National assessment of throughfall sensitivity to changes in storm magnitude for the forests of Iran


Pedram Attarod

Forestry and Forest Economics Department, Faculty of Natural Resources, University of Tehran, Iran.

Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, China.

Qiuhong Tang

Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, China.

John T. Van Stan II

Department of Geology and Geography, Georgia Southern University, USA.

Xingcai Liu

Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, China.



Aim of study: To understand throughfall (TF) sensitivity to variability in rainfall amount (Pg) for typical forest sites across the main climate types of Iran.

Area of study: Nine forest stands of several common native and introduced tree species situated in all common Iranian climate types, but located primarily in northern Iran.

Material and methods: A nondimensional relative sensitivity coefcient was employed to predict responses of TF to Pg changes. Projected Pg changes over the measurement sites for the period 2020-50 were estimated using one of the Coupled Model Intercomparison Project phase 5 (CMIP5) known as HadGEM2-ES under low and high emission scenarios (RCP 2.6 and 8.5).

Main results: TF displayed strong positive linear relationships with Pg at all sites [TF=0.66 Pg-0.16; R2=0.91]. The sensitivity coefcient ranged from 0.96-2.35 across the nine forest sites and large sensitivity coefcient differences were found between small (< mean annual Pg ) and large (> mean annual P g ) storms for arid and Mediterranean plantations. Shifts in P g and increased small storm frequency are predicted for these regions (2020-50) under low and high emission scenarios.

Research highlights: TTF sensitivity may be a useful variable when selecting tree species for afforestation to buffer expected shifts in P g due to climate change.

Additional Keywords: climate change; forest ecosystems; precipitation projection; throughfall sensitivity.

Authors' contributions: Pedram Attarod: Design and conception; Qiuhong Tang: Financial support; John T. Van Stan II: Language editor and scientifc comments, and Xingcai Liu: Data preparation.

Citation: Attarod, P., Tang, Q., Van Stan II, J-T., Liu, X. (2018). National assessment of throughfall sensitivity to changes in storm magnitude for the forests of Iran. Forest Systems, Volume 27, Issue 3, e019.

Received: 27 Aug 2018. Accepted: 03 Dec 2018.

Copyright © 2018 INIA. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 International (CC-by 4.0) License.

Funding: This research was supported by the University of Tehran (UT), Iran and by the International Partnership Program of Chinese Academy of Sciences (Grant No. 131A11KYSB20170113) and the National Natural Science Foundation of China (Grant No. 41790424).

Competing interests: The authors have declared that no competing interests exist.

Correspondence should be addressed to Pedram Attarod:





Material and methods






A significant portion of rainfall is intercepted on forest canopy surfaces, where it is either returned to the atmosphere as interception or redistributed to the ground as throughfall (TF) and stemflow. This rainfall partitioning by forest cover significantly alters hydrologic cycling along the soil-forest-atmosphere continuum (Miralles et al., 2010) and can impact feedbacks between the hydrologic cycle and global climate (Davies-Barnard et al., 2014). Stemflow is the proportion of rainfall that drains to the ground along the stem, usually accounting for <2% of annual rainfall in most forests (Van Stan & Gordon, 2018). The remaining majority (60-90%) of rainfall passes through canopy gaps or drips from the vegetation, reaching the surface as TF (Levia et al., 2011) where it can influence soil moisture (Raat et al., 2002), physicochemistry (Rosier et al., 2015), fine root distribution (Ford & Deans, 1978), and microbial processes (Moore et al., 2016). As these ecohydrological interactions play key roles in ecosystem functioning, understanding TF’s response to natural and anthropogenic variability is critical to forest management.

Studies across forest ecosystems agree that rainfall amount (Pg) is the principal variable driving stand-scale TF amount (Levia & Frost 2006; Levia et al., 2011)—often explaining >90% of inter-storm TF variability—and that TF-Pg relationships are shaped by canopy structures, like leaf area index, crown depth, etc. (Staelens et al., 2008; Toba & Ohta 2005). However, the authors are unaware of previous work quantifying TF sensitivity to changes in Pg—a commonly anticipated climate response to hydrologic intensification (Huntington, 2006). This is surprising since hydrologic intensification has been linked to increased extreme Pg in many regions (c.f., Tollefson, 2016). Moreover, afforestation, reforestation, urban forestry and other “greening” initiatives have grown in popularity and, therefore, increased the forest cover in most developed regions (i.e., McGovern & Pasher, 2016). In Iran, for example, the restoration of semi-arid and arid ecosystems through planting of low-demand and drought tolerant species has become a critical element of national ecosystem management plans (Attarod et al., 2015b).

