Introduction
Predicting the risk of an insect pest outbreak largely contributes to a more efficient management strategy. Prevision of pest population dynamics helps implement well-prepared and better-targeted control measures. In many cases, pest outbreaks are initially due to relatively small numbers of individuals colonizing the crop. These individuals find optimal conditions for their development and quickly reach high population levels threatening the crop. Knowing these optimal conditions and their impact on population dynamics is therefore critical to prevent an outbreak. As poikilotherm animals, insects depend primarily on the temperature of their environment to develop (Régnière et al., Reference Régnière, Powell, Bentz and Nealis2012). Pest distribution and population dynamics, therefore, can be largely predicted by running temperature-dependent models of development (Tonnang et al., Reference Tonnang, Juarez, Carhuapoma, Gonzales, Mendoza, Sporleder, Simon and Kroschel2013; Azrag et al., Reference Azrag, Pirk, Yusuf, Pinard, Niassy, Mosomtai and Babin2018). These tools also provide standard life history traits such as temperature thresholds and thermal constant that characterize the relationships between insect development and temperature (Wagner et al., Reference Wagner, Olson and Willers1991; Nielsen et al., Reference Nielsen, Hamilton and Matadha2008; Azrag et al., Reference Azrag, Murungi, Tonnang, Mwenda and Babin2017).
Although the coffee berry borer, Hypothenemus hampei (Ferrari) (Coleoptera: Curculionidae: Scolytinae) is the most devastating insect pest of coffee worldwide (Damon, Reference Damon2000; Jaramillo et al., Reference Jaramillo, Borgemeister and Baker2006; Vega et al., Reference Vega, Infante, Castillo and Jaramillo2009), there is much to learn about its thermal biology. This tiny and discreet beetle feeds on coffee berries at all maturation stages, leading to losses of both quantity and quality of coffee beans (Le Pelley, Reference Le Pelley1968; Vega et al., Reference Vega, Blackburn, Kurtzman and Dowd2003; Jaramillo et al., Reference Jaramillo, Chabi-Olaye and Borgemeister2010). Feeding damage is of two types: adult female feeding lesions on developing fruits lead to berry drop, and offspring feeding galleries in the berry endosperm (coffee seeds) lead to bean weight loss and increased vulnerability to disease infection. Economic losses due to H. hampei infestations globally are estimated at US$500 million annually (Pardey, Reference Pardey and Gaitán2015). In eastern Africa, infestation level can be as high as 80–90% in medium to low-elevation coffee plantations (<1500 m asl), causing serious economic loss to the predominant small scale farmers (Jonsson et al., Reference Jonsson, Raphael, Ekbom, Kyamanywa and Karungi2015).
Life history traits such as high reproductive rate and short life cycle (Jaramillo et al., Reference Jaramillo, Chabi-Olaye, Kamonjo, Jaramillo, Vega, Poehling and Borgemeister2009) may account for the success of H. hampei as a devastating pest of coffee. Another factor is that H. hampei spends most of its life cycle inside the coffee berry that makes damage difficult to detect at first sight, and chemical spraying usually inefficient (Brun et al., Reference Brun, Marcillaud, Gaudichon and Suckling1989). Biological and semiochemical control attempts have shown some success (Damon, Reference Damon2000; Dufour & Frérot, Reference Dufour and Frérot2008). In spite of this, until now, none of the control methods recommended for the control of H. hampei has achieved the complete eradication of the pest. Rather, new areas of Arabica coffee production are under increasing threat due to global warming (Jaramillo et al., Reference Jaramillo, Chabi-Olaye, Kamonjo, Jaramillo, Vega, Poehling and Borgemeister2009; Reference Jaramillo, Muchugu, Vega, Davis, Borgemeister and Chabi-Olaye2011). Thus, it becomes urgent to develop new tools and knowledge to support existing management strategies for the pest.
Literature dealing with H. hampei biology reports inconsistent data on pest development as influenced by temperature. For example, Barrera (Reference Barrera1994) reported a larval development time of 17 days at 27°C, while it was 13 days at the same temperature in the study by Fernández & Cordero (Reference Fernández and Cordero2007). The reason for this may be the use of different methodologies for rearing and observation. H. hampei life cycle inside coffee berries makes direct observation difficult and berry dissection may be a source of disturbance that may lead to assessment bias. A first output of the present study is a new rearing and observation method that overcomes this difficulty. This method allowed the monitoring of the development of H. hampei immature stages on a daily basis under seven constant temperatures in the laboratory. This paper therefore provides basic data for H. hampei immature stage development as influenced by temperature. The third output of this paper is a set of standard temperature-dependent models of development that characterize the relationship between H. hampei development and temperature and provide thermal requirements of this major pest of coffee. As components of a general phenology model for H. hampei, these models will help predict the distribution of H. hampei as influenced by temperature on coffee, in the context of global warming.
