The SP was first detected at outbreak levels within the Leiria National Pine Forest, in its southern sector, over an area of approximately 1,000 ha covered by dense, even-aged stands of young P. pinaster. This forest is a highly organized monoculture of P. pinaster, managed to produce high-quality timber for centuries, without any major known disturbances in the decades preceding the detection of the SP (for details, see Pimentel et al. 2006). The SP rapidly dispersed, reaching a maximum extent of 85,000 ha by 2014 (Fig. 1b; Pimentel et al. 2006; Godefroid et al. 2016). The megafires of 2017 destroyed 20,000 ha of coastal forests, encompassing most of the SP’s prime habitat (Fig. 1b). In the summer of 2018, surveys were carried out in the remaining forest to detect nests of SP larvae. This population was detected in only 5,000 ha of maritime pine forests around the town of Nazaré (Fig. 1b), while the WP was not detected in this forest area. Therefore, the SP is effectively isolated from the WP due to the burned area to the north of its distribution area and agricultural fields to the east and south.

For the present work, a forest site (SP) within the only remaining distribution area of the summer population was selected near Nazaré, as well as a similar forest north of the burned area, near Osso da Baleia, with no records of the SP but with the presence of the WP (WP_N). Two more coastal P. pinaster forests were selected more than 150 km south of the SP, with the presence of the WP: a forest site near Melides (WP_S1), part of a continuous coastal strip of pinelands of approximately 9,000 ha, and, approximately 50 km north, a forest site in a 4,000 ha private forest in the Peninsula de Setúbal (WP_S2; Fig. 1a).

To detail differences in the phenology of T. pityocampa populations in the four studied areas, searches for the moth’s immature stages on P. pinaster trees within 1–2 ha of each forest were conducted in June–July and September–October 2019, when egg batches and first larval instars of the SP and WPs were more likely to occur. This search was repeated in the following February to assess the occurrence of the final WP instars. Individual trees were assessed using binoculars. The occurrence of egg batches and the five different instars, easily distinguishable by size, color, hair distribution, and behavior, were recorded, along with the cohort of leaves on which the egg batch was laid and on which the different instars were feeding (see Suppl. material 1: figs S1–S4).

Previous studies have shown that T. pityocampa larvae require mean minimum winter temperatures above –6 °C and absolute minima not below –16 °C and feed only when daytime and nighttime temperatures exceed 9 °C and 0 °C, respectively (Battisti et al. 2005; Buffo et al. 2007; Hoch et al. 2009). Larval colonies construct thermoregulatory tents that can raise internal temperatures several degrees above ambient air, making solar radiation crucial for development once these thresholds are met (Pimentel et al. 2011; Poitou et al. 2021). Tolerance to high temperatures is less well defined: although adapted to cool winters, larvae and eggs can survive, and even develop, at 30–35 °C (Santos et al. 2011b; Robinet et al. 2013), while the SP exhibits even greater heat tolerance, developing at up to 36 °C and surviving short exposures to 42 °C (Santos et al. 2011b).

In this study, a 25-year climate series from 1990–2015, covering the entire period in which the T. pityocampa SP was detected and expanded its distribution, was obtained to explore the climatic conditions behind its evolution and provide a comparison with the WP inhabiting similar forests. Precipitation and temperature data were extracted from https://essd.copernicus.org/articles/11/1947/2019/ (retrieved 15 October 2023). Radiation data were obtained from http://climexp.knmi.nl/ from the E-OBS dataset (https://www.ecad.eu/download/ensembles/ensembles.php; retrieved 15 October 2023). From both sources, data were obtained for the following coordinates: WP_S1, 38.15445, -8.766842; WP_S2, 38.542983, −9.133881; SP, 39.612104, -9.055613; and WP_N, 40.033983, -8.872011. With these data, yearly averages were calculated for the climatic variables most likely to determine T. pityocampa larval survival and growth, both during winter (October–February) and summer (July–September): (1) number of days in winter in which the minimum temperature falls below 0 °C (T0 °C); (2) number of days in winter in which the maximum temperature falls below 9 °C (T9 °C); (3) number of days in summer in which the maximum temperature exceeds 30 °C (T30 °C); (4) number of days in summer in which the maximum temperature exceeds 36 °C (T36 °C); (5) average monthly radiation during winter (RadW); and (6) average monthly radiation during summer (RadS). Some basic climatic parameters were also assessed (Suppl. material 2: table SS1, fig. S7), as well as absolute minimum temperatures and minimum average temperatures for the coldest period of the year (December–January).