Iran has invested in vast tree-plantings throughout its major cities for urban greening and air pollution mitigation, including the Chitgar and Lavizan Forest Parks (Sadeghi et al., 2016) and the ongoing "Jam Afforestation Project" which is tasked with large-scale afforestation and reforestation in the Zagros region (FRWO, 2012). Restoration of the natural Caspian deciduous forests (that extend from the Alborz Mountains to the southern coast of the Caspian Sea) has also resulted in significant reforestation projects since the 1960s (Abbasian et al., 2015). Concerns have risen over the impact of these greening initiatives on the hydrological cycle (Sun et al., 2006; Wang et al., 2011), particularly for the 90% of Iran classified as arid or semi-arid (Ul Hassan et al., 2007). The first process in the rainfall-to-runoff pathway is the partitioning of Pg by forest canopies (Savenije, 2004) and, accordingly, an improved understanding of TF sensitivity to climate change and Pg is essential for addressing these concerns and quantifying the impacts of Iran’s (and other nations’) large-scale afforestation and reforestation efforts.

A sensitivity analysis is a technique used to determine how different values of an independent variable impact a particular dependent variable under a given set of assumptions. A method for estimating the “sensitivity coefficient” of a dependent variable (in this case, TF) on the relative changes of an independent variable (in this case, Pg) exists (McCuen, 1974) and has been rigorously applied to evapotranspiration and its principal meteorological drivers (Hupet & Vanclooster, 2001). However, the authors are unaware of its application to assess sensitivity of TF to its principal meteorological driver, Pg, for multiple species of contrasting canopy structure. Thus, the objectives of this study are to (1) collect rainfall and TF across common forest species of differing canopy structures and climates in Iran, then (2) quantify and compare their TF sensitivity coefficients. Accomplishing these aims will provide novel information to complement existing data used by forest managers during species selection for restoration and afforestation activities.

Material and methodsTop

Sites description

Data were collected in 9 forest stands of several common native and introduced tree species (Table 1) situated in all common Iranian climate types (Table 2), but located primarily in northern Iran (Fig. 1). The selected species represent a diversity of forest canopy architectures, ranging from the smooth-barked, broad-leaved canopy of Fagus orientalis (FO) to the rough-barked, needle-leaved canopy of Pinus eldarica (PE). Leaf phenology also differs among the selected species as, for example FO is deciduous and PE is evergreen. The PE and Cupressus arizonica (CA) throughfall sites are in southern Alborz Mountain Range near the city of Tehran, while the remaining sites—FO, Quercus castaneifolia (QC1-QC3), Acer velutinum (AV), Pinus brutia (PB), Cupressus sempervirens var. horizontalis (CS)—are in northern Alborz Mountain Range and Southern coasts of the Caspian Sea (Fig. 1). Stand structural (and TF) measurements for each forest stand were performed in 0.5 ha plots. Stand density ranged from 112 trees ha-1 in the FO forest to 1,600 trees ha-1 in the CS plantation (Table 1). Diameter at breast height (1.3 m, dbh) varied greatly among the measured forests, with the smallest values in CS (12 cm) and the largest values in QC3 (65 cm) (Table 1). Mean canopy coverage also exhibited a wide range across sites, from 45% for CS and 95% for FO (Table 1).

Table 1. Forest sites where throughfall (TF) data were collected, including their climate, geographical coordinates, elevation, and stand characteristics. “G” and “D” refer to growing and dormant seasons, respectively, and dbh refers to diameter at breast height.

Table 2. Details of weather stations included in this study, their climate types according to the De Martonne aridity index (IDM: Baltas, 2007), and the duration of their measurement record.