Materials and methods
Insect field collection for colony initiation
Experiments started with the collection of H. hampei adult females from field populations. Initial sampling was done in small holding coffee farms located on the Aberdare range, in Murang'a County, Kenya (sampling area between 0.710°S, 37.083°E and 0.695°S, 36.923°E, with elevation range ≈1300–1800 m asl). In this area, annual rainfall ranges ≈1200–1800 mm, with two rainy seasons, from mid-March to May and from October to mid-December, the former being the most important (Ovuka & Lindqvist, Reference Ovuka and Lindqvist2000). The annual mean temperature varies according to elevation, with ≈20°C at 1500 m asl. In the area, land use is dominated by small scale coffee farms, mixed with food crops such as maize, beans and banana, with trees such as grevillea, Grevillea robusta A. Cunn. ex R. Br. and macadamia, Macadamia spp. that provide shade to coffee trees. Coffee berries infested by H. hampei were collected from 13 coffee farms. Infested berries are easily detected by the holes the females drill, almost always at the apex of the berry, to penetrate the fruit. Berries were kept in 0.5-litre plastic containers (Foodmate, Kenpoly Manufacturers Ltd, Kenya), measuring 10.4 cm in mean diameter and 6 cm deep, for transportation to the coffee pest laboratory at International Centre of Insect Physiology and Ecology (icipe), Kenya. Container lids had a 4 cm diameter opening, covered with a fine mesh tissue for aeration. Enough infested berries were collected to fill 20 of these containers. In the laboratory, the containers were kept for three weeks in an incubator (SANYO MIR-553, Sanyo Electrical Ltd, Tokyo, Japan) set at 25 ± 0.5°C, with 80 ± 5% RH and 12:12 L:D photoperiod (Jaramillo et al., Reference Jaramillo, Chabi-Olaye and Borgemeister2010). Afterwards, the berries were dissected with a scalpel under a stereo microscope using the ×10 magnification and females were gently collected with a mouth aspirator. Approximately 12,000 H. hampei adult females were collected this way for subsequent experiments.
Egg production
Mature berries of Coffea arabica var. Ruiru 11 were collected from the same coffee farms where H. hampei were sampled for colony initiation. The berries were carefully checked for infestation and only non-infested fruits were collected and transported to the laboratory. Here, berries were washed with detergent (Teepol, Sudi Chemical Industry Ltd, Kenya), then thoroughly rinsed with water and finally placed on a paper towel for 2 h to dry at room temperature. Afterwards, the berries were distributed in 60 containers of the same type as those used for field collection (0.5-litre aerated plastic containers), with each container containing 80 berries. Then, approximately 200 reproductive H. hampei females obtained from field-collected infested berries were introduced into each container. After 18 h, newly infested berries were collected and transferred into new containers (each container had between 50 and 70 infested berries) of the same type as previously (0.5-litre aerated plastic containers) but, lined with a humidified mixture of plaster of Paris and activated charcoal to maintain a high level of humidity (Jaramillo et al., Reference Jaramillo, Chabi-Olaye and Borgemeister2010). These containers were incubated (as previously described) at 25 ± 0.5°C, with 80 ± 5% RH and 12:12 L:D photoperiod. Five days after infestation, the berries were dissected under a stereo microscope using the ×10 magnification and the eggs were carefully collected using a fine camel-hair brush and placed on discs made of paper towels in small plastic Petri dishes (3.5 cm wide, 1 cm deep). Each Petri dish contained 10–12 eggs to ease observation under the stereo microscope. We assumed that most of the eggs were ≤24 h old at that time (all of them were <48 h), since H. hampei females usually start laying eggs 4–6 days after they have penetrated the berry (Jaramillo et al., Reference Jaramillo, Chabi-Olaye and Borgemeister2010).
