y − 1) (P ≈ 0.05) (Fig. 4B,C).

Substituting values of cumulative µ_cold(autumn–winter) computed for the period 1 September to 31 May that bridges 2 years (i.e. y − 1 and y) for µ_cold(y − 1) in the multiple regression yielded a low predictive value (r ² = 0.287), with only µ_cold(autumn–winter) being significant (P < 0.05). This suggests that screwworm buildup occurs in Area I during the prior year (y − 1) and that conditions during the preceding winter period simply set the stage for outbreaks during summer in Texas in year y.

Favourability of weather for myiasis outbreaks during 1948–2010

The data used in estimating Eqn (3) include the effects of weather, the vagaries of stockman bias and the non‐significant effects of log₁₀ sterile insect released, and may be viewed as an ecological niche model for screwworm outbreaks in Texas. Using Eqn (3), we examined the favourability of weather for myiasis outbreaks in Texas before (1948–1961), during the SIT eradication period (1962–1982) and after (1983–2010). Observed and predicted total myiasis are shown in Fig. 5(A). During 1948–2010, average ${\overline{\mu }}_{\text{cold}}$ at McAllen was 7.92 ± 1.28 per year and average rainfall in the NW Mexico reference area was 58.73 ± 24.82 mm/year. Annual deviations from these averages (i.e. Δµ_cold and Δrain) are depicted as histograms in Fig. 5B,C, with deviations favourable for screwworm outbreaks indicated in black.

Eradication of screwworm in tropical areas of Mexico

Myiasis data from Mexico are not available, and Eqn (3) for Texas cannot be used to examine the favourability of weather for myiasis outbreaks in this region. Instead, we used an age‐structured, physiologically‐based demographic model (PBDM) (Gutierrez & Ponti, 2014) to capture the fly's weather driven dynamics and to explore SIT eradication of the fly in Mexico. The bio‐demographic functions for development, survival and reproduction in the PBDM depend on temperature, and additionally oviposition site density for reproduction (see below). Although Thomas (1993) asserted that ‘… evidence is lacking that laboratory‐adapted strains are competitively impaired in the field …, laboratory adapted strains clearly deviate from the wildtype’ questions arose whether reproductively isolated population could impede the progress of the eradication campaign (Richardson et al., 1982), whether sterile flies were competitive with wild ecotypes (Bush et al., 1976) and whether aerial fly release procedures were efficient (Krafsur & Garcia, 1978; Krafsur, 1987). Lachance et al. (1982) found that allelic frequencies were similar in wide geographically separated populations and no genetic mating incompatibilities were found. Data to estimate the effects of competitiveness and release procedure are not available, although a very rough estimate of their combined effect can be made heuristically by comparing observed field release rates with base PBDM predictions of sterile fly release rates required for eradication, assuming sterile flies have the same mating competitiveness, and the release procedures are non‐limiting.

Absent SIT effects, the number of fertile eggs (ΔE(t,T)) deposited by all fertile mated females (W _m) and newly mated females (0.5W _u) at time t (i.e. day) is computed using Eqn (4) (Gutierrez et al., 2014).

Simplifying the notation:

Only half of virgin females are mated per day (Krafsur, 1985), and the average maximum per capita reproductive rate per female per day R is approximately 67 eggs/days. This potential reproduction is corrected for sex ratio (sr = 0.5) and scaled by concave symmetrical functions for the effects of temperature on reproduction (0 ≤ ϕ_T(T(t)) ≤ 1, 14.5 °C ≤ T ≤ 43.5 °C; estimated from Thomas & Mangan, 1992) and for adult survival (0 ≤ ϕ_lx(T) < 1) (Gutierrez et al., 2006). A ratio‐dependent functional response model ($0\le {\varphi }_{\text{search}}\left(t,T\right)=\left(1-e^{-0.0001\mathrm{\Delta t}\left(\mathrm{T}\left(\mathrm{t}\right)\right)·\mathrm{H}/{\mathrm{W}}_{\mathrm{m}}\left(\mathrm{t}\right)}\right)<1$) (Gutierrez & Baumgärtner, 1984) was used to estimate the success of fertile females in finding wounds (H km⁻²) given a low search rate (0.0001 per dd) where Δt(T(t)) is physiological time in degree days > 14.5°C (dd) on day t. In the field, H varies in time and space in unknown ways (Matlock et al., 1996; Matlock & Skoda, 2009) and hence a constant (H = 100) was used for comparative purposes.