Figure 1. Locations of the measurement sites (triangles) and synoptic meteorological stations (numbers) in the northern provinces of Iran. See Tables 1 and 2 for species abbreviations. Meteorological stations are numbered as: (1) Mehr-Abad; (2) Nou-Shahr; (3) Sari; and (4) Gorgan.

Meteorological data were obtained from the nearest synoptic meteorological stations recording reliable long-term meteorological data (Fig. 1). The range in meteorological data records is from 1951 to 2015 (Table 2). There exists relatively long distances between the Mehr-Abad, Sari, and Gorgan meteorological stations (Fig. 1), but there is no significant topography between the measurement sites and the meteorological stations, and no closer meteorological stations exist in the region. Due to lack of meteorological station inside the Caspian forest, we used meteorological data recorded by Nou-Shahr meteorological station regardless of elevation difference. The ranges of annual precipitation (P) and temperature (T) in the weather station sites are 16.4-17.4 ºC and 230-1291 mm, respectively (Table 2). The ‘‘De Martonne’’ climate classification, i.e., De Martonne aridity index (IDM), as described by Baltas (2007), ranged roughly from 49 to 8.5, so that the study sites were grouped into very humid (FO, QC1), semi-humid (QC2, AV, PB), Mediterranean (QC3, CS), and arid (PE, CA) climates (Table 1).

Field measurements

Field measurements were performed from July-2008 to March-2014 during growing and dormant seasons (Table 1). A discrete rain event was defined as a period with >0.1 mm of rainfall. The minimum inter-event dry period between discrete storms was 4-10 h, depending on the site. The effect of pre-storm canopy wetness was assumed negligible as the canopy is assumed to be dry after the minimum inter-event time, which is a common assumption among rainfall partitioning studies (Carlyle-Moses & Gash, 2011). Snowfall was ignored. Pg at each forest site was measured by 3-10 funnel-type plastic collectors with 10 cm funnel diameter and 20-30 cm heights, placed in the nearest open area away from the forest stands. The average of water amount measured in all rain-gauges at a site was used to estimate Pg for each site. Quantities of water in the collectors were measured manually using a graduated cylinder. Storms that reached only the weather stations but not the forest sites, or vice versa, were ignored. Pg volumes were measured at the same time as TF volumes at each site, either immediately after a storm or at sunrise following a night storm.

TF was measured using 20-50 rain collectors of the same design as those used to quantify Pg. TF collectors were randomly distributed beneath the canopy and fabric covered the neck of the collectors to avoid litter, needles, and debris from entering (Abbasian et al., 2015). TF data based on multiple field campaigns that used a different number of collectors is a common, often necessary, issue in national-to-international scale studies (i.e., Wallace et al., 2013; Návar, 2017); however, it is important to note that this methodological variability may introduce uncertainty to the stand-scale TF estimates (Vose et al., 2016). Variability about mean Pg and TF will be expressed in standard error (SE) throughout.

Throughfall sensitivity coefficients

A practical method of presenting a sensitivity analysis is to plot relative changes of a dependent variable (in this case, TF) against relative change of an independent variable (i.e., Pg) as a curve (e.g., Singh & Xu, 1997; Goyal, 2004, Attarod et al., 2015a). This is different from a standard regression of TF versus Pg (Fig. 2a) as the mean sensitivity coefficient is calculated as the slope of a correlation between the percent changes in Pg against percent changes in TF (Fig. 2b). The sensitivity coefficient represents the fraction of change in Pg transmitted to the change of TF, i.e. a sensitivity value of 0.1 would suggest that a 10% increase in Pg may be predictable to increase TF by 1%. Negative coefficients would indicate that a decrease in TF would result from an increase of Pg, which is not expected due to the universally reported positive relationship between Pg and TF (Friesen et al., 2015). In the present study, the sensitivity coefficient of TF was determined in climate classifications only in response to changes in Pg on an event-basis.

Figure 2. Example regression of throughfall (TF) against storm magnitude (Pg) for data collected at the Fagus orientalis site. Through shifting Pg by any percentage (ΔPg), the regression equation (in panel a) can be used to calculate a corresponding percent change in TF (ΔTF). These shifts are (b) plotted and a regression calculated where the slope is a nondimensional coefcient of “sensitivity.”