Development and survival of immature stages
Eggs obtained with the method described above were incubated in laboratory incubators of the same model as previously mentioned, but now set at the following seven constant temperatures: 15, 18, 20, 23, 25, 30 and 35°C (± 0.5°C), with 80 ± 5% RH and 12:12 L:D photoperiod. For each temperature, between 100 and 200 eggs were observed daily for a month under a stereo microscope (×10 magnification) to detect hatching and assess the incubation period. The eggs that did not hatch during this period were observed for two additional weeks and then, if not hatched, recorded as dead.
After emerging, the larvae were reared individually on fresh coffee seeds. Mature coffee berries were dissected to extract the two seeds from inside. Then, a slit approximately 1.5 mm deep and 2 mm wide was dug on the seed using a sharp scalpel blade, where the larva was carefully placed using a fine camel hair brush. The seed with the larva was gently wrapped with aluminium foil to maintain the larva in conditions as close as possible to those inside the berry and to prevent the larva from escaping. Each seed with larva was labelled and placed in a well (2.5 cm in diameter and 2 cm deep) of a 12-well plate (Costar, Corning Inc., NY, USA). Larvae were transferred to new fresh seeds every 4 days to prevent them from boring deeply in the seed. The larvae were monitored daily under a stereo microscope to record the pupation and mortality. After pupation, the pupae were carefully extracted from the seeds and kept in the same well plates lined with paper towel and monitored daily until adults emerge. This rearing method enabled us to directly observe the development and survival of all immature stages and follow the same individuals from egg to adult.
Model parameterization
The impact of temperature on the development and mortality of H. hampei immature stages was described with linear and non-linear functions using the Insect Life Cycle Modelling software (ILCYM, version 3.0) (Tonnang et al., Reference Tonnang, Juarez, Carhuapoma, Gonzales, Mendoza, Sporleder, Simon and Kroschel2013). ILCYM includes a model builder that facilitates the fitting of non-linear functions to the observed data. These models allow the calculation of the thermal requirements of the insect by describing the temperature dependency of development time, development rate and mortality rate for each life stage. The best-fitted models were selected based on their coefficient of determination (R 2) and Akaike's information criterion (AIC) (Tonnang et al., Reference Tonnang, Juarez, Carhuapoma, Gonzales, Mendoza, Sporleder, Simon and Kroschel2013).
Modelling the development time distribution
The frequency distributions of insect development time are usually skewed, and it is assumed that the distributions have the same shape at different constant temperatures (Sharpe & DeMichele, Reference Sharpe and DeMichele1977). In a first step, cumulative frequencies of development times were plotted, for each life stage and temperature, against ln-transformed development times (normalized development time). Then, common binary distribution models were fitted to observed data in a parallel line approach to estimate the development time. The estimated development time was the median of the distribution; in other words, the time required for 50% cumulative frequency of individuals in each life stage to complete the development. The best-fitted models were complementary log–log (CLL) model for the egg stage and the complete development from egg to adult, and probit model for the larva and pupa stages. The mathematical expressions of the CLL and probit functions are given in table 2 (Tonnang et al., Reference Tonnang, Juarez, Carhuapoma, Gonzales, Mendoza, Sporleder, Simon and Kroschel2013).
Modelling the effect of temperature on the development rate
The development rate was calculated at each constant temperature and for each immature stage, and for the complete development from egg to adult as the inverse of the median development time (development rate = 1/development time). We chose median development time because distributions of insect development time usually have similar shapes and the use of median in this case yields one standard curve for all temperatures (Wagner et al., Reference Wagner, Wu, Sharpe and Coulson1984). In addition, the median is less sensitive to outliers compared to the mean times, especially when the distribution of the development is skewed to the longer times (Wagner et al., Reference Wagner, Wu, Sharpe and Coulson1984). The calculated development rate was plotted against temperature and fitted to linear models following the formula:
$$r\,(T)\, = \; a\, + \,bT\; $$where r (T)is the development rate at temperatureT; a is the intercept and bis the slope of the regression line. The minimum temperature threshold (T min), at which the development rate = 0, was estimated using the intercept and slope of the regression line: T min = −a/b; while the thermal constant k (in degree days) was estimated using k = 1/b.
