Climate influence seed production in managed and unmanaged Nothofagus pumilio forests of Southern Patagonia

Background: Annual seed production is key to understand natural forest dynamics and to apply sustainable forest management. This process is subjected to variations according to annual and seasonal climatic conditions, locally affected by El Niño-Southern Oscillation (ENSO) and the Southern Annular Mode (SAM) seasonality. Recognising how these variables affect the dynamics of harvested forests is useful for developing forest management strategies. Therefore, the objective of this study was to analyse annual seed production (SP) in


Introduction
Forest dynamics around the world are subjected to climate variability, being seeding a critical process for maintaining forest structure over time (Srur et al. 2018;Martinez Pastur et al. 2016;LaMontagne et al. 2020) within a context of climate change (IPCC 2022).In Tierra del Fuego, Argentina (52° 40' to 55° 03° S, 65° 07' to 68° 36' W), native forests are dominated by Nothofagus pumilio (Poepp.& Endl.)Krasser (commonly named lenga), a deciduous species that is naturally distributed in Argentina and Chile from 33º to 56º S (Veblen et al. 1996), growing as natural monospecific formations.This species abundantly regenerates with seeds after natural and human disturbances (e.g., harvesting), following natural dynamics based on gaps openings (Veblen 1989;Alan & Veblen 1993).The proposed management of these monospecific forests is based on the natural regeneration of the harvested stands (Martínez Pastur et al. 2019), by modifying light availability and soil moisture (Martínez Pastur et al. 2009, 2011b), where the remaining canopy overstory acts as a seed source creating favourable microclimatic conditions for seedling growth and survival (Chen et al. 1995(Chen et al. , 1999;;Martínez Pastur et al. 2011a).Therefore, seeding is an essential process to guarantee forest continuity (Rosenfeld et al. 2006;Martinez Pastur et al. 2011a) and conservation of the natural genetic resources (Marchelli & Gallo 1999;Premoli & Kitzberger 2005) being the first years-after-harvesting (YAH) crucial for seedling recruitment (Martínez Pastur et al. 2011a, 2011b).During the 21 st century, variable retention harvesting has been used in N. pumilio forests for timber purposes (extracting 30% basal area as aggregates and 15% as dispersed retention), combining economic and conservation objectives (Martínez Pastur et al. 2009, 2019).
Nothofagus pumilio forests concentrate seed fall at the beginning of the autumn with a strong interannual variation (Herrera et al. 1998;Torres et al. 2015;Rodriguez-Souilla et al. 2023a).This process is mainly regulated by biological and environmental factors (e.g., topography, soil conditions, and changes in the forest structure) (Hernández et al. 1992;Hacket-Pain et al. 2019;Keyejo et al. 2020), and also influenced by the regional climate during primordia development and flowering (McKee & Richards 1996;Martinez Pastur et al. 2008;Torres et al. 2015).In particular, seeding cycles are well described for other Nothofagus forest species (Alley et al. 1998;Monks & Kelly 2006), where the variation over the years (Kelly 1994;Kelly & Sork 2002) depends on forest structure characteristics such as stand basal area, crown cover and site quality (Övergaard et al. 2007;Soler et al. 2010;Bahamonde et al. 2013;Rodríguez-Souilla et al. 2023a).However, silvicultural treatments changes micro-climatic conditions (Promis et al. 2010), increase wind exposure that brings about less flower stability and reduced pollination (Cellini 2010;Toro Manriquez et al. 2016), competition among remnant trees (Martinez Pastur et al. 2008), and changes in nutrient and carbon pool dynamics (Chaves et al. 2023).These factors may exert varying impacts on the diverse silvicultural treatments, which are intricately linked to climate variability (Torres et al. 2015).