Projected Pg changes over the measurement sites

General Circulation Models (GCMs) are numerical models representing physical processes in the atmosphere, ocean, cryosphere, and land surface. These models are the most advanced tools currently available for simulating the response of the global climate system to increasing greenhouse gas concentrations. GCM simulations for the fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC) have become available (Taylor et al., 2012; Miao et al., 2014). Comparing to the IPCC AR4, the GCMs in AR5 include a more varied set of model types (i.e., climate/earth system models with more interactive components such as atmospheric chemistry, aerosols, dynamic vegetation, ice sheets and carbon cycle) (Liu et al., 2013). Several improvements in the physics, numerical algorithms and configurations are implemented in the IPCC AR5 models with a new set of scenarios called Representative Concentration Pathways (RCPs) for energy and industry CO2 emissions (Moss et al., 2010). The RCPs span a large range of stabilization, mitigation and non-mitigation pathways. Phase 5 of the Coupled Model Intercomparison Project (CMIP5) is a standard experimental protocol for studying the output of GCMs which provides a community-based infrastructure in support of climate model diagnosis, validation, intercomparison, documentation and data access (Jones et al., 2011). Longer time-scale (“centennial”) experiments have been performed at the Met Office Hadley Centre with the HadGEM2-ES Earth System model—one of the CMIP5 climate models (Collins et al., 2011; Jones et al., 2011; Miao et al., 2014). The HadGEM2-ES model used here is one of the state-of-the-art GCMs and involves many typical and advanced representations of land and ocean processes (Jones et al., 2011). The HadGEM2-ES climate data has been widely used for climate studies (Huntingford et al., 2013). The r1i1p1 ensemble of HadGEM2-ES was used in this study. The r1i1p1 ensemble is the most accessible ensemble in the CMIP5 archive.

To understand Pg variations under changing climate over the measurement sites, we focused on the precipitation projection for two scenarios: RCP 2.6 and RCP 8.5. RCP 2.6 represents a “low” emissions scenario featured by the radiative forcings of 2.6 Wm-2 and atmospheric CO2 concentration of 421 ppm by 2100. RCP 8.5 represents a “high” emission scenario with the radiation forcing of 8.5 Wm-2 and CO2 concentration of 936 ppm by 2100. The time resolution of projected storm magnitude (i.e., Pg) is daily. Future climate data in grids with 0.5°×0.5° horizontal resolution across measurement sites were obtained from the HadGEM2-ES model projections (Hempel et al., 2013).

Delta change method

Because of the limitations of coarse-resolution GCM climate data (used in these predictions), the delta change (DC) method is generally used to derive scenarios of future climate (e.g., Fischer et al., 2007; Shahid, 2011; Chung & Nkomozepi, 2012). The method consists of simply scaling the observed climate data using monthly change factors calculated from the differences in climatology predicted by GCMs for the current and future periods. In this way, Pg for future time periods was derived by scaling the historically observed climate data (OBS) by the GCM-computed change. This results in a new Pg time series scaled according to the GCMs, but based on historical observations. In this study, the OBS from 1974 to 2004 (reference period) were scaled to derive the 30-yr future climate scenario for the period 2020-2050. Relative change factors (ΔVar) were then applied to the observed flux variable to calculate future Pg. Following Leng & Tang (2014), the DC method are formulated as Eq., (1):


where VarΔ is the scaled flux variable using the DC method and VarOBS is the observed flux variable in the historic period. The suffixes i and j stand for the day and the month, respectively, and ΔVar is the monthly DC factor, which is calculated as follows:


where Varfuture (j) and Varcurrent (j) are the mean values of the time series for month j for the future and current time periods by GCMs.


Overview of rainfall and throughfall events

There were 290 Pg events recorded from 2008-2014 across all measurement sites and Pg ranged from 0.5-54.7 mm. Mean Pg was 10.8±0.7 mm across all forest sites, but site-specific mean Pg ranged from 17-20 mm for forests in the very humid, semi-humid, and Mediterranean climate types (Table 3). Mean Pg from the arid forest sites was 4.4 mm (Table 3). At arid sites, 67% of collected storms were < mean Pg; however, for other climate types, this fraction was generally around 50% (Table 3). “Large” storms (Pg > 15 mm) were frequent at the very humid forest sites (85%), yet only 5% of storms exceeded 15 mm at arid sites (Table 3). “Small” storms (Pg < 5 mm) were the norm for arid forests in this study (70%: Table 3). The respective percentages of Pg < 5 mm or Pg > 15 mm were similar for the semi-humid and Mediterranean forests (Table 3).

