Olea europaea L. is one of the oldest and most abundant tree crops in the Mediterranean regions, where it is of essential socioeconomic and ecological importance (Michalopoulos et al., 2020; Caselli and Petacchi, 2021). Approximately 95% of the global demand for olive oil is satisfied by southern European countries, with Spain, Italy, and Greece being the main olive oil producers (Fraga et al., 2021). Given the importance of olive cultivation, it is imperative to effectively manage pests that can cause significant production losses. More than 255 species, including insect pests, mites, nematodes, and pathogenic microorganisms, are potentially harmful to O. europaea (Caselli and Petacchi, 2021). Most of the yield loss is caused by the key pest Bactrocera oleae (Rossi) (Diptera: Tephritidae). In addition, moths [e.g., Prays oleae (Bernard)] contribute locally or occasionally to yield decline (Caselli and Petacchi, 2021). Insects can also be pathogen vectors, such as Philaenus spumarius L., the primary vector of Xylella fastidiosa subsp. pauca, which is responsible for the olive quick decline syndrome (Elbeaino et al., 2014; Sevarika et al., 2022). In a scenario of climate change, the ability to predict the evolution of the cycle of pests and plants, assessing the risks associated with their interaction, represents a critical challenge for the implementation of a proper control strategy.

Herbivorous insects, especially those with low thermal thresholds, are highly sensitive to changes in climate (Deutsch et al., 2008). Climate change may alter plant–pest phenological events, such as flowering and leaf unfolding, insect overwintering, and migration (Gordo and Sanz, 2005). Depending on the insect species, higher mean daily temperatures and extreme climate events, no longer sporadic, could cause the extension of suitable geographical ranges for herbivores but also the disruption of the synchronization of biological cycles among herbivores and their natural enemies (Forrest, 2016; Bonsignore et al., 2020; Skendžić et al., 2021), with an expected substantial increase in crop losses (Deutsch et al., 2018). As a result, monitoring methods and pest management programs must be reviewed and, if necessary, adapted to the new climatic changes that are occurring. In recent decades, pest control relied mainly on the use of broad-spectrum insecticides, with negative effects related to the decline of natural enemies, the occurrence of insecticide resistance in the target population, environmental pollution, and human health (Damos et al., 2015). Integrated pest management (IPM) is now commonly adopted, foreseeing the combined use of synthetic insecticides with more sustainable methods (Stetter and Lieb, 2000). The use of successful prediction models can play a pivotal role within IPM, also in combination with machine learning algorithms, used to manage complex datasets (McQueen et al., 1995; Damos, 2015; Raza et al., 2015; Benos et al., 2021; Caselli and Petacchi, 2021; Skendžić et al., 2021). Degree day models are a conventional instrument for predicting insect phenology (AliNiazee, 1979; Jones et al., 1991; Song et al., 2003; Rebaudo and Rabhi, 2018; Barker et al., 2020). However, physiologically based population modeling merges information on insect development and crop phenology to achieve more accurate predictions (Gutierrez et al., 2009; Rossini et al., 2022). Since a precise indication of pest outbreaks is necessary in an IPM context, machine learning has been applied in decision support systems (Capalbo et al., 2017; Ip et al., 2018; Rossini et al., 2022). A decision support system is a set of computer programs, mathematical models, and heuristic information that operate synergistically to improve decision-making (Nestel et al., 2019).

The olive fruit fly, B. oleae is an important pest of Olea spp. in Europe, Asia, Africa, and North America (Varikou, 2022). Bactrocera oleae is expected to expand due to global warming, thus colonizing areas at higher latitudes and altitudes (Petacchi et al., 2015; Marchi et al., 2016). Furthermore, an increase in average temperatures can affect the adult phenology in spring or can halt egg development in summer, but, on the other side, it can prolong the oviposition period in autumn, possibly resulting in increased yield losses (Caselli and Petacchi, 2021). The economic damage of B. oleae is estimated at more than one billion US dollars per year, only in the Mediterranean region (Van Asch et al., 2015).

The larval stage is responsible for both qualitative and quantitative damage due to its feeding activity within the olive mesocarp, leading to a strong decline in oil quality and premature fruit drop (Gömez-Caravaca et al., 2008). IPM programs against B. oleae are primarily based on monitoring of adults, sampling of olives to evaluate active infestation, and eventually pesticide treatments. In recent years, understanding the population dynamics of B. oleae has become a major focus of research on this pest. Since dimethoate use has been banned due to its toxic effects (Commission Implementing Regulation (EU), 2019/1090), alternative prevention-based control strategies are now recommended. Olive orchard monitoring is required due to the numerous factors that affect B. oleae infestations such as temperature, weather conditions, geographical location, olive tree variety, and management practices (Wang et al., 2009, 2013; Johnson et al., 2011; Rizzo et al., 2012; Petacchi et al., 2015; Volpi et al., 2020). Moreover, the olive variety can determine the fruit susceptibility to B. oleae attacks. Preferences are based on factors such as the size and shape of the fruits and their concentration of phenolic compounds (Varikou et al., 2022; González-Fernández et al., 2023). Furthermore, the mineral element content of the fruit may also influence female choice, making fruits that contain higher amounts of K and Fe more susceptible to attacks (Garantonakis et al., 2016). Concerning the use of digital tools, so far, some predictive models, machine learning algorithms, and decision support systems have been used for B. oleae monitoring and control (Ordano et al., 2015; Petacchi et al., 2015; Marchi et al., 2016; Miranda et al., 2019; Benhadi-Marín et al., 2020; Volpi et al., 2020; González-Fernández et al., 2023). However, proper calibration of these tools for the Umbria region (Central Italy) is lacking. To fill the knowledge gap about the effect of climate and environment on B. oleae in the Umbria region, we analyzed the dynamics of B. oleae infestation over 7 years (from 2015 to 2021) in 79 olive groves in total.