At the beginning of 2019, in each of the four studied forests, 10 healthy P. pinaster trees were selected, and three branches were tagged on the south-facing side of each tree. Starting on 1 February and throughout April 2019 and 2020, the sites were visited weekly, and observations were conducted on the terminal bud of each tagged branch. The timing of leaf flushing was determined as the first observation of leaf tips detaching from the bud axis (Suppl. material 1: fig. S5).

In all four studied areas, at three different dates (July and October 2019 and February 2020), three large branches were collected at different heights from the south-facing side of six healthy P. pinaster trees, and samples of leaves were collected. To ensure that foliage chemistry and morphology measurements were not influenced by tree-damage responses to branch removal, a different set of six trees was sampled at each date rather than resampling the same individuals. Maritime pines produce long-lived leaves, and preliminary observations in the four studied areas indicate that they can have up to five cohorts of leaves (year of formation; see Suppl. material 1: fig. S6). Analyses were carried out on composite samples of leaves of the same cohort from the three branches collected from each tree. At all three dates, the two younger cohorts of foliage were analyzed, since preliminary observations indicated that these were preferentially selected by the pine processionary moth. These cohorts are referred to as current-year flushed leaves (Y-0), 1-year-old leaves (Y-1), and 2-year-old leaves (Y-2).

In the laboratory, undamaged fascicles of leaves were randomly selected, and the sheath holding the fascicles together at the base was then removed. Length was measured individually in sets of 30 leaves per tree/cohort, with the average considered the leaf length (cm) for that sample. These sets of leaves were then dried in an oven at 60 °C for 72 h, and the moisture content was determined from the difference in mass between the fresh and dry samples.

Several chemical components of leaves potentially associated with insect development were analyzed. First, the elemental composition of nitrogen (N), which is crucial for protein synthesis and growth, and carbon (C), which is generally associated with plant secondary metabolism and defenses, was determined. The carbon-to-nitrogen (C/N) ratio was also calculated, as it is an important parameter for assessing foliage palatability to insects (Schädler et al. 2003). Additionally, other potentially nutritive components, such as sugars and fatty acids, as well as secondary metabolites associated with plant defenses, such as phenols and terpenes, were also analyzed.

For the analysis of elemental composition, 2 mg of dry leaves from each sample was milled into a fine powder with a ball mill (Type MM2, Retsch, Germany). The elemental composition (N and C) of the leaf material was determined on dried, homogenized samples using a EuroVector EA3000 by Dumas combustion, with Wheat Flour Standard OAS (Elemental Microanalysis, UK; 1.47% N, 39.53% C) as the elemental composition reference material. Results were expressed as % N and % C per dry weight.

