**Matteo Pecchi 1, Maurizio Marchi 2,\* , Marco Moriondo <sup>3</sup> , Giovanni Forzieri <sup>4</sup> , Marco Ammoniaci <sup>5</sup> , Iacopo Bernetti <sup>1</sup> , Marco Bindi <sup>1</sup> and Gherardo Chirici <sup>1</sup>**


Received: 20 July 2020; Accepted: 24 August 2020; Published: 26 August 2020

**Abstract:** Forests provide a range of ecosystem services essential for human wellbeing. In a changing climate, forest management is expected to play a fundamental role by preserving the functioning of forest ecosystems and enhancing the adaptive processes. Understanding and quantifying the future forest coverage in view of climate changes is therefore crucial in order to develop appropriate forest management strategies. However, the potential impacts of climate change on forest ecosystems remain largely unknown due to the uncertainties lying behind the future prediction of models. To fill this knowledge gap, here we aim to provide an uncertainty assessment of the potential impact of climate change on the forest coverage in Italy using species distribution modelling technique. The spatial distribution of 19 forest tree species in the country was extracted from the last national forest inventory and modelled using nine Species Distribution Models algorithms, six different Global Circulation Models (GCMs), and one Regional Climate Models (RCMs) for 2050s under an intermediate forcing scenario (RCP 4.5). The single species predictions were then compared and used to build a future forest cover map for the country. Overall, no sensible variation in the spatial distribution of the total forested area was predicted with compensatory effects in forest coverage of different tree species, whose magnitude and patters appear largely modulated by the driving climate models. The analyses reported an unchanged amount of total land suitability to forest growth in mountain areas while smaller values were predicted for valleys and floodplains than high-elevation areas. Pure woods were predicted as the most influenced when compared with mixed stands which are characterized by a greater species richness and, therefore, a supposed higher level of biodiversity and resilience to climate change threatens. Pure softwood stands along the Apennines chain in central Italy (e.g., *Pinus*, *Abies*) were more sensitive than hardwoods (e.g., *Fagus*, *Quercus*) and generally characterized by pure and even-aged planted forests, much further away from their natural structure where admixture with other tree species is more likely. In this context a sustainable forest management strategy may reduce the potential impact of climate change on forest ecosystems. Silvicultural practices should be aimed at increasing the species richness and favoring hardwoods currently growing as dominating species under conifers canopy, stimulating the natural regeneration, gene flow, and supporting (spatial) migration processes.

**Keywords:** ecological modelling; Mediterranean area; forest management; future spatial projection; silviculture

#### **1. Introduction**

Climate change represents an important challenge for ecologists, biologists, and modelers whose research interest is the study of the potential effect of climate change on ecosystem services provided by forests [1–4]. The use of predictive models and statistical tools in scientific literature has increased since the 1980s [5–7], aimed at stimulating the most likely effect of climate change. A predicted spatial movement of ranges and suitable envelopes has been often the main result in many research papers. This shift across a geographic or an altitudinal gradient [8] represents one of the possible responses of forest tree species to climate change [9,10]. The colonization of a new environment depends to the landscape fragmentation, species-specific seed dispersal ability, as well as the nutrient availability in the new environment [11]. However there is scientific evidence that this is already underway both in altitude [12,13] and in latitude [14,15]. In a climate change framework, forest management and planning efforts must be oriented toward maintaining and improving biodiversity and ecosystem services, assuring the long-term availability of forest resources and their biological functioning [16,17]. The development of a sustainable forest management strategy (SFM) is a very urgent topic in forestry and environmental sciences for human well-being [18–20] and for carbon sequestration purposes [21,22]. Information about the ecological requirements of different tree species are fundamental for its implementation [23–25] allowing conservation plans, ecological restoration actions [26], as well as the detection of threatened areas and also possible refuges [27–29].

The ecological modelling of the spatial distribution of living organisms, both animals and plants, is currently known as Ecological Niche Modelling (ENM) or Species Distribution Modelling (SDM). These techniques can be used to link the spatial distribution of a target species with some ecological drivers (climate, soil data, etc.) often extracted in a GIS environment. The gathered information is then used as input data for mathematical models where the ecological information is used as predictor (i.e., independent variable). Even if sometimes criticized as not a reliable predictive method in a changing climate [30], they still represent the most used tool to support forest management strategies worldwide [31–34]. SDM/ENM are statistical algorithms which have provided greater flexibility and good performance in deriving and modelling the ecological requirements of single species or ecological groups from its spatial distribution, assuming an equilibrium with climate. When future scenarios prediction is the final aim of SDMs, many uncertainties lay behind the final prediction [35]. These uncertainties can be summarized into three main sources: (i) the parameter uncertainty i.e., imperfect species occurrence data, unavailableness of important predictor variables; (ii) the model uncertainty, that it is linked to the choice of different SDM algorithms and their complexity; (iii) the climate uncertainty which includes both the interpolation error and the climate change scenarios uncertainties [36–38]. To deal with uncertainties, many modelling efforts were developed such as probabilistic predictions [39] and the use of ensemble modelling strategies [40–42] where average models are calculated from different algorithms, deriving confidence intervals and weighted means according to the predictive power. While variability between different modelling techniques is typically low [31], the variability in climate data is more relevant [36,38]. The variability in climate data is strictly associated with general circulation models (GCMs) and regional circulation models (RCMs) patterns for the same study area. This is one of the main issues responsible for the wide range of results obtained by different research groups worldwide and often on the same environment of forest tree species [31,43–45]. Overall RCMs are believed to be more reliable for small-scale studies thanks to the statistical downscaling procedures involved in their development from original coarse GCMs where novel climatic surfaces with higher spatial resolution are obtained [46–49]. In addition to downscaled surfaces and RCMs, specific downscaling tools for custom queries [50–52] are often used in literature.

According to the provided evidence, many uncertainties are still masked under the predictions generally provided by researchers in their studies, with climate as one of the main drivers. At the current time the SDM technique has been successfully used in Italy for some occurring tree species using National Forest Inventory data [27] or wide-range projections and broad spatial distribution data and forest categories [53], or spatial analysis on species richness [54]. However an extensive study on the whole country for a wide range of tree species and evaluating the modelling uncertainties is still missing The aim of this paper is to evaluate the uncertainties behind an SDM procedure in the Mediterranean environment (Italy) to support future SFM strategies. In this work, several future scenarios for 19 species, among the main forest tree species in Italy, were realized using six GCMs and one RCM, quantifying the discrepancies between them and within species when different climatic data are used. Suitability maps were obtained for Italy to provide indications to forest planners regarding the possible consequence and impact of climate change in Italian forest systems. Then adaptive forest management strategies were proposed dealing with potential impacts of climate change and uncertainties detected behind the modelling efforts.