The best model to explain the probability of attack in July–August retained five explanatory variables. Attack was negatively affected by DEM and TEMP (MAR–MAY), but positively affected by TEMP (NOV–FEB), CDD (PLANT), and SWB (the results of the best-fitted generalized mixed-effect model are reported in Table 2 ; Figure 1 ). Concerning the attacks in September–October, these were negatively affected by the increase of SWB and TEMP (MAR–MAY) but positively by TEMP (NOV–FEB) and TEMP (−7D) ( Table 3 ; Figure 1 ). Concerning the day of the first attacks ( Table 4 ), the increase in CDD (PLANT), DEM, and SWB delayed the occurrence of the attacks in the July–August period. On the contrary, an increase in LAKES, TEMP (NOV–FEB), TEMP (MAR–MAY), and TEMP (−7D) had a positive effect in anticipating the attacks. For the September–October period, higher values of DEM and CDD (PLANT) delayed the attacks, while an increase in TEMP (MAR–MAY) and TEMP (−7D) anticipated the attacks by B. oleae ( Table 5 ).

This study represents the first attempt to understand B. oleae population dynamics in Umbria (Central Italy) using a combination of landscape and agroclimatic variables, which could be further leveraged in studies related to larger territories with similar geographical characteristics. The assessment of B. oleae infestation was carried out over two crucial periods of the year, that is, in July–August, during the fruit growth period, and in September–October, during the preharvest period.

Our analysis reveals that olive orchards located at higher altitude expected lower attacks during summer (July–August period). In these sites, attacks are also delayed during the year (July–August and September–October periods). Similarly, Helvaci et al. (2018) reported that the infestation rate was inversely related to both altitude and relative humidity in olive orchards of Northern Cyprus. Furthermore, temperatures in winter and spring have a strong effect on determining attacks and timing of infestation. Average daily temperatures were approximately 7°C in November–February, i.e., when B. oleae is mostly overwintering, and determined an attack probability of approximately 50% (averaged across years). Temperature increase in this period has a positive effect in enhancing future attacks throughout the year, e.g., with an increase in the early season of ca. 35% probability of attacks at 8.5°C average temperatures. Our analysis suggests that temperatures also anticipate the timing of attacks during the early season period of approximately 2 d per degree. Insects that overwinter as pupa in soil have increased survival when temperature increases above 1°C (Bale and Hayward, 2010). Winter temperatures above 0°C gradually reduce the mortality of B. oleae, which can successfully overwinter with large populations (Hatherly et al., 2005; Wang et al., 2013; Petacchi et al., 2015; Marchi et al., 2016). Similarly, Marchi et al. (2016) registered an anticipation of B. oleae appearance in spring and higher infestation rates of juvenile forms (i.e., eggs, first and second instar larvae alive and dead) during the early season period (July to August) in years characterized by mild winters. Helvaci et al. (2018) detected a higher infestation rate as a consequence of higher winter air temperatures.

The increase of the average temperatures in the March–May period negatively affects attack probability but remarkably anticipates the occurrence of first attacks. Given the massive influence of temperature on insects and the fact that higher temperatures often lead to shorter life spans, the combination of a mild winter and a hot spring could potentially be responsible for the premature decline of overwintering adults (Preu et al., 2020), possibly resulting in a lower attack probability. In non-irrigated orchards, such as those investigated here, warmer spring temperatures may also negatively affect pupal survival, whereas irrigation may prevent pupal desiccation (reviewed by Yokoyama, 2015). In addition, during spring, most of the newly emerged B. oleae adults disperse from olive orchards, where fruits are not available, to seek flowers and nectar for survival (Paredes et al., 2023). During these migratory flights, newly hatched adults can also encounter abandoned olive groves, where they can find fruits from the previous year to oviposit. In this scenario, the reduced infestation found in the surveyed olive groves might be due to the higher spring temperatures followed by an increased frequency and distance of migratory flights (Economopoulos et al., 1978; Mazomenos et al., 2002; Ragaglini et al., 2007; Skouras et al., 2007; Marchini et al., 2017; Ortega et al., 2022). The part of the population that did not migrate, on the other hand, may have been responsible for the early attacks recorded.

