Episodes of drought-induced tree mortality have been recently observed worldwide ( Allen et al. , 2010 ). Under a climate change scenario, those effects may boost potential vegetation changes over large geographical areas, with major impacts on forests diversity and ecosystem function ( Carnicer et al. , 2011 ). Declination has also been reported in oak forests ( Haavik et al. , 2015 ), including the Persian oak forests ( Quercus brantii Lindl. var. persica (Jaub & Spach) Zohary) in Zagros Mountains ( Fatahi, 1995 ). Persian oak covers more than half of the Zagros forest area (west Iran), representing the most important tree species of this region ( Bordbar et al. , 2010 ; Hassanzad-Navroodi et al. , 2015 ). These forests support valuable ecological and economic resources, as for instance the use of seeds in traditional medicine, fuel wood, charcoal, and timber hardwood ( Fatahi, 1995 ). As a consequence, conservation of Persian oak forests is an impending management concern, currently challenged by drought-induced tree mortality and insect outbreaks.

Oak decline is a disease complex that involves the interaction of several biotic and abiotic factors (see Manion, 1991 ). The effects of decline on oak forests represent one of the most serious forest disease issues, with impacts that range from partial crown dieback to tree death ( Starkey & Oak, 1989 ; Brasier, 1996 ; Rizzo & Garbelotto, 2003 ; Lloret et al. , 2004 ; Heitzman et al. , 2007 ; Fan et al. , 2008 ). Due to the increasing risks (beetles attacks, root fungi, drought events, etc.,), forest owners and stakeholders require new adaptation guidelines to increase the resilience of those oak forests subjected to potential decline ( Cescatti & Piutti, 1998 ; McEwan et al. , 2011 ; Haavik et al. , 2015 ).

Tree mortality drives the functional and structural dynamics of forest ecosystems, both as a sink/source of carbon and nutrients, and as a mechanism of forest structure and diversity change ( Franklin et al. , 1987 ; Allen & Breshears, 1998 ). Understanding tree mortality patterns is crucial to understand current forest community composition and to predict changes in stand structure and species composition ( Chapman et al. , 2006 ; Heitzman et al. , 2007 ; Galiano et al. , 2010 ; Carnicer et al. , 2011 ; Lechuga et al. , 2017 ). At a regional scale, tree mortality may be mainly influenced by climate, for instance extreme events, such as drought and high temperatures ( Anderegg et al. , 2012 ). Tree mortality can occur across a wide range of spatial and temporal scales, from the gradual death of individual trees ( Ogaya & Peñuelas, 2007 ; Camarero et al. , 2016 ) to the abrupt mortality of thousands of trees over large areas due to severe disturbance events ( Allen et al. , 2010 ).

As regards the need to improve our understanding of forest response to the current climate change, it might be of great importance to investigate site and stand structure effects, buffering or enhancing decline process of forests decline ( Heitzman et al. , 2007 ; Camarero et al. , 2016 ; Colangelo et al. , 2017a ). As regards the specific case of oak trees, water deficit is the most common stressing factor related to physiological weakness ( Jenkins & Pallardy, 1995 ; Lloret et al. , 2004 ; Ogaya & Peñuelas, 2007 ; Fan et al. , 2008 ; Haavik et al. , 2015 ), triggering episodic tree declines and mortality ( Fan et al. , 2012 ; Keyser & Brown, 2016 ; Sánchez-Salguero et al. , 2017 ). Furthermore, following drought, other events, such as beetle outbreaks, affect mortality of oak trees or increase its severity ( Kabrick et al. , 2008 ; Fan et al. , 2008 ; Hosseini, 2012 ). As a result, mortality events are usually associated with complex interactions among endogenous factors, such as human use-related stand characteristics and genetics, and exogenous factors that include site conditions, insects and pathogens ( Heitzman et al. , 2007 ; Urbieta et al. , 2008 ; McEwan et al. , 2011 ; Haavik et al. , 2015 ).

The effects of drought are more common visualized at the regional-scale ( Keyser & Brown, 2016 ; Sánchez-Salguero et al. , 2017 ), but its influence on crown dieback and tree death will be modulated at the local-scale, where factors such as topography, soil and stand structure might be as important as regional climate to understand oak forests dynamics ( Jenkins & Pallardy, 1995 ; Chapman et al. , 2006 ; Gea-Izquierdo et al. , 2009 ; Galiano et al. , 2010 ). Topographic characteristics such as slope and aspect also strongly influence soil moisture, and therefore drought-induced tree mortality may vary accordingly ( Das et al. , 2008 ). Consequently, spatial and temporal variation in site moisture conditions may influence the spatial pattern and severity of drought-induced forest decline, while tree mortality patterns at local scale are very difficult to forecast ( Das et al. , 2008 ; Anderegg et al. , 2012 ).