Seeding cycles and occurrence of masting events (Piovesan & Adams 2001;Schauber et al. 2002;Bogdziewicz et al. 2019) depend on physiological and climatic driving factors (Isagi et al. 1997;Liebhold et al. 2004;Bahamonde et al. 2011), particularly during the previous spring and summer (Richardson et al. 2005), but also years prior to seed dispersal (Lowry 1966;Eis 1973).The seed production response of N. pumilio to annual climatic variations have been previously reported (Richardson et al. 2005) in Tierra del Fuego (Torres et al. 2015), being strongly affected by both the Atlantic and Pacific oceans, the Antarctic climate, the Andes (Berman et al. 2013;Garreaud et al. 2013), and by large-scale climatic modes such as El Niño-Southern Oscillation (ENSO, with its phases La Niña and El Niño) and the Southern Annular Mode (SAM, with its positive and negative phases) (Silvestri & Vera 2009).Particularly, ENSO (with a periodicity ranging from 2 to 7 years) influences on precipitation patterns in the southernmost regions of South America, on both sides of the Andes Cordillera, particularly through the westerlies (Schneider & Gies 2004).Additionally, SAM exhibits considerable variability, with a decorrelation time of 10 days (Robinson 2000) and two phases characterized by significant seasonal variations (Fogt & Marshall 2020).SAM plays a crucial role in driving temperature and precipitation fluctuations, centered around the intrusion of cold air masses from Antarctica into middle latitudes.The occurrence of ENSO and SAM together (due to the teleconnection phenomena) can amplify climate variability in South America, especially during spring and summer (Silvestri & Vera 2003;Fogt & Bromwich 2006, 2011;Hill et al. 2009;Han et al. 2017;Lim et al. 2019).Besides, the nature of this relationship varies monthly (Vasconsellos et al. 2019).La Niña events tend to promote a positive SAM polarity more frequently, whereas El Niño events are associated with negative SAM polarity.Nonetheless, it is essential to acknowledge that other combinations, resulting from internal variability, may also occur, albeit less frequently (Fogt & Marshall et al. 2020).The utilization of climatic indices such as those developed for ENSO and SAM to comprehend the connections between climate and forest dynamics proves invaluable, especially in regions with limited access to climatic data.
Similarly, Fletcher (2015) demonstrated a distinct correlation between these climatic events and seeding patterns in New Zealand.Furthermore, Rodriguez Souilla et al. (2023b) applied these indices to assess regeneration height growth, while Srur et al. (2018) employed them to study the establishment of the N. pumilio tree line.In this context, the main objective of this study was to explore annual seed production (SP) in N. pumilio stands harvested under variable retention compared to unmanaged primary forests at Tierra del Fuego, Argentina, related to the activity of climatic modes over a 17-year period (2006-2022).We address the following questions: (i) Is there a differential response in SP between harvested and primary forests?
(ii) How does climatic variations brought about by ENSO and SAM activity influence SP over growing seasons?and (iii) Is there an interaction between silvicultural treatments and ENSO and SAM?
We hypothesised that climatic events with higher temperatures, especially during flowering and seed formation (spring season) act as a trigger for high SP, finding different responses between treatments due to different microclimatic conditions.