To assess the content of sugars, fatty acids, and different secondary metabolites, dry leaves were ground in a knife mill (Retsch SM 2000) to particle sizes below 2 mm and sieved. The 0.250–0.450 mm fraction was used for chemical analysis. To extract the soluble lipophilic compounds, samples were extracted in a Soxhlet apparatus for 6 h with dichloromethane. The extraction yield was determined by the mass difference of the solid residue after drying at 60 °C overnight and for 1 h at 105 °C and was reported as mg per g of the original sample. Approximately 2 mg of dried extract was trimethylsilylated in 100 μL of pyridine (Sigma-Aldrich, ≥99.8% purity, St. Louis, MO, USA) with the addition of 100 μL of bis(trimethylsilyl)trifluoro­acetamide (BSTFA; Sigma-Aldrich, ≥99.0% purity, St. Louis, MO, USA), with the mixture kept for 30 min in an oven at 60 °C. Compounds with hydroxyl and carboxyl groups were trimethylsilylated into trimethylsilyl (TMS) ethers and esters, respectively. GC–MS analyses were then carried out using an Agilent 5973 mass-selective detector (MSD) (Wilmington, DE, USA), and compounds were separated in a Zebron 7HG-G015-02 column (30 m, 0.25 mm i.d., 0.1 μm film thickness), using helium as the carrier gas (1 mL/min). The chromatographic conditions were as follows: injector, 280 °C; oven temperature program, 100 °C (1 min), rate of 10 °C/min up to 150 °C, rate of 4 °C/min up to 300 °C, rate of 5 °C/min up to 370 °C, and rate of 8 °C/min up to 380 °C (5 min). The MSD source was kept at 220 °C and operated in electron-impact mode with an energy of 70 eV, and data were collected at a rate of 1 scan s^–1 over a range of m/z 33–700. Compounds were identified by comparing their mass spectra (MS) with the equipment’s mass spectral library (NIST Mass Spectral Library) and literature data. For semi-quantitative analysis, the areas of peaks in the total ion chromatograms of the GC–MS analysis were integrated, and the relative proportion of each peak was expressed as a proportion of the total chromatogram area. Each aliquot was injected in triplicate, with the mean considered the result (only standard deviations below 5% were considered). From these data, the proportions corresponding to terpenes and fatty acids were calculated, and the values in mg per g of dry weight of leaf sample (mg/g DW) were extrapolated for these two types of compounds.

To assess hydrophilic compounds, the dichloromethane-extracted samples were extracted in a 50:50 (v/v) ethanol/water mixture in an ultrasonic bath at room temperature for 16 h. The suspension was filtered, and the resulting solution was used for the quantitative analysis of total phenolic content (TPC) and soluble sugar content (SSC). The TPC was determined according to the Folin–Ciocalteu method (Singleton and Rossi 1965). Briefly, 100 μL of the extract sample was added to 4 mL of diluted Folin–Ciocalteu reagent (1:10 v/v), followed by 4 mL of aqueous sodium carbonate (7.5 g/L). The mixture was incubated for 30 min in the dark and measured at 750 nm using a spectrophotometer (Shimadzu UV-160A). TPC was calculated from a standard curve with gallic acid (range, 0.014–0.762 mg/mL) and expressed as mg of gallic acid equivalents per dry gram of leaves (mg/g DW). The analyses were carried out in triplicate, with the averages considered the final results. SSC quantification was carried out using the phenol–sulfuric acid assay (Zhang et al. 2009). Aliquots of 0.1 mL of sample were reacted with 2.5 mL of concentrated sulfuric acid (96%) and 1 mL of phenol solution (5%, w/v). The solution was homogenized using a vortex mixer. After cooling, samples were evaluated at an absorbance of 490 nm using a spectrophotometer (Shimadzu UV-160A). The calibration curve was obtained with a D-glucose standard solution at six concentration levels (10–100 mg/L). The results were expressed as mg glucose equivalents per g of dry sample (mg/g DW).

Chi-square tests (χ²) were used to assess whether the distribution of the different immature stages of T. pityocampa among the different cohorts of pine leaves deviated from that expected under a random distribution. A two-way repeated-measures ANOVA was applied to assess differences in pine tree leaf-flush timing between Years and Areas, with trees as the repeated unit. For this analysis, the leaf-flushing date of the tip of each branch within each tree was transformed into a Julian date, and the average for each tree was calculated. The ANOVA was based on the Julian date value calculated for each tree.