More hypotheses could be drawn for the observed negative effect of increased spring temperatures on attacks and could, for example, consider the modification of the chemical and physical factors involved in the susceptibility of olives to B. oleae (Tognetti et al., 2006; Malheiro et al., 2015).

The accumulation of degree days relevant to the olive phenology positively determined B. oleae attacks but only for the July–August period, confirming the results of previous studies conducted in other Italian regions (Marchi et al., 2016; Volpi et al., 2020).

Notably, higher temperatures in the 7-day period before the monitoring day increase the attack probability only in the late season and anticipate the occurrence of attacks in both early and late seasons. Other variables rather than temperature may affect the interactions between olive fruit fly and its host, such as soil water content and irrigation type. For example, we detected that an increase in the soil water content increases the attack probability by B. oleae. A low water level can lead to loss of plant turgor and premature fruit drop. Conversely, adequate water supply improves fruit turgidity, which elicits higher B. oleae attacks on fruits (Marchi et al., 2016). Irrigation practice was not considered within the model selection because in Central Italy productive olive plantations are not irrigated. Vicinity to lakes is known to increase the attack probability by B. oleae because of the peculiar microclimate caused by lakes (Gutierrez et al., 2009). However, in our analysis, proximity to lakes was of marginal importance and never affected the probability of attacks. An explanation is that a possible variability in the attacks in orchards gradually distant from lakes could have been captured by other variables considered (e.g., winter and spring temperatures).

Collectively, our results support a potential application of IPM control strategies leveraging predictive models in the future implementation of decision support systems. The main bottleneck of this application is the evaluation of false negatives that can underestimate the infestation rates in olive orchards (Volpi et al., 2020). For these reasons, different interpolation methods must be tested before choosing the best-fit model for a specific geographic and climatic area (Petacchi et al., 2015). In the case of B. oleae, adequate monitoring networks are needed to implement a proper program, as insect distribution is affected by environmental and landscape factors (Thies et al., 2003; Krasnov et al., 2019). Despite the difficulties of precision agriculture strategy for pest control, there are previous studies confirming the successful implementation of digital tools (Miranda et al., 2019). Sciarretta et al. (2019) reported a substantial reduction in the number and volume of pesticide applications, for the control of the Mediterranean fruit fly (medfly), Ceratitis capitata (Wiedermann), in areas managed with a dedicated decision support system and electronic monitoring traps. Given the close relationship between insect development and temperature, further studies on the impact of temperature variations on the spatial and temporal distribution of B. oleae populations are urgently needed. In addition to air temperature, the emergence of B. oleae can be affected by soil temperature and moisture. For example, low soil temperature and high soil moisture due to rain can increase B. oleae pupal mortality (Neuenschwander et al., 1981; Wang et al., 2013). Conversely, low soil humidity may be responsible for pupal desiccation, drastically reducing adult emergence in the spring (Wang et al., 2013; Yokoyama, 2015). The use of specific sensors to measure soil parameters would be necessary to further elucidate the effect of soil temperature and moisture on the overwintering population of B. oleae.

Moreover, traditional control methods (i.e., cover and bait sprays) based on conventional insecticides could become inefficient if the olive fruit fly continues to increase its resistance (Skouras et al., 2007; Pereira-Castro et al., 2015; Marchi et al., 2016). Alternative approaches, by combining powder dust or kaolin with propolis, exhibited an interesting and concrete possibility to reduce fruit infestations by B. oleae (Daher et al., 2022). An integration of different olive fruit fly management methods is recommended, including the use of chemical, biotechnical, and biological control (Lantero et al., 2023). The complexity of the landscape surrounding olive groves has been shown to reduce the abundance of B. oleae and other insect pests (Ortega et al., 2018; Villa et al., 2020). In addition, a more complex environment has been found to enhance the effectiveness of biocontrol agents (Boccaccio and Petacchi, 2009; Ortega et al., 2018). Therefore, agricultural spatial planning should consider the impact of landscape complexity on insect development. Our analysis allowed the identification of key environmental variables influencing the probability and timing of B. oleae attacks that could be used in distribution modeling at the regional scale, or even in more complex modeling methods, based on machine learning algorithms. Further analyses should also evaluate the effect of landscape complexity and composition on olive fruit fly infestations occurring in Central Italy.

Data will be provided upon request directed to GR (gabriele.rondoni@unipg.it).

The manuscript presents research on animals that do not require ethical approval for their study.

GR: Conceptualization, Formal analysis, Methodology, Writing – original draft, Writing – review & editing, Supervision. EM: Methodology, Writing – original draft, Writing – review & editing, Conceptualization, Formal analysis. VAG: Methodology, Writing – original draft, Writing – review & editing. ElC: Methodology, Writing – review & editing, Writing – original draft. AB: Conceptualization, Investigation, Methodology, Writing – review & editing, Supervision. GN: Writing – review & editing, Conceptualization, Investigation, Methodology. RP: Writing – review & editing. FF: Conceptualization, Supervision, Writing – review & editing. AN: Conceptualization, Methodology, Supervision, Writing – review & editing. ErC: Conceptualization, Methodology, Supervision, Writing – review & editing.