Nothofagus antarctica (G. Forster) Oerst. (commonly named 'ñire') is one of the main deciduous native tree species in Tierra del Fuego. It covers more than 200,000 ha ( Soler, 2012 ; Martínez Pastur et al. , 2013 ), and naturally occur in different environments near drier northern areas close to the ecotone with the Fuegian steppe. These forests are mainly used for sheep and cattle production ( Peri et al. , 2016a ). In this context, silvopastoral systems have become an economical, ecological and productive alternative for these forests ( Martínez Pastur et al. , 2012 ). These systems combine trees and natural grasslands or pastures under grazing in the same unit of land, and provide a diversification of farm income, either directly from the sale of timber and animals and/or indirectly by the provision of stock shelter and beneficial effects due to its potential on ecosystem service provisioning ( Peri et al. , 2016b ).

For silvopastoral systems in secondary forest stands, planned thinning may reduce the time required to yield products of a desired quality, concentrate growth on selected trees, increase wood production by utilizing trees that would die due competition (self-thinning) and also improve the understorey biomass production (and consequently increase animal production) by increasing incoming radiation. In Tierra del Fuego (dominant tree height >8 m, and rainfall >350 mm/yr) intensive thinning were proposed, leaving a 30-40% of canopy cover ( Peri et al. , 2016a , 2016b ). This remaining crown cover provides protection from desiccating strong winds, improving the microclimatic conditions for understorey plants growth and tree regeneration. However, the thinning not only provides synergies with other forest components, but also produce greater trade-offs with other ecosystem services ( Gyenge et al. , 2011 ; Martínez Pastur et al. , 2017 ; Peri et al. , 2017 ) and biodiversity components ( Lencinas et al. , 2008 , 2015 ; Quinteros et al. , 2010 ; Peri et al. , 2016a ). These changes are poorly studied in the long-term, and can be influenced by changes in the climate conditions ( Kreps et al. , 2012 ) or biological threats ( e.g . attacks of defoliators) ( Paritsis & Veblen, 2010 ; Piper et al. , 2015 ).

To date, there is enough information to support most of the silvopastoral proposals, including forest component, natural cycles, regeneration, understorey and biodiversity ( Lencinas et al. , 2008 , 2015 ; Quinteros et al. , 2010 ; Gyenge et al. , 2011 ; Ivancich et al. , 2011 , 2014 ; Soler, 2012 ; Soler et al. , 2013 ; Bahamonde et al. , 2013 , 2015; Martínez Pastur et al. , 2013 ; Gargaglione et al. , 2014 ; Echevarría et al. , 2014 ; Gönc et al. , 2015 ; Peri et al. , 2016a , b , 2017 ). However long-term results along the forest cycle, and its resilience deserve special attention. These long-term data can allowed to design new management alternatives ( Martínez Pastur et al. , 2016 ; Peri et al. , 2016c ). In this sense, long-term plots are essential to determine the economic feasibility of the intermediate treatments, define baselines and impacts of different silvicultural treatments and provide monitoring methodologies ( Peri et al. , 2016b ). For this, the objective of the present work was to analyse the effectiveness of the different levels of thinning on tree growth, remnant forest structure and microclimatic variables along seven years after the cuttings in a secondary Nothofagus antarctica forests in Tierra del Fuego, Argentina.

Original forest structure belongs to a typical secondary forests with 71-82% BA of dominant and co-dominant trees, with different degrees of health (28% BA of healthy trees, 51% BA of trees with minor damages and 21% BA of unhealthy trees). Trees were mostly affected by fungi and Misodendrum sp. attack. The measured forest parameters did not present significant differences (data not shown) prior thinning among the areas selected for each treatment: TD=2183-2793 trees/ha, DBH=12.4-14.3 cm, BA=34.0-38.7 m²/ha, and TOBV=163.0-190.5 m ³ /ha. After thinning, these variables presented significant differences over time due to the growth of remnant trees (Fig. 1). TD, BA and TOBV were significantly reduced from C > LT > HT, while DBH increased from C < LT < HT, and only DBH presented differences among the years. LT reduced in 75% TD, but only reduced 45% BA and 36% TOBV, while HT reduced in 87% TD, and 63% BA and 56% TOBV. Beside this, average DBH of the thinned stand increased 48% in LT and 73% in HT.