In our study, the relationship between development rate and temperature fitted well to linear regressions for all temperatures. However, this relationship is usually not linear for the highest temperatures of development as demonstrated for other insect species (Sharpe & DeMichele, Reference Sharpe and DeMichele1977). Therefore, non-linear models were also used to describe this effect for each immature stage. In addition, non-linear models allow the assessment of the maximum temperature threshold (T max). The Logan model (Logan et al., Reference Logan, Wollkind, Hoyt and Tanigoshi1976) predicted well the effect of temperature on development rates for egg and pupa stages, while the modified version (five parameters) of the biophysical Sharpe and DeMichele model (Sharpe & DeMichele, Reference Sharpe and DeMichele1977) gave the best fit to the larval stage and for the period from egg to adult. The mathematical expressions of the models are presented in table 4 (Tonnang et al., Reference Tonnang, Juarez, Carhuapoma, Gonzales, Mendoza, Sporleder, Simon and Kroschel2013).
Modelling the effect of temperature on the mortality rate
Mortality rate was calculated for each life stage at given temperatures from the number of surviving individuals. Then, a modified version of the Wang model (Wang et al., Reference Wang, Lan and Ting1982) was applied to describe the effect of temperature on the mortality rate of each immature stage, while a second-order polynomial function was used for the mortality rate for the period from egg to adult. The mathematical expressions of these models are presented in table 5 (Tonnang et al., Reference Tonnang, Juarez, Carhuapoma, Gonzales, Mendoza, Sporleder, Simon and Kroschel2013).
Statistical analyses
The effect of temperature on H. hampei developmental time (in days) was assessed for each development stage and for the complete development from egg to adult. Data for egg incubation period, larva and pupa development time, and egg to adult development were separately subjected to generalized linear model (GLM) with a Poisson distribution as recommended by O'Hara & Kotze (Reference O'Hara and Kotze2010). R programming environment (R Core Team, 2016) was used for calculations with temperature as an independent variable. Once significant differences were detected, data were submitted to post hoc analysis for mean comparison using Tukey's test at α = 0.05.
Results
Development time
Development occurred between 15 and 30°C for egg and between 18 and 30°C for larva and pupa (table 1). The impact of temperature on the observed development times was significant for every H. hampei immature stages, as well as for the complete development time from egg to adult (egg: χ2 = 436, df = 848, P < 0.0001; larva: χ2 = 416.9, df = 745, P < 0.0001; pupa: χ2 = 317.98, df = 920, P < 0.0001; egg to adult: χ2 = 106.26, df = 81, P < 0.0001). The mean observed development time for egg ranged between 4.6 ± 0.1 and 16.8 ± 0.5 days at 30 and 15°C, respectively. The longest mean development time for the larva was 39.5 ± 0.6 days at 18°C, while the shortest was 12.5 ± 0.2 days at 30°C. To complete its development to adult, the pupa stage took an average of 13.6 ± 0.3 days at 18°C and 3.0 ± 0.1 days at 30°C. Mean total development time from egg to adult was 63.4 ± 0.8 days at 18°C and, 18.0 ± 0.2 days at 30°C (table 1).
Table 1. Observed mean development time and development time simulated from the models (median of the distribution) for immature stages of H. hampei reared in the laboratory at different constant temperatures.
Means are in days with SE and n the initial number of eggs observed for each temperature.
Means in each column followed by the same letter are not significantly different (Tukey's HSD, P = 0.05).
The distribution of development times for egg stage and the complete development from egg to adult was well described by a CLL model (R 2 = 0.94–95, AIC = 515.75–1030.1) (fig. 1, table 2). By contrast, the distributions of development times for larva and pupa stages fitted well a probit distribution model (larva: R 2 = 0.96, AIC = 710.4; pupa: R 2 = 0.97, AIC = 339.8). Simulated values for development time (median development time of the distribution) obtained from these models were consistent with observed development times (mean development times) (table 1), attesting the quality of model fitting.
Fig. 1. Cumulative distribution of developmental times of H. hampei egg, larva, pupa and egg to adult. Curves are fitted models: complementary log–log (CLL) model for egg stage and complete development from egg to adult, and probit model for the larva and pupa stages. Bars indicate 95% confidence intervals for median development times estimated from the models.
Table 2. Parameters (a = y-intercept, b = common slope) and goodness of fit estimators (R 2 and AIC) of models fitted to cumulated frequency distributions of development times of H. hampei immature stages reared at six constant temperatures.
1 CLL distribution: f (x) = 1 − exp ( − exp(a i + b ln x)).
2 Probit distribution: 
$f\,(x)\, = \,\emptyset \,(a_i + b\,{\rm ln}\,x)$.
Models: CLL and probit distributions: f(x) is the probability to complete development at time x, ln x is the natural logarithm of the days observed, a i is the intercept of the regression line corresponding to temperature i, and b is the common slope of the regression line in all cases.