Study site
The study was conducted in monospecific Nothofagus pumilio forests located at Los Cerros Ranch (54º22' S, 67º51' W) in the central area of Tierra del Fuego (Argentina), covering an area of about 2,200 ha.In these forests, variable retention harvesting was implemented for timber production.This harvesting method retains a percentage of the original forest structure by leaving aggregates (a circular patch of 30 m radius per ha) and dispersed dominant trees (10-15 m 2 ha -1 of basal area) evenly distributed between the aggregates (Martínez Pastur et al. 2009).The studied forests have medium site quality (class II-III according to Martínez Pastur et al. 1997), with dominant tree height of 22-24 m, which presented a range of 700-900 m 3 ha -1 of total over-bark volume and 70-80 m 2 ha -1 of basal area before harvesting.This study frames within the longterm PEBANPA network of permanent sample plots (Parcelas de Ecología y Biodiversidad de Ambientes Naturales en Patagonia Austral, INTA-UNPA-CONICET, Argentina) (Peri et al. 2016).

Sampling design
Six stands (5-20 ha each) were selected for measurements at Los Cerros Ranch (Figure 1): three stands were harvested with variable retention, while three stands of undisturbed forests were considered as control (primary unmanaged forests, PF).Within the harvested stands, sampling plots were randomly located covering three different conditions according to the retention levels (see Martínez Pastur et al. 2011a): (i) within aggregate retention (AR) supposed not to present harvest effects, (ii) dispersed retention under the influence of AR (<20 m from the aggregate edges) (DRI), and (iii) dispersed retention outside the influence of AR (DR).A total of 72 sampling plots were selected (3 managed stands x 3 treatments x 6 plots, plus 3 primary forest stands x 6 plots), and monitored annually between 2006 and 2022 (Figure 1).Rodríguez-Souilla et al. (2023a) showed for the same experiment and time period, significant differences for stand variables among silvicultural treatments and time periods.
Litter traps (50 x 30 x 20 cm hard plastic baskets) were randomly placed in each plot within control and harvested stands, which were kept in their fixed locations throughout the monitored years, protected with sticks around them to reduce the possibility of disturbance by livestock.Each trap was collected every May and manually sorted in seeds and leaves of N. pumilio trees; other components of the litter were discarded (e.g., small branches, fungi, epiphyte plants, miscellaneous, and other understory species).The reason to collect litterfall in May follows the results of Martínez Pastur et al. ( 2008), who found that most of the seeds fall between March and April, and 90% of the leaves during April.Seeds were counted, and production at each sampling area was quantified (SP, million ha -1 year -1 ).

Regional climate data
Climatic data was obtained from the ERA5 (Hersbach et al. 2020) grid corresponding to the studied area during 30 years : mean, minimum and maximum monthly air temperatures (Tº = mean temperature, Tº MIN = minimum temperature, Tº MAX = maximum temperature) and monthly accumulated precipitations (PP = rainfall).For our purposes, climatic data were aggregated by season (WIN = winter, July-September; SPR = spring, October-December; SUM = summer, January-March; GS = full growing season, October-March), in order to capture the potential seasonal variations in climate patterns related to tree phenology.For the codes used for climatic variables see the list of abbreviations.
The years covering the post-harvest period (2006-2022) were characterised through the oscillation of occurrence of hemispheric modes (ENSO and SAM, in their joint behaviour) in order to assess their influence over seeding (divided in the season periods mentioned before).Values of ENSO (Southern Oscillation Index -SOI) and SAM (Marshall 2003) were obtained from open-access data (https://www.weather.gov/fwd/indicesand https://stateoftheocean.osmc.noaa.gov/atm/sam.php).Each period was classified according to its prevalence as positive (+) or negative (-) ENSO-SAM values (ENSO-/ SAM-, ENSO+/SAM-, ENSO-/SAM+, ENSO+/SAM+).Positive and negative ENSO values represent La Niña or El Niño events, for cold and warm phases at tropical latitudes, respectively.Positive SAM values indicate the contraction of the westerly wind belt towards Antarctica, whereas negative values signify its northward expansion.
The period 2006-2022 showed anomalies compared to the period 1993-2022.Air temperatures anomalies were +0.13 °C for winter, +0.17 °C for spring, and +0.17 °C for summer.Precipitation anomalies were -17 mm for winter, +10 mm for spring, while summer did not exhibit anomalies.The years and seasons with the highest temperature anomalies were: winters 2011,

Statistical analyses
Relationships among the 24 climatic variables and SP for each forest treatments (PF, AR, DRI, DR) were estimated by applying Pearson correlation (p <0.05).For that, ENSO and SAM average values for each season were considered.Then, univariate analyses were performed for comparisons.Different analyses of variance (ANOVA) were conducted, applying independent analyses sequentially to account for time-dependence of the data:

Annual seed production
Seed production annually varied with high (e.g., 2009,2012,2019,2022) and low values (e.g., 2010, 2014, 2020) (Figure 3).The average SP of the studied period was 9.35 million ha -1 year -1 for PF, 7.16 million ha -1 year -1 for AR, 2.25 million ha -1 year -1 for DRI and 1.08 million ha -1 year -1 for DR.All treatments showed a similar pattern, being in the order PF>AR>DRI>DR, except for some exceptional years when AR>PF (e.g., 2011 and 2017).Treatments showed significant SP differences  mainly between PF and AR when compared with DRI and DR (except 2021, where AR did not differ between retention levels).