These changes in the forest structure influenced tree growth at individual and the stand levels (Table 1). An increase in the G-DBH (33% and 71% compared to C in LT and HT, respectively), and a decrease in G-BA and G-TOBV were observed in the control compared with thinned treatments (-12% and -8% for LT, and -38% and -32% for HT for G-BA and G-TOVB compared to C, respectively) (Fig. 2). The ratio between G-TOBV and G-BA greatly increased in thinned treatments compared with the control (70% in LT and 90% in HT). There was a great variation in growth over time, where the highest values occurred in the second year after thinning, and decreased in the following years to date. These growth rates influenced the class diameter distribution of the managed stands compared to the control (Fig. 3) where the individual size increased with the thinning intensity. Beside this, the recorded heavy attack of one defoliator, Ormiscodes amphimone (Fabricius), during the growing season of 2012-2013 greatly impacted on tree growth by decreasing the main studied variables (-18% G-DBH, -44% G-BA, -34% GTOBV, and -39% for G-ratio compared with the average of previous and following seasons). The impact of the defoliation mainly impacted HT (Fig. 1). However, other variables recovered the expected values in the following growing seasons.

After thinning, the crown cover and radiation presented significant differences due to the changes in the forest structure and tree growth (Figs. 4 and 5). CC and RLAI were significantly reduced from C > LT > HT, while radiation (DIR, DIF and TR) increased C < LT < HT (Fig. 4). LT reduced in 11% CC and 25% RLAI, increasing near 40% the radiation values. On the other hand, HT reduced in 27% CC and 57% RLAI, increasing more than 100% the radiation values inside the forests. Along the years, crown development increased the canopy cover and the RLAI, decreasing 25% of the radiation levels measured just after the thinning. These variables also were impacted by the attack of the defoliator obtaining -8% CC, -15% RLAI, and 12-19% radiation levels compared with the average of previous and following years (Fig. 6). The crown dynamic and radiation recovered the expected values in the following growing seasons as well as the growth variables.

These changes in the forest structure and the crown variables influenced on other micro-climatic variables (Fig. 7). Air temperature was the less influenced, however, it was observed that C treatment mitigate the extreme values during summer (low average values) and winter (high average values). Soil temperature and air humidity did not greatly changed between the C and LT treatments, while HT presented differences of -2°C and 15% compared with C, respectively.

Nothofagus antarctica forests are characterised by a lower timber values in Tierra del Fuego forests due to over-mature forest structures growing in low site quality ( Ivancich et al. , 2011 , 2014 ; Peri et al. , 2016b ), but greatest amount of understory biomass due to lower canopy closeness is produced ( Lencinas et al. , 2008 ; Soler, 2012 ). These characteristics determine that these forests were mainly used for sheep and cattle breeding. However, human activities during the last 100 years greatly changed the forest structure of many over-mature stands. Clear-cuts and human fires convert large areas in over-stocked secondary forests with low amount of understory biomass ( Soler, 2012 ; Soler et al. , 2013 ). These forest structures must be managed to recover the desired characteristics for silvopastoral management that appears as the most attractive economical alternative through thinning practices/intervention ( Peri et al. , 2016a ).

Thinning objectives varied with the silvicultural management defined for each forest type, e.g. in southern Patagonia thinning had been applied in N. pumilio and N. betuloides to increase specifically the quantity and quality of timber-wood ( Martínez Pastur et al. , 2001 , 2002 ; Peri et al. , 2002 , 2013 ). However, thinning for silvopastoral management must consider multiple objectives, e.g . in N. antarctica forests the interventions try to improve both the quality and quantity of timber, and at the same time enhance pastures biomass by increasing the radiation levels ( Martínez Pastur et al. , 2013 ; Peri et al. , 2016a , b ).

The proposed thinning levels for the studied secondary forests can be considered very intensive (75-87% tree removal) compared with other thinning proposals in the region ( Martínez Pastur et al. , 2001 , 2002 ; Peri et al. , 2002 , 2013 ), because the need of increase the radiation levels into the forests (50-100%). Also, at these latitudes, the thinning allows rainfall to reach the forest floor by reducing the canopy interception ( Caldentey et al. , 2005 ). These two variables (light and soil moisture availability) were the most influential ones over understory biomass growth ( Lencinas et al. , 2008 ; Quinteros et al. , 2010 ; Soler, 2012 ; Gönc et al. , 2015 ).