Development rate
Temperature had a significant effect on the development rate of immature stages of H. hampei as well as on the total development from egg to adult (P < 0.001) (tables 3 and 4). Linear models showed that the minimum temperatures required for immature stage development (T min) were 10.5, 13.1, 15.0 and 13.0°C, for egg, larva, pupa and egg to adult respectively (fig. 2, table 3). The thermal constants k was estimated at 78.1, 188.7, 36.5 and 312.5 DD for egg, larva, pupa and complete development from egg to adult, respectively. The Logan model gave the best fit for the egg and pupa stages (R 2 = 0.89–99 and AIC = −45.87 to −2.31) (fig. 2, table 4), while for the larval stage and total development from egg to adult, modified version (five parameters) of the Sharpe and DeMichele model gave the best fit (R 2 = 94–0.95 and AIC = −26.09 to −20.3). The maximum temperature threshold (T max) was estimated at 35.2, 34.4, 33.9 and 32°C for egg, larva, pupa and from egg to adult, respectively (fig. 2, table 4).
Fig. 2. Temperature-dependent developmental rate of H. hampei egg, larva, pupa and egg to adult. Observed values are the black points, with bars representing standard deviation of the mean. Fitted models are the dashed straight lines for linear regression and solid lines for the Logan model (egg and pupa) and Sharpe and DeMichele model (larva and egg to adult). Dashed lines above and below represent the upper and lower confidence bands.
Table 3. Estimates of the linear regression describing the effect of temperature on H. hampei development rate (1/day).
k, thermal constant in degree days (DD); T min, minimum temperature threshold; R 2, coefficient of determination; AIC, Akaike's information criterion.
Table 4. Model parameters of Logan and Sharpe and DeMichele models describing the effect of temperature on H. hampei immature stage development rate (1/day).
For Logan models Y, ρ and v, model parameters (mean ± SE); T max, maximum temperature threshold (in °C); and for Sharpe and DeMichele model, P, T o, H A, T L and H L, model parameters (mean ± SE); R is the universal gas constant (1.987 cal degree−1 mol−1), F, F-test statistic; df, degree of freedom; P, probability value; R 2, coefficient of determination; AIC, Akaike's information criterion.
Effect of temperature on the mortality rate
Temperature had a significant effect on the mortality rate of the egg and larval stages and for the total development from egg to adult (egg: F = 68.04, df = 2, 3, P < 0.05; larva: F = 141.45, df = 2, 3, P < 0.01; egg to adult: F = 59.76, df = 2, 4, P < 0.05) (table 5). However, the tested temperatures did not have a significant effect on the mortality of the pupa stage (F = 14.73, df = 1, 3, P = 0.18). For all immature stages, the best-fitted model was the Wang 2 (R 2 = 0.98–0.99 and AIC between −22.92 and −21.20), while second-order polynomial functions gave a good fit to mortality for the total development from egg to adult (R 2 = 0.98, AIC = −14.9) (fig. 3, table 5). The larval stage had the highest mortality rate for the tested temperatures with 100, 50, 38, 28, 33 and 80% at 15, 18, 20, 23, 25 and 30°C, respectively (fig. 3). The thermal window for H. hampei survival from egg to adult was estimated from the second order polynomial function between 16.1 and 30.3°C, and the optimum temperature for survival was estimated at 23.2°C (fig. 3).
Fig. 3. Temperature-dependent mortality rate of H. hampei immature stages fitted to Wang 2 function for egg, larva and pupa, and to second order polynomial function for the complete development from egg to adult. The points are observed values and the solid curves are the selected model output. Dashed lines above and below represent the upper and lower 95% confidence bands of the models.
Table 5. Model parameters of Wang 2 function (Tl, h, B and H ± SE) and second order polynomial function (b 1, b 2 and b 3 ± SE) testing temperature effect on H. hampei immature stage mortality rate.
F, F-test statistic; df, degree of freedom; P, probability value; R 2, coefficient of determination; AIC, Akaike's information criterion.