Climatic variables and host-quality variables might operate simultaneously in shaping population differences. However, the absence of historical data on host-quality characteristics limits temporal assessment of resource changes, and the geographic restriction of the summer population to a single area limits spatial assessment of host-quality variation. Therefore, a formal statistical framework to partition the relative contributions of climate and host resources to population divergence is not feasible with the available data. Consequently, both hypotheses were evaluated independently by assessing whether observed differences between WP areas and the SP area in climatic conditions and host-resource quality were consistent with predictions of the abiotic hypothesis and resource-matching hypothesis, respectively. Both mechanisms likely contribute interactively to life-history evolution, and the separate analyses should be interpreted as complementary perspectives on these drivers rather than mutually exclusive explanations.

To test whether climatic differences support the abiotic hypothesis, six climatic variables were compared across the four study areas: (1) number of days in winter with minimum temperature below 0 °C (T0 °C); (2) number of days in winter with maximum temperature below 9 °C (T9 °C); (3) number of days in summer with maximum temperature above 30 °C (T30 °C); (4) number of days in summer with maximum temperature above 36 °C (T36 °C); (5) sum of radiation during winter (RadW); and (6) sum of radiation during summer (RadS). For day-count variables (T0 °C, T9 °C, T30 °C, and T36 °C), generalized linear models with a negative binomial error distribution and log-link function were fitted. For radiation variables (RadW and RadS), linear models were used.

To test whether host-resource quality differences support the resource-matching hypothesis, nine foliage chemistry and morphology variables were compared across the four study areas: (1) soluble sugars; (2) phenolic content; (3) leaf length; (4) water content; (5) carbon content; (6) nitrogen content; (7) carbon-to-nitrogen ratio; (8) fatty acids; and (9) terpene content. Linear mixed models with tree identity as a random intercept were fitted for variables with meaningful between-tree variability (soluble sugars, phenolic content, leaf length, and nitrogen content). Nitrogen data were log-transformed to meet model assumptions. For variables with negligible between-tree variance (carbon, carbon-to-nitrogen ratio, and fatty acids), linear models were used instead. All models included Cohort (leaf age class), Season, Area, and two interaction terms (Season × Cohort and Season × Area). These interaction terms were included to detect differences in seasonal effects between leaf ages and among areas, given that seasonal variation was expected to affect the size and composition of recently flushed leaves differently from older foliage and to be intrinsically linked with local environmental conditions.

In all applied models, pairwise comparisons of estimated marginal means were performed using Tukey’s post hoc adjustment to identify the source of variation. For comparisons between Areas, contrast analyses were performed to test whether the SP area differed significantly from the pooled WP areas. Applying contrasts after model fitting, rather than grouping populations a priori, preserved the full data structure, yielded more precise estimates of uncertainty, and enabled testing of biologically meaningful comparisons (SP vs. WP) while maintaining the full model structure. Model assumptions were assessed using the DHARMa package (Hartig 2024), and post hoc analyses were performed using the emmeans package (Lenth 2023). All analyses were performed in R (version 4.4.2).

The occurrence of first-instar T. pityocampa larvae was restricted to a short period of 2–3 weeks in all populations, although the timing of this period varied among distribution areas. In the two southern areas (WP_S1 and WP_S2), first instars were recorded during the last 2 weeks of October. In the SP distribution area, they occurred during the first half of July, whereas in the northern WP_N area, they were observed during the first 3 weeks of September. Fifth-instar larvae appeared in October in the SP, while in all WP areas they were recorded in February (Fig. 2).

Figure 2. 

Thaumetopoea pityocampa larval development in the four forests represented in Fig. 1. (WP_S1 and WP_S2, which exhibit similar patterns, are combined and represented as WP_S: The figure also indicates the timing of Pinus pinaster leaf sampling.