SIT affects only unmated females and the effects enter the model Eqn (4) as scalar functions for mating competitiveness (0 < ϕ_comp < 1) and, as a linear scalar (0 < ϕ_release < 1) for the relative efficacy of sterile fly releases for mating with virgin females (Eqn 5). The ϕ functions can be viewed as survivorship terms.

Specifically, 0 < ϕ_comp = W _♂/(cS _♂ + W _♂) ≤ 1 is the proportion of W _u mated by wild type males (W _♂) competing with sterile males (S _♂) with coefficient 0 < c ≤ 1 being the competitiveness of S _♂ relative to W _♂ (Krafsur, 1994). Hence, 0.5ϕ_compW_u(t) is the number of W _u × W _♂ matings, and 0.5(1 − ϕ_comp)ϕ_releaseW_u(t) is the number of W _u × S _♂ matings removing females from the reproductive pool. The remaining unmated females is W_u(t + 1) = 0.5W_u(t) + ΔW_u(t), where ΔW_u(t) is the new females emerging from pupae during t.

Estimating the efficacy of SIT

The daily population dynamics of screwworm without and with SIT were simulated at four distinct ecological locations on a north–south transect: Uvalde, Uvalde County and McAllen, Hidalgo County, Texas, and at Tampico, Tamaulipas and Tuxtla‐Gutierrez, Chiapas, Mexico (Fig. 6). Observed daily weather at each location was used to drive the model: years 1961–1980 for Uvalde and McAllen and 1990–2000 weather for Tampico and Tuxtla‐Gutierrez. As initial conditions, H = 100, and 0.25 individuals of both sexes of each life stage km⁻² were used, yielding the relative levels of flies observed in the field (Matlock et al., 1996). Simulated pupal numbers was used as the metric of fly density.

The efficacy of SIT was estimated heuristically at the four locations by comparing observed field release rates to PBDM predictions of release rates required for eradication assuming no detrimental effects on sterile flies. Specifically, W _u ×   S _♂ mating is assumed to be frequency dependent (i.e. 0 < ϕ_comp = cS _♂/(cS _♂ + W _♂) ≤ 1, given c = 1) and the released sterile flies are assumed to be optimally placed (ϕ_release = 1) in proximity to wild virgin females. The bi‐weekly release rate of sterile flies was varied in the different simulations until all screwworm life stages decline to zero.

Uvalde, Texas is located north of the transition zone, and has historically experienced high incidence of myiasis (e.g. Figs SM2 and SM3 in the Supporting information, File S1). During late spring and summer, simulated populations grow rapidly (Fig. 6A) but, as the autumn temperatures cool, reproduction and survival decline to zero (µ_adult) (Fig. 6A) requiring annual reinitialization of fly populations in the simulation. Annual µ_cold (i.e. the sum of daily µ_adult) during 1961–1983 ranged from 13–18, leading to local extinction of the fly, thus obviating the need for local eradication efforts.

McAllen, Texas in the transition zone had average µ_cold of 7.92, allowing low level overwinter survival, although cold temperatures during 1979 caused near local extinction (see above) (Figs 5B and 6B). In the model, bi‐weekly releases of two sterile flies km⁻² (i.e. both sexes) were sufficient for eradication (Fig. 6B′); a level grossly at odds with the observed weekly release rates of 39–896 sterile flies km⁻² (Matlock et al., 1996). During the 1972 outbreak, 1890 sterile flies were released per case of myiasis (Fig. 3B) (see also Fig. SM1 in the Supporting information, File S1) and yet the number of reported cases in Texas remained at approximately 16 500 per month during summer (Fig. 7B). Using the simulated average daily density across all years (e.g. 12.67 adult flies km⁻²/day) (Fig. 6B) as the initial density, bi‐weekly releases of 15 sterile flies were required for simulated eradication (see Fig. SM5 in the Supporting information, File S1).