Discussion
Observation method
Here, we developed and validated a new observation method that allowed direct monitoring of the complete development from egg to adult of a large number of H. hampei individuals, whilst maintaining rearing conditions similar to those found in a coffee berry. This is the recommended approach for accurate assessment of the impact of temperature on insect demography (Tonnang et al., Reference Tonnang, Juarez, Carhuapoma, Gonzales, Mendoza, Sporleder, Simon and Kroschel2013). Due to the cryptic nature of the pest, life cycle observation has been challenging in the past. Different methods and approaches were used in both field and laboratory that include those in Baker et al. (Reference Baker, Barrera and Rivas1992), where coffee berries on trees in the field were artificially infested with H. hampei and sampled after every 3–4 days for dissection to observe immature stages. Fernández & Cordero (Reference Fernández and Cordero2007) used parchment coffee beans moistened for 24 h to feed the larval stage in the laboratory. By contrast, Brun et al. (Reference Brun, Gaudichon and Wigley1993) reared H. hampei on artificial diets in the laboratory, which made the observation easier. However, artificial diets are but proxies to natural food for H. hampei and thus might affect the development process of the pest. In the recent past, Jaramillo et al. (Reference Jaramillo, Chabi-Olaye, Kamonjo, Jaramillo, Vega, Poehling and Borgemeister2009) developed an observation technique based on artificial infestation of coffee berries in the laboratory, where berries were dissected on a daily basis to observe the development of groups of individuals. The most innovative improvement of our method is probably the way we monitored the larva development on fresh coffee seeds and followed the same individuals from egg through to the adult stage. On the one hand, packing the fresh coffee seed hosting the larva in an aluminium foil allowed us to mimic the living conditions inside the berry, especially darkness and high level of humidity. On the other hand, the aluminium package was easily opened without damaging or disturbing the larva, which is not ensured when berries are dissected. This method will be adapted to assess H. hampei female fecundity and adult longevity for the calculation of life table parameters.
Temperature-dependent development models
Although insects do not develop at a constant temperature in nature, development models obtained from laboratory studies provide useful information on their thermal biology, such as thermal thresholds. As such, they help understand and predict the pest distribution and population dynamics in plantations according to temperature (Tonnang et al., Reference Tonnang, Juarez, Carhuapoma, Gonzales, Mendoza, Sporleder, Simon and Kroschel2013; Azrag et al., Reference Azrag, Pirk, Yusuf, Pinard, Niassy, Mosomtai and Babin2018). To the best of our knowledge, the only study that adopted a similar approach for H. hampei is that of Jaramillo et al. (Reference Jaramillo, Chabi-Olaye, Kamonjo, Jaramillo, Vega, Poehling and Borgemeister2009). In that study, a combination of a linear regression and the modified version of the Logan model was used to characterize the relationship between the development rate and temperature for all life stages. In our study, ILCYM software was used to fit the linear regression and to select the best nonlinear model amongst 58 models used to describe this relationship for insects (Tonnang et al., Reference Tonnang, Juarez, Carhuapoma, Gonzales, Mendoza, Sporleder, Simon and Kroschel2013).
The linear model well predicted the development rate for all the tested temperatures, confirming what has been found for a number of insect species, including H. hampei (Sporleder et al., Reference Sporleder, Kroschel, Quispe and Lagnaoui2004; Jaramillo et al., Reference Jaramillo, Chabi-Olaye, Kamonjo, Jaramillo, Vega, Poehling and Borgemeister2009; Azrag et al., Reference Azrag, Murungi, Tonnang, Mwenda and Babin2017). The Logan and five parameters of the Sharpe and DeMichele models (Logan et al., Reference Logan, Wollkind, Hoyt and Tanigoshi1976; Sharpe & DeMichele, Reference Sharpe and DeMichele1977) were the best nonlinear models. The biophysical model of Sharpe & DeMichele has a biological significance in predicting insect development rate (Sharpe & DeMichele, Reference Sharpe and DeMichele1977). It includes thermodynamic parameters associated with the development, such as the enthalpy of enzyme activation (Sharpe & DeMichele, Reference Sharpe and DeMichele1977). In addition, the model can be fitted in different forms (i.e. four, five and six parameters), which makes it flexible in fitting different temperature ranges (Tonnang et al., Reference Tonnang, Juarez, Carhuapoma, Gonzales, Mendoza, Sporleder, Simon and Kroschel2013). On the other hand, Logan model (Logan et al., Reference Logan, Wollkind, Hoyt and Tanigoshi1976) has a restricted number of parameters and it considers enzyme-catalysed biochemical reaction rate at optimum temperature.