Overall, T. pityocampa females preferred younger leaves for oviposition, with Y-0 receiving more egg batches than expected by chance (χ² = 18.17, df = 3, N = 23, P < 0.001), with no egg batches deposited on Y-2, although, in the case of the SP, more egg batches were deposited on Y-1. In contrast, first instars were found to be more evenly distributed between Y-0 and Y-1 (Y-0: χ² = 5.67, df = 3, N = 17, P = 0.129; Y-1: χ² = 0.67, df = 3, N = 12, P = 0.881) but were never found feeding on Y-2. As larvae grew larger, Y-2 began to be used as food, except in the case of the SP, where only Y-0 and Y-1 were consumed. In fact, considering only the three WPs, during the fourth and fifth larval instars, larvae fed less than expected on Y-0 (χ² = 10.18, df = 2, N = 22, P = 0.006) and more than expected on Y-1 (χ² = 7.07, df = 2, N = 48, P = 0.029). No significant deviation from the expected distribution on Y-2 was found (χ² = 1.84, df = 2, N = 33, P = 0.399) (Suppl. material 2: fig. S8).

In studied areas, absolute minimum temperature were: WP_S1 –6.5 °C, WP_S2 –0.6 °C, SP –4.9 °C, and WP_N –3.6 °C. Average ± SE minimum temperatures of the two coldest months of the winter (December and January) were: WP_S1 4.9 ± 0.4 °C, WP_S2 8.3 ± 0.2 °C, SP 5.1 ± 0.3 °C, and WP_N 6.9 ± 0.4 °C. During the winter period, it was extremely rare for daily temperatures not to exceed 9 °C, which occurred on only three days in the entire 25-year climate series across all studied areas. Daily temperatures fell below 0 °C more frequently, averaging 12 days per year in WP_S1 and SP, which had the highest values, resulting in a significant contrast between the SP and the studied WPs. Additionally, average radiation between October and February was higher in the two southern areas (WP_S1 and WP_S2) than in the two northern areas (SP and WP_N), also resulting in a noticeable contrast between the SP and the studied WPs (Table 1; Suppl. material 2: fig. S7).

The number of days in summer when the temperature exceeded 30 °C was 3–5 times higher in the two southernmost areas, while days with more extreme temperatures, above 36 °C, were relatively rare but tended to occur more frequently in the southernmost area. Winter radiation was also higher in the two southernmost areas than in the two northernmost areas. Overall, these differences resulted in a significant contrast between the SP and the studied WPs (Table 1; Suppl. material 2: fig. S7).

The onset of leaf flush differed significantly between Areas and the 2 years of the survey (two-way ANOVA: Areas, F_3,35 = 4.07, P = 0.014; Year, F_1,35 = 144.8, P < 0.0001; Interaction, F_3,35 = 0.84, P = 0.482). The onset occurred earlier in 2020 (average ± SD, 11 March ± 4.8 days) than in 2019 (7 April ± 7.5 days). Contrast analysis indicated an earlier onset of leaf flush in the SP (15 March ± 13.9 days) than in the WPs (t₃₅ = –3.44, P = 0.002). No significant differences were found among the WP areas (28 March ± 18.6 days), according to pairwise comparisons with Tukey adjustment (P > 0.95 for all comparisons).

The leaf cohort Y-0 was shorter than Y-1, with a significant interaction between Cohort and Season due to the lowest values for Y-0 being found in July (Table 2). Moreover, Y-0 exhibited higher moisture content than Y-1, with a sharp decline during summer across all areas (Table 2). In July, contrast analysis indicated that trees in the SP distribution area had higher moisture content than trees in the WP distribution areas (Table 3; Fig. 3a) (see Suppl. material 2: fig. S9a, b for the complete data set).

Figure 3. 

Seasonal changes in the means and standard errors of several chemical parameters measured in Pinus pinaster leaves in the four forests represented in Fig. 1: a. Water content (% of fresh mass); b. Fatty acids (mg/g dry weight [DW]); c. Nitrogen (N; % dry weight); and d. C:N ratio. These parameters are represented because the analyses revealed significant contrasts between the SP and WP Thaumetopoea pityocampa populations (see Tables 2, 3). Data refer to averages of the two most recent leaf cohorts, Y₀ and Y₁.