In tropical areas such as Tampico and Tuxtla‐Gutierrez, Mexico, fly populations are endemic and, despite periods of cool temperatures, local extinction is unlikely (Fig. 6C,D). Simulated fly populations at Tuxtla‐Gutierrez show less variability than at Tampico where winter temperatures are cooler during winter. The effects of cool temperatures in this region are seen in field data from Pozo‐Rico, Veracruz, Mexico, where oviposition rates fell to zero and adult mortality increased as temperatures declined to near the developmental threshold of 14.5 °C (see Fig. SM6A,C in the Supporting information, File S1) (data from Krafsur et al., 1979). Simulated eradication at Tampico required bi‐weekly releases of approximately 3 sterile flies/km², whereas 23 sterile flies were required at Tuxtla‐Gutierrez (Fig. 6C′,D′). The simulated approximately 7.7‐fold difference is a result of cooler weather at Tampico. The low simulated biweekly values contrast sharply to with weekly field releases of 386 sterile flies/week/km² in Mexico (Krafsur, 1985). The observed weekly release‐rate at Tuxtla‐Gutierrez was approximately 33.5‐fold higher than the simulated bi‐weekly values for a comparative efficacy of 3%. The low field efficiency of SIT is further shown by the modest sterile mating rates that ranged from 10% to 31% in Mexico (Krafsur et al., 1979, 1980) and from 18.6% to 71.4% in Mexico and Guatemala (Krafsur et al., 1987); with this at low native fly densities (e.g. 7.2 to 53.9 females/km² in Mexico). The efficacy of SIT releases corrected for mating success (e.g. 18.6–71.4% above) was computed as the percentage of unmated females × 0.03, yielding an efficacy range of 0.1–2.4% with an average of approximately 1.7%. These values serve to inform qualitatively why massive numbers of sterile males were required to eradicate relatively low field populations. Furthermore, field evidence suggests that the effect of ϕ_release ≫ ϕ_comp (Krafsur & Garcia, 1978; Krafsur,1987).

Role of screwworm life history and weather in fly eradication in Mexico‐Central America

Ecologists venture that adaptive life histories strategies evolve as a consequence of density and environmental factors (Oizumi et al., 2016). Indicative of an r‐strategy (Pianka, 1970) in screwworm are its very high reproductive potential, allowing rapid opportunistic exploitation of oviposition sites (Thomas & Mangan, 1989), a potential population doubling times of 14 days (Matlock & Skoda, 2009), and an aggregating behaviour of screwworm males and low dispersal of unmated females that enhance mating before dispersal (Krafsur, 1978; Krafsur & Garcia, 1978), as well as promiscuous mating in males (polygyny) and single mating in females. The r‐strategy potential of screwworm is demonstrated by outbreaks during some years induced by low densities of mated female migrants invading non‐endemic temperate areas of Texas.

However, in the tropics, growth rates of endemic field population are low (Thomas & Mangan, 1992) with field doubling times ranging from 54 to 139 days (Matlock & Skoda, 2009), oviposition site densities (wounds) are low and can be viewed as an environmental carrying capacity (Krafsur et al., 1979), and there appears to be a low innate success rate in finding oviposition sites (Krafsur et al., 1979, Krafsur et al., 1980). These attributes in the tropics in addition to promiscuous mating in males (polygyny) and single mating in females makes screwworm highly susceptible to massive SIT releases, this despite an estimated low average efficacy of approximately 1.7%.

Furthermore, screwworm has boom to bust dynamics (Krafsur, 1998) as the oviposition sites and weather allow (Krafsur et al., 1980). Bust phases occur with declining temperatures that decrease fly vital rates and increase mortality rates (e.g. McAllen, Texas, and Tampico, Mexico), suggesting that the added load of massive sterile fly releases during bust periods, even if not all females are sterile mated, could drive intrinsically low screwworm populations to demographic mate limited ‘Allee’ extinction (Courchamp et al., 2008). In tropical areas such as Tuxtla‐Gutierrez, cold weather effects are weak (Fig. 6D) and hence higher simulated levels of SIT releases were required compared with Tampico where low non‐freezing temperatures occur (Fig. 6C). The contrasting simulated dynamics of screwworm at McAllen and Tuxtla‐Gutierrez are shown in Fig. SM5 in the Supporting information, File S1.