For the mortality rate, the best model was the Wang 2 model (Wang et al., Reference Wang, Lan and Ting1982). This model gives a very good fit to the data over a wide range of constant temperatures (Tonnang et al., Reference Tonnang, Juarez, Carhuapoma, Gonzales, Mendoza, Sporleder, Simon and Kroschel2013). It showed that the mortality rate of H. hampei immature stages was around 100% at 13°C but, decreased with an increase in temperature and reached a minimum value at 23°C. Thereafter, the mortality increased again and reached 100% at around 34°C for all immature stages. This model was previously used to predict the relationship between temperature and mortality of immature stages for different tropical pests, such as the mealybug Phenacoccus solenopsis Tinsley (Fand et al., Reference Fand, Tonnang, Kumar, Kamble and Bal2014) and the leaf miner Liriomyza huidobrensis Blanchard (Mujica et al., Reference Mujica, Sporleder, Carhuapoma and Kroschel2017).
Life cycle
The complete life cycle of H. hampei was obtained under constant temperature in the range 18–30°C, with a total developmental time of about 63 days at 18°C and 18 days at 30°C. The egg incubation period ranged from 4.6–16.9 days under a temperature range of 30–15°C, and it might have been slightly underestimated with our method (by a day less). Jaramillo et al. (Reference Jaramillo, Chabi-Olaye, Kamonjo, Jaramillo, Vega, Poehling and Borgemeister2009) reported an egg incubation period of 4.7–12.0 days under the temperature range 33–20°C. In that study, the authors did not get oviposition by females at 15°C. Our result at 25°C is similar to that reported by Brun et al. (Reference Brun, Gaudichon and Wigley1993), who obtained an incubation period of 5 days at the same temperature for a population reared on an artificial diet. On the other hand, Ruiz et al. (Reference Ruiz, Bustillo-Panley, Flórez and González1996) found an incubation period (3.3 days) at 26°C shorter than ours. The variation in incubation period between these studies may be linked to the methods used to produce eggs for the experiments and to the incubation conditions, such as observation settings, relative humidity and photoperiod, which might have played a role.
For larval stage, development times we obtained are in agreement with findings of Jaramillo et al. (Reference Jaramillo, Chabi-Olaye, Kamonjo, Jaramillo, Vega, Poehling and Borgemeister2009) for all tested temperatures, with the exception of those obtained at 23°C. These authors reported a larval development time of 17 days, which differs from 21 days we reported here. In fact, in our study, only 12 individuals completed the larval stage in 17 days, which is the lowest value at 23°C. By contrast, 80% of the individuals completed this stage in a time between 20 and 23 days. The larva development time assessed by Chami (Reference Chami2003) at 25°C (28.1 days) was much longer than ours (15.2 days) at the same temperature. Here again, these variations may be due to methods and conditions used to maintain and monitor H. hampei larvae. The development time of pupa we obtained at 25°C was similar to those recorded by Bergamin (Reference Bergamin1943) and Chami (Reference Chami2003).
Thermal requirements
The minimum temperature thresholds (T min) we obtained are similar to those reported by Jaramillo et al. (Reference Jaramillo, Chabi-Olaye, Kamonjo, Jaramillo, Vega, Poehling and Borgemeister2009), with the exception of the egg stage, for which we found 10.5°C compared to 16.7°C in that study. The thermal constant for the complete development from egg to adult in our study (312.5 DD) is also comparable to the 262.5 DD reported in the study by Jaramillo et al. (Reference Jaramillo, Chabi-Olaye, Kamonjo, Jaramillo, Vega, Poehling and Borgemeister2009). Again, differences may be due to experimental conditions and observation methods. Another explanation may be that H. hampei individuals used in Jaramillo et al. (Reference Jaramillo, Chabi-Olaye, Kamonjo, Jaramillo, Vega, Poehling and Borgemeister2009) were collected in a different region in Kenya, with a different climate. We cannot rule out that populations from different geographical areas have slightly different thermal requirements.
Our models gave a thermal window of 13–32°C for H. hampei to complete its development from egg to adult. On the other hand, the thermal window for survival obtained from mortality models is narrower with about 16–30°C, with optimal temperature for survival around 23°C. This is in line with other studies that showed that insect development rate response to temperature is different from survival response, due to some other factors such as the diet and manipulations, which contribute to the mortality rate beside temperature (Sporleder et al., Reference Sporleder, Kroschel, Quispe and Lagnaoui2004; Mujica et al., Reference Mujica, Sporleder, Carhuapoma and Kroschel2017). For example, this was true for the survival response to temperature of the whitefly Bemisia tabaci (Gennadius) and the leaf miner L. huidobrensis (Blanchard), which differed from developmental rate response (Bonato et al., Reference Bonato, Lurette, Vidal and Fargues2007; Mujica et al., Reference Mujica, Sporleder, Carhuapoma and Kroschel2017).