Regarding the chemical composition of P. pinaster leaves, Y-0 and Y-1 showed similar levels of N, C, and C/N ratio (Table 2), with C being inversely correlated with N (Pearson r = –0.452, P < 0.0001). Generally, there was a decrease in N levels from July to October, followed by an increase in February (Fig. 3c), with the opposite trend in C/N (Fig. 3d). Contrast analysis indicated that N levels were generally lower and C/N ratios generally higher in the SP than in the WPs (Table 2), with differences being significant in October and February but not in July (Table 3; Fig. 3c, d) (see Suppl. material 2: fig. S9c–e for the complete data set).

The contents of fatty acids, sugars, phenols, and terpenes tended to be higher in Y-1 than in Y-0. Secondary metabolites included in the phenolic and terpenic fractions, as well as sugar content, generally increased from July to February, whereas fatty acid content showed the opposite trend (Table 2). Contrast analysis indicated that the SP had lower levels of fatty acids than the WPs, but this difference was significant only in July (Table 3; Fig. 3b) (see Suppl. material 2: fig. S9f–i for the complete data set).

On the west coast of Portugal, P. pinaster bud break occurs in early spring (mid-March to early April), with leaf elongation extending through summer. This follows the typical phenological pattern of Mediterranean woody plants, similar to that observed in other temperate regions (Maseyk et al. 2008). However, unlike cooler temperate climates, Mediterranean conditions, characterized by high summer temperatures, shift photosynthetic activity to the cooler, wetter winter months (Troeng and Linder 1982; Yang et al. 2020). Consistent with this pattern, an increase in foliar nitrogen from autumn through winter was observed, accompanied by a corresponding decline in C/N, since foliar nitrogen is largely associated with photosynthetic proteins (Harley et al. 1987; Epron et al. 1992; Maseyk et al. 2008). The subsequent decline in N likely results from its remobilization to support new tissue growth beginning in early spring (Coleman 1986; Turgeon 1989), while the concurrent rise in foliar C likely reflects carbohydrate accumulation under drought stress, when photosynthesis exceeds growth demand (Herms and Mattson 1992).

In T. pityocampa populations exhibiting the standard winter life cycle (WP_S1, WP_S2, and WP_N), larval development begins shortly after the end of the Y-0 needle elongation phase and coincides with rising foliar N levels and decreasing C/N ratios throughout autumn and winter, when the pine canopy holds its full complement of leaves. Consequently, larval growth occurs during the period of peak resource quantity and quality, with the final instars—those with the highest rates of food intake and body mass increase (Pimentel et al. 2011)—developing as N concentrations in pine foliage reach their maximum. Larvae complete development and pupate in February, before P. pinaster bud break and before leaf N begins to be remobilized for new growth. Previous studies have shown that extending larval development into spring results in smaller and less fecund adults (Pimentel et al. 2011), suggesting reduced host quality during this period (Heisswolf et al. 2009; Chorbadjian et al. 2019).

Nevertheless, the T. pityocampa WP life cycle involves trade-offs. Newly hatched larvae emerge at the end of summer, when foliar N is lowest and C/N ratios are highest, conditions that can limit survival, particularly for the sensitive first instars (Montgomery 1982; Hódar et al. 2002; Zalucki et al. 2002). Feeding on pine foliage, naturally poorer in N than the ancestral broadleaf hosts (Sardans et al. 2011), has likely favored prolonged larval development (Morehouse and Rutowski 2010). The early start of larval feeding may thus represent an adaptive compromise: exposing neonates to suboptimal summer foliage but ensuring that later instars can exploit the period of peak foliage quality in late winter and early spring.

This strategy contrasts sharply with that of most temperate forest Lepidoptera, which synchronize egg hatching with spring bud burst to exploit young leaves of high nutritional value (Feeny 1970; Chorbadjian et al. 2019). However, it is not entirely unique. The early instars of the brown-tail moth Euproctis chrysorrhoea (Linnaeus, 1758), a widespread defoliator of broadleaf trees, also feed in late summer and autumn before overwintering and resuming feeding in early spring, coinciding with bud burst (Frago et al. 2010). Although the role of foliage quality in this species remains unclear, it illustrates the diversity of life-history strategies that defoliators have evolved to synchronize with host phenology.