Tree decline has been related to stand structure, otherwise supporting that tree response to water stress also relies on tree-to-tree competition ( Das et al. , 2011 ; Lechuga et al. , 2017 ). Episodic tree mortality is usually related to dry site conditions and high stand density ( Chapman et al. , 2006 ; Gea-Izquierdo et al. , 2009 ). Specifically, trees subjected to higher competition show lower radial growth and they are more prone to die following extreme drought events ( Das et al. , 2008 ; Galiano et al. , 2010 ; Linares et al. , 2010 ; Colangelo et al. , 2017b ). As a result, contrasting stand structure may act modulating the individual tree responses to drought ( Lechuga et al. , 2017 ). Persian oak forests display a wide range of stand structural heterogeneity ( Erfanifard et al. , 2009 ), which might be, at least in part, due to contrasting land use practices ( Fatahi, 1995 ; Urbieta et al. , 2008 ). Hence, understanding the effects of drought and tree mortality must take into account stand structure in a spatially explicit context ( Das et al. , 2008 ).

Extensive oak mortality is a significant threat to Zagros forests health and economic value of their inhabitant. In this study, we attempt to investigate recent climate trends and the effect of topography, soil, stand structure and tree-to-tree competition on widespread tree mortality in Zagros forests in west of Iran. Specifically we aim to (1) quantify the extent to that recent drought events support a long-term increasing aridity at regional scale; (2) identify stands that are at high risk for oak decline, as regards site conditions (elevation, aspect, slope, soil); and (3) identify the mortality pattern for individual trees within declining stands, according to individual characteristics (size, coppice, standard, etc.) and tree-to-tree competition.

The study area is mainly comprised by forests dominated by Quercus brantii var. persica Lindl., Pistacia atlantica Desf., Acer monspessulanum subsp. cinerascens (Boiss.) Yalt., Crataegus pontica C. Koch, Amygdalus orientalis L., and Cerasus microcarpa C.A. Mey. Boiss. These mixed oak forests grow on contrasting aspects, slopes, and elevations, ranging from 1500 to 2000 m a.s.l. ( Fatahi, 1995 ). The limits of the study area were determined on a topographic map with a scale 1:50000, to account for a wide range of environmental variables, and contrasting forest stand structures and mortality. Geographic information system analyses were performed by using ArcGIS 9 ArcMap version 9.2 ( ESRI, 2006 ) to investigate digital layers of slope, aspect and elevation. Aspect data were grouped within four categorical factors: N, north; S, south; W, west; and E, east aspect, respectively. Slope data were grouped within three categorical factors: 0-30%, 30-60%, > 60%. Elevation data were grouped within two categorical factors: 1500-1700 and 1700-2000 m a.s.l. The obtained categorical factors were overlapped, accounting for 24 combinations of topographic units (TUs; 3 slopes × 4 aspects × 2 elevation categories, respectively), however only 20 combinations of TUs were suitable for further field sampling (Fig. 1; Table S1 [suppl.]).

Each TU was sampled in the field by mean of three circular plots, 1000 m ² surface, randomly established (Fig. 2). Within each plot, site characteristics (slope, aspect, elevation) were measured. We performed a soil profile per TU to measure soil properties. Soil depth was measured in the field, while physical and chemical soil properties were measured in the laboratory using a soil sample per TU ( Sparks et al. , 1996 ). We determined texture and soil moisture, total nitrogen percentage (N), organic carbon percentage (OC), organic matter percentage (OM), and total neutralizable material percentage (TNV, CaCO ₃ ).

In each plot, we recorded all tree and shrub species and measured its diameter at 1.3 m from the ground (DBH) for all trees above 5 cm DBH, maximum and minimum crown diameter, crown dieback percentage, and mortality density. Tree habit (coppice or standard) was also noted (coppice trees are considered those originated by new growth from the stump or roots after cutting), using for coppices equivalent diameter estimation ( Grier et al. , 1992 ):

The study area ranges from 1510 to 1880 m elevation. Soil samples texture was mainly clay (Fig. S1 [suppl.]), with mean values of total organic carbon and nitrogen percentages of 1.66% and 0.17%, respectively, while neutralizable material percentage was on average 43.6%. Persian oak is the dominant species within the study area, accounting for 95.3% of the total tree density and 93.6% of the total basal area (Table S2 [suppl.]). Crataegus pontica and Pistacia atlantica were the second and third most abundant tree species; the relative contribution of P. atlantica to stand basal area was higher, than that of C. pontica . Acer cinerascens showed the lowest stand density and basal area among trees as well as the highest mortality rate (48%), while Amygdalus orientalis and Cerasus microcarpa were minor components among the shrub strata (Table S3 [suppl.]). The average stand canopy cover was 32.5%, the majority of which belongs to Q. persica (31.5%). Many trees in the study area (especially Q. persica ) are sprouts of coppices. Mortality was significantly higher for coppices, expressed as tree density: 23.83 trees/ha vs . 11.5 trees/ha, for coppices and standard respectively (t-test, p =0.002; Table 1). Nonetheless, given that total density of coppices was also higher, expressed as percentage, mortality rate was higher among the standard growth form (Table 1), however, this difference on mortality percentage was not significant (t-test, p =0.106).