By contrast with screwworm, a climatically adapted species such as olive fly with high reproductive rates, abundant hosts and other r‐selected attributes has proven difficult to eradicate using SIT methods. This may have been a result of asynchronous mating activity between the wild and released sterile populations, and/or low competitiveness of the radiation‐sterilized mass‐reared flies (Ant et al., 2012). SIT eradication efforts of pink bollworm in the SW U.S.A. were coupled with the highly effective GMO Bt cotton that greatly reduces wild populations (http://www.westernfarmpress.com/cotton/pink-bollworm-eradication-cotton-2017). However, claims of SIT eradication of pink bollworm in the San Joaquin Valley of central California are flawed because the moth cannot overwinter there and Bt cotton is not widely grown (Gutierrez et al., 2006). In general, claims of establishment of invasive species below detectable levels or of eradication (e.g. fruit flies) (Papadopoulos et al., 2013) are suspect if the effects of weather on the dynamics are ignored.

Potential reinvasion and prospective effects of climate change

Screwworm is endemic to the Caribbean and South America, and periodic cases of myiasis occur in North America (Alexander, 2006) as exemplified by the severe outbreak of myiasis in deer (Odocoileus virginianus clavium) populations in the Florida Keys in 2016. This infestation was eradicated by releasing 188 million sterile flies (USDA‐APHIS, United States Department of Agriculture – Animal and Plant Health Inspection Service, 2017a; Skoda et al., 2018). SIT containment of the fly continues in Panama, and the early detection and control of sporadic infestations in North America remains the mainstay of screwworm management. In 2000, USDA‐APHIS began producing millions of sterile flies annually in its Panama production facility, and releasing them over eastern Panama and areas of Colombia at a cost of $15 million annually (USDA‐APHIS, 2017b). Despite this effort in Panama, ‘… an increased number of [myiasis] cases in … clusters could be due to SIT failure, the regular transport of screwworm‐positive animals …, movement of screwworm‐positive wildlife and a lack of fly control in neighboring Colombia’ (Maxwell et al., 2017). Such problems are potential harbingers of difficulties that may be encountered if attempts are made to extend eradication across the vast tropical and subtropical areas of South America (Gutierrez & Ponti, 2014); a problem that may be exacerbated by climate change.

Cold temperatures limit the potential northward endemic range of the fly, although an increase in average temperatures by year 2050 of 2 °C (or more), as posited by the Intergovernmental Panel on Climate Change (IPCC, Intergovernmental Panel on Climate Change, 2014), would increase its potential endemic range in the U.S.A. However, projections of climate change effects using climate model data are fraught with difficulty because of the complexity of modelling NAMS that make any sound prediction of future weather (Gutzler et al., 2005) and fly dynamics problematic. Several regional climate models have been developed (see supplemental materials, File S1). Using high resolution, bias‐corrected NASA climate model data for 2045–2075 vs. 1975–2005 (Thrasher et al., 2012; NASA, 2015), the model predicts important increases in the prospective endemic range of screwworm northward into the southern U.S.A. and at higher elevations in Mexico (Fig. 8A vs. B). Under climate change, a wide spread reinvasion would greatly challenge low efficacy SIT intervention, as occurred during the large 1972 outbreak, and as suggested by the predicted outbreaks during 1982–2010 (Fig. 5). Furthermore, climate warming would increase cattle tick populations (and wounds) (Pérez de León et al., 2012) and their geographical range, and this would exacerbate the outbreak potential for screwworm. On the positive side, considerable progress has been made with respect to improving SIT eradication technologies (Scott et al., 2017), although these advances need to be augmented by an improved forecasting system of screwworm dynamics on fine time and spatial scale across a large geographical region. The development of this system requires that the weather driven biology of the fly (and of cattle tick) be better documented (Gutierrez & Ponti, 2014) than has occurred previously.