Thaumetopoea pityocampa females selectively lay their eggs on Y-0 foliage, which has higher water content and lower concentrations of defensive compounds such as phenols and terpenes than older needles. Since less-mobile neonates initially feed near the hatching site, this preference may partially compensate for the overall low host quality at the time of hatching, as these traits benefit neonates and early instars that are more susceptible to physical and chemical defenses (Zalucki et al. 2002). However, Y-1 needles, similar in N content but richer in other nutrients such as fatty acids and sugars, become the main resource for later instars that are better adapted to overcome plant defenses. Consequently, larvae shift from Y-0 to Y-1 leaves as they grow, with fourth- and fifth-instar larvae feeding on foliage of all ages. This pattern differs from that of other evergreen defoliators such as pine sawflies, which typically avoid current-year foliage because of its higher concentrations of secondary metabolites (Haack and Mattson 1993; Chorbadjian et al. 2019). The broader host use by T. pityocampa may reflect its later start in the growing season, when leaves are fully expanded, as well as physiological adaptations of Mediterranean pines to seasonal drought stress.

Leaves of P. pinaster trees from the forest hosting the SP exhibited the lowest N concentrations and highest C/N ratios among all studied sites (Table 3). In this forest, during the studied period, N content in February was even lower and C/N ratios higher than in July, contrasting with the pattern observed in the other three P. pinaster forests (Fig. 3c, d). Conversely, in July, needles had higher water content than those in forests hosting only WPs, potentially making them more suitable for early instars. The forest in the SP distribution area was also characterized by markedly earlier leafing phenology than the other sites. This earlier onset likely influences the seasonal dynamics of leaf chemistry. Although the causes of this advanced phenology remain unclear, they may involve local climatic or edaphic factors promoting an early growing season in host trees (Campôa et al. 2026). This could trigger early nitrogen remobilization from mature foliage, potentially explaining the unusually low winter foliar N levels.

Foliage chemistry is known to vary among years and to be influenced by climate (Jamieson et al. 2015). Because long-term data on host foliage chemistry are not available, it can only be assumed that similar host conditions have prevailed in the studied forests over the past decades, since foliage chemistry and morphology were assessed for only one season and leaf phenology for two. Nevertheless, the pattern of early phenology observed in the SP forest was consistent across two consecutive years, despite substantial interannual variability, suggesting a regional trend. If these conditions have been stable over time, they form the ecological background against which the SP evolved. If low foliar N in the SP forest during winter represents stable, long-term conditions, larvae developing during this period would potentially face poor nutritional conditions. Such a mismatch between defoliator and host phenology could have exerted selective pressure favoring a shift in larval development to summer. Early-summer hatching in the SP likely provides early instars with more favorable feeding conditions, as newly expanded needles are softer and more digestible than those in the other studied forests. However, throughout summer, needle chemistry in the SP distribution area is still characterized by relatively low N concentrations and high C/N ratios, indicating that nutrient limitation may continue to constrain larval growth and fecundity compared with WPs, whose larvae feed when foliar N peaks in winter. This suggests a potential trade-off between resource quality and developmental rate: while SP larvae feed on foliage of generally lower nutritional value, they complete development more rapidly and can attain higher population densities (Pimentel et al. 2006). Furthermore, the feeding trial of Rocha et al. (2020) showed that reduced leaf quality impaired larval development only in the WP, suggesting that SP larvae are better adapted to low-quality foliage, an adaptation consistent with the evolution of a summer life cycle under persistently poor winter foliar conditions in P. pinaster forests. However, these conclusions should be interpreted cautiously given the lack of long-term phenological data from the studied areas and the geographic restriction of the summer population to a single site, which limits independent replication of the summer phenotype.