Implication for pest management
Arabica coffee does not tolerate too high temperature and the crop is usually grown high in tropical mountains or highlands. The range of elevation favouring Arabica coffee cultivation varies considerably across the tropical belt; in east Africa, the crop is usually found in the range 1000–2000 m asl (Garedew et al., Reference Garedew, Lemessa and Pinard2017; Liebig et al., Reference Liebig, Babin, Ribeyre, Läderach, van Asten, Poehling, Jassogne, Cilas and Avelino2018). There, H. hampei usually thrives in coffee plantations below 1400 m asl, where the climate is warmer. At world scale, most studies showed that infestation by the pest decreases with an increase in elevation (e.g. Jaramillo et al., Reference Jaramillo, Muchugu, Vega, Davis, Borgemeister and Chabi-Olaye2011; Avelino et al., Reference Avelino, Romero-Gurdián, Cruz-Cuellar and Declerck2012). According to our results, the temperature range 20–28°C is the most suitable for H. hampei immature stage development and a constant temperature of 23°C is optimal for their survival. Such temperature range roughly matches conditions of low elevation plantations of Arabica coffee in east Africa, where H. hampei causes the highest damage to the crop (Jaramillo et al., Reference Jaramillo, Muchugu, Vega, Davis, Borgemeister and Chabi-Olaye2011). Our results therefore confirm that, provided that H. hampei fecundity relationship to temperature shows a similar trend, low-elevation coffee in east Africa should be considered as the highest in terms of infestation risk by the coffee berry borer and should focus attention on its management.
Another major factor impacting temperature in coffee plantations is shade. As an understorey plant, coffee grows well under shading trees and shading canopy largely affects pests and diseases through microclimate. For H. hampei, a well-managed shade is sometimes considered as a promising strategy to reduce the temperature and keep pest outbreaks at bay (e.g. Teodoro et al., Reference Teodoro, Tscharntke and Klein2009; Jaramillo et al., Reference Jaramillo, Muchugu, Vega, Davis, Borgemeister and Chabi-Olaye2011). However, the impact of shade on coffee infestation by H. hampei is far from obvious and mechanisms involved are complex (Avelino et al., Reference Avelino, Romero-Gurdián, Cruz-Cuellar and Declerck2012; Jonsson et al., Reference Jonsson, Raphael, Ekbom, Kyamanywa and Karungi2015; Mariño et al., Reference Mariño, Pérez, Gallardo, Trifilio, Cruz and Bayman2016). Our results will help understand the impact of shade on H. hampei populations through microclimate and thus will contribute to more precise and efficient recommendations for shade management on coffee.
Finally, climate change is expected to impact the distribution of many insect pests worldwide. For H. hampei, the threat is expected to worsen in east Africa, with the extension of the distribution to coffee areas at altitudes higher than today, where Arabica coffee is particularly renowned for its quality (Jaramillo et al., Reference Jaramillo, Muchugu, Vega, Davis, Borgemeister and Chabi-Olaye2011). The models we developed in the present study will help predict H. hampei distribution under global warming and assess the risk in terms of production loss. Our study will therefore contribute to improving mitigation strategies against this important pest.
Conclusions
In conclusion, we developed an observation method that allowed for the first time the monitoring of the development of the same individuals of H. hampei from egg to adult. The models presented here gave good predictions for immature stage development and survival according to temperature. Models also provided thermal requirements for immature stage development. With the incorporation of oviposition models and validation under fluctuating temperature, it will help understand and predict H. hampei distribution on coffee production areas. This information will be incorporated in pest management programmes for better control of this major pest of coffee, in the context of climate change.
Acknowledgements
We acknowledge the financial support for this research by the following organisations and agencies: The Centre de Coopération Internationale en Recherche Agronomique pour le Développement (CIRAD), Montpellier, France; UK's Department for International Development (DFID); Swedish International Development Cooperation Agency (Sida); the Swiss Agency for Development and Cooperation (SDC); and the Kenyan Government. The first author AGA Azrag was supported by a German Academic Exchange Service (DAAD) In-Region Postgraduate Scholarship as well as the National research foundation of South Africa to AA Yusuf & CWW Pirk. The views expressed herein do not necessarily reflect the official opinion of the donors.