Canopy of N. antarctica quickly reacted to the cuttings by recovering the original levels in the light thinning treatment 4 years after cutting, as well as it was observed for other Nothofagus species ( Martínez Pastur et al. , 2002 ; Peri et al. , 2013 ). Heavy thinning favoured greatest diametric growth (tree variable) with the best G-ratio (stand variable) which maximized the performance of individual trees and increased by 100% the radiation levels. Considering these individual tree growth rates, the timber volume can be increased in the stand ( Ivancich et al. , 2014 ) and the radiation levels is enough to develop good-quality pastures for cattle breading ( Peri et al. , 2016b ). For this, heavy thinning levels offered the best combination for silvopastoral management purposes, maintaining for longer periods the benefits of the silvicultural interventions.

The value of resilience in forest ecosystem management has been widely accepted ( Yan et al. , 2011 ). Ecosystem resilience is one of the key target for sustainable management. This represents the ability of one ecosystem to absorb impacts before a threshold is reached where the ecosystem changes into a different state (ecological point of view), or the capacity of one ecosystem to return to its more-or-less exact pre-disturbance state (engineering point of view) ( Gunderson, 2000 ; Carpenter et al. , 2001 ). Nothofagus in general and particularly N. antarctica forests were described as resilient ecosystem considering a wide range of natural and anthropogenic disturbances ( e.g. Frangi et al. , 2015 ; Peri et al. , 2017 ). The studied forests were converted to pastures in the 50s of the last century by clear-cuts, but abundantly regenerated in the following years resulting in a secondary forests with complete crown cover. After thinning, trees quickly reacted in diameter and crown development. This determines frequent interventions to maintain the open degree of the canopy ( e.g . LT recover the canopy cover almost completely to date, while HT still present an appropriate openness degree). This reaction is desirable in terms of timber production, but increase the costs of interventions for silvopastoral purposes.

The recover after the caterpillar attack which significantly reduced the alive canopy is another good example of resilience of these forests, where after disturbance, the values arrived to the expected pre-disturbance levels. This kind of attacks can be related with a loss of growth, accelerating the die-back processes and mortality, related to C and N storage ( Piper et al. , 2015 ). Beside this, it is also interesting to observe (Fig. 4) that the attack was higher in the more intensive thinning, producing greater losses and lower recovery rate in the following season compared to LT. This can be explained by the changes in the microclimatic conditions ( e.g . increase of temperature) favouring higher consumption rate by the Ormiscodes amphimone larvae ( Paritsis & Veblen, 2010 ). These authors also relate an increase in the population of defoliators due to foliage quality, and the higher thinning can produce and increase in the quality of the tree leaves due to major availability of limiting resources ( e.g. soil nutrients, soil moisture and light) ( Soler et al. , 2011 ).

Tree diameter was the most useful variable for monitoring growth in forests with timber production purposes ( e.g. Martínez Pastur et al. , 2001 , 2002 ; Peri et al. , 2002 , 2013 ). However, in forests managed for silvopastoral purposes, the pastures allowance must be also monitored. This variable can be estimated using the tree crown cover as a proxy ( Peri et al. , 2006a , b ). Here we tested the use of hemispherical photographs for monitoring crown and radiation variables using simple models and available software ( e.g . Martínez Pastur et al. , 2011) . This methodology allowed to measured variation in these variables through the years, which greatly influenced on several ecological processes related to productivity in agroforestry systems ( e.g. moisture availability, nutrient cycles, decomposition, regeneration) ( Soler, 2012 ; Bahamonde et al. , 2013 , 2015 ; Martínez Pastur et al. , 2013 ; Peri et al. , 2016a ). Finally, these monitoring also allowed us to detect changes in the control treatment ( e.g. increasing crown cover), which can be related to the natural development of the stand ( e.g. Ivancich et al. , 2011 , 2014 ) or climate change effects ( Kreps et al. , 2012 ) which occurring in the last decades in Tierra del Fuego ( Ivancich et al. , 2012 ).

In summary, intensive thinning in Nothofagus antarctica secondary forests increased individual tree growth rates and doubling the radiation levels at the understory level, achieving the main objectives of the silvopastoral management. These forests exhibit a desirable resilience to forestry interventions and natural disturbances ( e.g. heavy defoliator attack), with a quickly reaction in the canopy cover growth. Monitoring of thinning for silvopastoral purposes must include easy and cheap measuring variables, e.g. diameter growth as a proxy for timber production objectives and hemispherical photos (crown cover and radiation) as a proxy for pasture production. Long-term monitoring allowed identifying reliable indicators, which can be used to develop new sustainable management alternatives.

To Benjamin Roberts and Juan Apollinaire of San Pablo Ranch (Tierra del Fuego) for their valuable collaboration in the establishment and monitoring of the long term plots in their land.