Climate characterisation
Mean temperature values showed significant differences only for TºWIN (p = 0.018), being warmer for ENSO+ (Table 1).SAM+ was related to higher temperatures.Interaction did not occur for Tº SPR.Similar trends occurred for minimum and maximum temperatures, with significant interactions for Tº MAX SPR (p = 0.022) and Tº MIN WIN (p = 0.023) (Table 2).On the other hand, rainfall showed significant differences for the different SAM events, particularly for SPR and SUM, where SAMshowed higher values.ENSO/SAM significantly changed mean temperatures and accumulated rainfall depending on time of the year (Figure 4).Regarding temperatures, ENSO+/ SAM+ winters were related to warmer values.In spring, ENSO-/SAM-showed higher values, followed by ENSO+/ SAM-and ENSO+/SAM+.During the summer, the lowest temperatures were recorded in ENSO-/SAM-.Regarding rainfall, only the combinations with SAM-showed higher rainfall during winter and summer.

Seed production related to climate
There were significant correlations between climatic variables and SP for the studied treatments (Table 3).For example, rainfall (PP SUM and PP GS) only correlated negatively with SP in PF (-0.34 and -0.36 respectively).Minimum and maximum temperatures (Tº MIN SPR, Tº MIN SUM, Tº MAX SPR) showed positive (negative) correlations with maximum (minimum) temperatures across all treatments.ENSO and SAM showed both positive correlations with SP, especially for spring and summer.PF was the treatment that, on average, showed the highest Pearson values related to climatic variables.
Annual seed production varied according to seasonal climate and silvicultural treatment (Figure 5).SP in PF drastically changed with ENSO/SAM values for spring and summer, finding that the combination ENSO+/SAM+ triggers more SP, followed by ENSO-/SAM+, while during winter ENSO-/SAM-seed production was reduced.On the other hand, SP for AR did show significant differences during winter, with SP being favoured by ENSO+/SAM+.During spring and summer, ENSO+/SAM+ also presented high values for SP, although ENSO+/SAM-became more prominent during summer.DRI and DR showed similar patterns of SP, showing the highest values during winter and spring occurring ENSO+/SAM+ and ENSO-/SAM-, but not in summer, where, similar to AR, ENSO+/SAMwas the one that presented highest SP.
The multiple regression models that best fitted SP (2 models x 4 treatments = 8) had an r 2 -adj.between 73.2% and 85.2% (Table 4).Main trends showed that the equations included minimum and maximum winter and summer temperatures (PF, DRI and DR), as well as mean spring temperature (AR, DRI, DR).When incorporating ENSO and SAM values, ENSO SPR and ENSO GS prevailed.

Discussion
Our results showed interannual variation in SP for N. pumilio forests (Cellini 2010;Rodriguez-Souilla et al. 2023a) with an approximate 3-year cycle occurrence of high seed production pulses that seems enough to TABLE 1: Two-way ANOVA to test differences in mean temperature and accumulated precipitation (see codes on List of abbreviations), considering positive (+) and negative (-) ENSO and SAM seasonal occurrence as main factors.Different letters indicated significant differences using Tukey test at p <0.05.F = Fisher test, p = probability.