The studied pine forests experience mild winters, with minimum temperatures well above the thresholds known to limit the survival and development of T. pityocampa larvae (Battisti et al. 2005; Buffo et al. 2007; Hoch et al. 2009). Such climatic conditions are typical of lowland coastal forests within the species’ central range and therefore do not, by themselves, explain the absence of winter-developing populations in the SP area. However, compared with WP sites, the SP region tends to experience more frequent frost events and lower solar radiation during autumn and winter (Table 1), conditions that could prolong larval development during this period (Pimentel et al. 2011; Poitou et al. 2021). This is consistent with the pattern observed in the northernmost winter population (WP_N), which shows a longer larval development period, starting more than a month earlier than the WP_S1 and WPS₂ populations (Fig. 2).

In contrast, SP develops in summer in an area where extremely hot days are relatively uncommon. These milder summer conditions may facilitate larval development during this season (Godefroid et al. 2016). However, because WP_N experiences similar summer temperatures (Table 1), climate alone cannot account for the absence of winter development in SP. Moreover, T. pityocampa larvae can tolerate high temperatures better than previously reported, with SP populations showing rapid local adaptation to thrive at temperatures exceeding 36 °C (Santos et al. 2011b). Thus, abiotic constraints appear insufficient to fully explain the observed phenological shift.

Instead, the shift from winter to summer development in SP may reflect the combined influence of climatic and host-related factors. In this region, cooler and less radiant winters, together with the reduced nutritional quality of pine foliage during this season, may have created selection pressure favoring earlier adult emergence and reproduction. Over time, this mismatch between larval development and host phenology may have driven the transition to a summer life cycle.

The authors have declared that no competing interests exist.

No ethical statement was reported.

The authors accept full responsibility for the content of the manuscript, including the disclosure of any use of AI.

No AI tools were used in the preparation of this manuscript.

This study was funded by the Portuguese Foundation for Science and Technology (FCT) as part of projects PTDC/BIA-ECO/31655/2017 and PTDC/ASP-SIL/29774/2017, which received additional funding from the COMPETE 2020 and Portugal 2020 programs through the European Regional Development Fund (FEDER). FCT also provided funding through PhD grants to J. Campôa (2021.05104.BD), M. Estorninho (2020.05618.BD), Q. Fortes (PRT/BD/154688/2022), and H. Patrício (2023.04037.BDANA); to R. Simões through a PhD grant from the SUSFOR Doctoral Programme (PD/BD/128259/2016); and to C.S. Pimentel through grant SFRH/BPD/80867/2011 and contract DL57/2016/CP1382/CT0009. This research was supported by the Forest Research Centre, a research unit funded by FCT (UIDB/00239/2025; https://doi.org/10.54499/UID/00239/2025).

C.S.P., O.C., C.M., I.M., and T.C. conceived and designed the study and obtained the funding; J.C., M.E., R.S., H.P., O.C., I.M., T.C., and C.S.P. participated in the field surveys and sampling; M.E., Q.F., R.S., H.P., and I.M. performed the chemical analyses; J.C. and C.S.P. analyzed the data and wrote the manuscript; and all authors edited and approved the manuscript.

J. Campôa https://orcid.org/0000-0002-5900-7085

M. Estorninho https://orcid.org/0000-0002-3719-1272

Q. Fortes https://orcid.org/0000-0002-3471-8825

R. Simões https://orcid.org/0000-0001-6539-9452

H. Patrício https://orcid.org/0000-0001-6075-8882

O. Correia https://orcid.org/0000-0002-1053-0561

C. Máguas https://orcid.org/0000-0002-4396-7073

I. Miranda https://orcid.org/0000-0003-0280-5219

T. Calvão https://orcid.org/0000-0002-9080-5173

C.S. Pimentel https://orcid.org/0000-0002-8364-8990

The data underpinning the analysis reported in this paper are deposited in the Figshare Data Repository at https://doi.org/10.6084/m9.figshare.33292368.