The most frequent tree diameters were those among 15-25 cm DBH classes (Fig. 4a), as well as the most frequents tree sizes of dead trees (Fig. 4b), nonetheless, basal area percentage of dead trees was higher among tree diameters belonging to 25 and 30 cm DBH classes (Fig. 4c). The mean DBH decreased, while stand mortality increased, as stand density becomes higher (Fig. 5a; Table S4 [suppl.]). The competition index (CI) of alive trees increased significantly according to stand density (Fig. 5b; p <0.02), while the CI of the dead trees was significantly higher than that of the alive trees (within plots) but was not related to stand density (among plots; Fig. 5b, see also Tables S5 and S6 [suppl.]).

Crown dieback was also significantly related to stand density, similar to mortality did; thus, moderate and severely defoliated trees increasing as stand density was higher. As regard environmental variables, total stand density was positively correlated to elevation (Fig. S2 [suppl.]). Nonetheless, expressed as percentage, higher elevation stands showed the lower fraction of healthy trees, as well as the higher fraction of trees with crown damages (Fig. 6). However, mortality rate ( i.e. , expressed as percentage) was only marginally significant regarding elevation, supporting that the main effect relies on stand density (Fig. 6; Fig. S2 [suppl.]). This result was also supported by multiple regression analyses (Table 2), where mortality percentage was significantly related to aspect (higher mortality percentage at southern and western aspects), but did not showed significant relationship with elevation or slope. According to soil variables, mortality percentage was higher on shallower soils, while regarding stand structural variables, total basal area showed a positive relationship (and negative for mean DBH), indicating that mature stands, with higher mean DBH and lower basal area, show lower mortality rates. On the other hand, canopy cover showed a negative relationships, i.e. higher canopy cover is related to lower mortality percentage, likely reflecting the covariance of defoliation and mortality (Table 2).

The extent to which site factors such as topography and aspect influence the severity Persian oak decline is poorly understood. Several researches performed on different oak forests have reported that shallow soils and/or dry aspects account for the greatest amount of oak decline ( Starkey & Oak, 1989 ; Rizzo & Garbelotto, 2003 ; Heitzman et al. , 2007 ; Fan et al. , 2008 ; Colangelo et al. , 2017a ). Trees growing in shallower and poorer soils appear to persist close to their limits of climatic tolerance, while populations in wetter and fertile soils sites may have larger physiological buffers ( Johnson et al. , 2016 ). Shallow soils are frequent at higher elevations, likely affecting the obtained spatial patterns of tree mortality. Drought effect is enhanced by lower water availability and limited rooting depth, as may occur in clay soils (Fig. S1 [suppl.]), enhancing drought sensitivity of trees ( Johnson et al. , 2016 ). Despite in our study area mortality was inversely related to soil deep (Table 2), there were not consistent patterns regarding soil texture or the amount of organic material and nitrogen. These results suggest that factors other than (or perhaps in addition to) soil properties, topography and aspect, such as species composition, tree age and size structure, and competition, are important determinants of oak decline, likely confounding climate- and site-related influences ( Ogaya & Peñuelas, 2007 ; Bordbar et al. , 2010 ; Lechuga et al. , 2017 ).

Persian oak decline likely involves the interaction of multiple biotic and abiotic factors, with impacts that range from partial crown dieback to tree death. Such decline complex may be framed in terms of predisposing factors, inciting factors, and contributing factors ( sensu Manion, 1991 ). Some suggested predisposing factors include high density of lower-size DBH classes, shallow and rocky soils (Table 2). As regards the inciting factors, it is likely that regional drought spells are related to oak decline ( Jenkins & Pallardy, 1995 ; Fan et al. , 2012 ; Keyser & Brown, 2016 ), while contributing factors may include different oak borers and diseases ( Haavik et al. , 2015 ).

An important indicator of the health of a tree is the condition of its crown ( Rizzo & Garbelotto, 2003 ; Lloret et al. , 2004 ; Ogaya & Peñuelas, 2007 ). Declining trees are initially point out by foliage wilt and browning followed by progressive branch dieback ( Starkey & Oak, 1989 ). Trees with vigorous, healthy crowns tend to have higher growthrates. By contrast, trees with damaged crowns have a reduced capacity for photosynthesis ( Brasier, 1996 ). Many stressors have been related with crown dieback including site factors and tree-to-tree competition ( Heitzman et al. , 2007 ; Camarero et al. , 2016 ; Colangelo et al. , 2017a ). In our study, elevation showed a significant relationship with crown dieback (Fig. 6), nonetheless, expressed as percentage, mortality rate was not related to elevation, supporting that the main effect relies on stand density.

We thank the Handling Editor and two anonymous reviewers whose throughout revision and comments greatly helped to improve a previous version of the manuscript.