Index
Tº    (Martinez Pastur et al. 2009, Rodriguez-Souilla et al. 2023b).We showed that climate affects annual seed production, as shown in Table 3 and Figure 5.This could be explained by differences in the allocation of resources at different times of the year (e.g.carbon and other nutrients), as well as differences in growing season length and modification of tree phenological stages (Torres et al. 2015;Bogdziewicz et al. 2019).In our study, warm springs may lead to more uniform and abundant flowering and pollination success (Allen & Plat 1990), and therefore higher SP, associated to the critical phase of floral primordial development (Schauber et al. 2002;Martinez Pastur et al. 2007).In addition, warmer temperatures during spring (primordia development and flowering) and summer (seed formation) would drive greater net carbon uptake (Allen & Plat 1990;Richardson et al. 2005).Although further studies are needed to extrapolate these relationships to a larger geographical scale, the results obtained are consistent with the hypothesis proposed in this study.Thus, minimum and maximum temperatures explained SP, having a negative effect for minimum temperature and a positive effect for maximum temperatures, especially during spring and summer seasons, as was reported by Torres et al. (2015).TABLE 4: Regression models for annual seed production and climatic variables (codes on List of Abbreviations) for each forest treatment (PF = primary forests, AR = aggregated retention, DRI = dispersed retention under the influence of AR, DR = dispersed retention without influence of AR).p = probability, r 2 -adj = coefficient of determination.SEE = standard error of estimation (million ha -1 year -1 ), MAE = mean absolute error (million ha -1 year -1 ).The seasonal influence of simultaneous ENSO and SAM events on SP has also been demonstrated in the present study, where growing season was restrained to the summer months (Villalba et al. 2010) and regional climate determined its duration.In consonance with climatic variables (Tº and PP), ENSO+ and SAM+ events tend to favour high SP for N. pumilio forests.This is consistent with Srur et al. (2018) who reported that higher temperatures may generate greater seedling abundance and growth.Also, the effects of ENSO and SAM on physiological processes in different forests has been reported (Schauber et al. 2002;Hadad et al. 2021), being modified according to how their seasonality in temperature and rainfall affect different parts of the world.Besides, studies have reported that higher temperatures associated to low rainfalls are correlated with years of high SP, especially during the spring (e.g., flowering time reported by McKee & Richards 1996;Buechling et al. 2016;Gallego Zamorano et al. 2018).Low winter temperatures can extend the dormant season, therefore limiting physiological processes that affect annual flowering development (Fritts 1976), soil nutrient acquisition (Smaill et al. 2011), which in turn reduces SP.
The main differences in SP between treatments occurred between PF and harvested treatments (AR presented mixed effects compared to other treatments).This is likely due to the microclimatic conditions generated in harvested stands (Chen et al. 1999;Martinez Pastur et al. 2007;Rodriguez-Souilla et al. 2023a) by altering flowering and fruiting of the remaining trees (Rodríguez-Souilla et al. 2024).In turn, the models obtained in this study can account for the influence (positive or negative) of climatic variables on SP, including ENSO and SAM indices to improve predictions.While rainfalls were not a limiting factor for SP, temperatures (mainly for harvested) in SAM events play a forcing role on its variability.So, SP in PF was influenced by the growing season accumulated precipitation, effect also reported by Smaill et al. (2011) for Nothofagus solandri var.cliffortioides forests In the context of climate change with periodic increases in temperature and variations in annual rainfall (IPCC 2022) and the occurrence of climatic extreme events, it is likely that annual seed production could be altered (Monks et al. 2016).In addition, the cycle and frequency of masting events for Nothofagus forests (detected only for the year 2012 in this work) might be modified under climate change conditions (Cuevas 2000(Cuevas , 2002;;Övergaard et al. 2007), although it may be insensitive to increases in global mean temperature (Kelly et al. 2013;Hacket-Pain & Bogdziewicz 2021), where constant heavy flowering or desynchronization may happen.The findings presented in this study may have important implications for forest management as a tool for understanding forest dynamics related to seeding, a key process for forest regeneration (Rodriguez-Souilla et al. 2023b) in a context of high climate variability.
(i) 17 one-way ANOVAs to analyse SP for each year of the studied period considering forest treatments as main factors;(ii) two-way ANOVAs to analyse temperatures and rainfalls values considering ENSO and SAM as main factors during the studied seasons;(iii) 6 one-way ANOVAs to analyse climatic factors (2: temperatures and rainfall values) considering ENSO and SAM as main factors during the studied seasons (3: winter, spring and summer); and(iv) 12 two-way ANOVAs to analyse SP considering ENSO and SAM as main factors for each forest treatments (4: PF, AR, DRI, DR) during the studied seasons (3: winter, spring and summer).

FIGURE 1 :
FIGURE 1: Locations of sampling sites.Primary forests (PF) and treatments (AR= aggregated retention, DRI = dispersed retention under the influence of AR, DR = dispersed retention outside the influence of AR).

FIGURE 2 :
FIGURE 2: Boxplot for basal area (BA, m 2 ha -1 ) and crown cover (CC, %) in the different forest treatments (PF = primary forests, AR = aggregated retention, DRI = dispersed retention under the influence of AR, DR = dispersed retention without influence of AR).

TABLE 2 :
Two-way ANOVA to test differences in mean minimum and maximum temperature (see codes on List of Abbreviations), considering positive (+) and negative (-) ENSO and SAM seasonal occurrence as main factors.Different letters indicated significant differences using Tukey test at p <0.05.F = Fisher test, p = probability.

TABLE 3 :
Pearson correlation coefficient (PC) and p-value (p) for seed production for each forest treatment (PF = primary forests, AR = aggregated retention, DRI = dispersed retention under the influence of AR, DR = dispersed retention without influence of AR) against different climatic variables (see codes on List of Abbreviations).Significant p values are highlighted in bold.