Table 3. Characteristics of recorded rainfall events (Pg) in the measurement sites with respect to the different climate types. n refers to the number of recorded events.

Strong positive linear relationships between event mean TF and Pg were observed across all sites [TF=0.66 (Pg) - 0.16; R2 = 0.91] and at each individual site (Table 4). Slopes of the regression lines between TF and Pg (often assumed to be linked to varying canopy structures) were wide-ranging: 0.39 for CS to 0.81 for FO (Table 4). Relative TF (TF:Pg) varied roughly from 40% (Mediterranean CS) to 75% (very humid QC1) (Table 4). In arid and Mediterranean climates, minimum TF:Pg were measured at zero, however, in other climates the minimum was approximately 47% (Table 4). TF:Pg was, surprisingly, nearly the same when averaged for all the needle-leaved forests (55.2%) and broad-leaved forests (50%) across the climate zones (Table 4). Marked differences in TF:Pg on average, however, were observed for plantations versus natural forests, where plantations were found to produce 53.4% of TF compared to 71.9% for natural forests. For all sites, QC1 produced the highest TF:Pg in very humid conditions (QC1: 74.5%) and the lowest TF:Pg under Mediterranean conditions (CS: 39.3%) (Table 4). Generally higher standard errors were observed for TF:Pg in the Mediterranean (3.25%) and arid sites (3.35%) compared to forests in other climates (1.34%) (Table 4).

Table 4. Relationships between throughfall (TF) and rain event magnitude (Pg), percent of event based average relative throughfall (TF: Pg), and related statistics. See Table 1 for the tree species represented by the location codes. n refers to the number of recorded events.

Historical annual storm characteristics

Historical records at the meteorological stations show that mean annual precipitation decreased from very humid (1291 mm y-1) to arid climate (230 mm y-1) (Table 5). Compared with other climates, the arid climate had relatively constant annual precipitation, varying only by 9 mm y-1 between 1951-2015 (Table 5). The ratio of maximum Pg to mean annual precipitation was respectively 11%, 16%, 18%, and 22% for the semi-humid, very humid, Mediterranean, and arid climates (Table 5). Although mean Pg had a descending trend from very humid (10.5 mm) to arid (3.9 mm), the proportion of mean Pg to maximum Pg in different climates ranged from 8.5% in the semi humid climate to 5.0% in the very humid climate (Table 5).

Table 5. Historical characteristics of annual precipitation, number of storms, and other related rain event magnitude (Pg) characteristics recorded in the meteorological stations nearest to the measurement sites.

Throughfall sensitivity to changes in storm magnitude

TF exhibited varying degrees of sensitivity to Pg, showing large fluctuations (i.e., doubling in sensitivity) for species planted in different climates (Fig. 3). TF from the arid PE forest was most sensitive to fluctuations in Pg (= 2.35) and TF from the Mediterranean QC3 forest was least sensitive (= 0.96) (Fig. 3). Needle-leaved plantations (i.e., PE, CA) generally had higher sensitivity (> 1.5) to changing Pg compared to broadleaved forests (~ 1.0) (Fig. 3). For broadleaved forests, QC1 was less sensitive than FO in a very humid climate (Fig. 3). QC1, QC2, and QC3 had similar sensitivity coefficients regardless of climate: very humid (=1.09) to semi-humid (=1.07) and Mediterranean climates (= 0.96) (Fig. 3). TF sensitivity was tested for larger and smaller storm sizes than the mean Pg in each climate (Table 6). Excluding AV, PB, and QC1, TF was found to be more sensitive to small storms (Table 6). Plantations in arid or Mediterranean climates had large differences in sensitivity coefficient between small and large events, e.g. 2.95 vs. 1.14 for PE and 1.78 vs. 0.74 for CS plantation (Table 6). Interestingly, the FO forest had nearly identical sensitivity values for storms smaller (1.06) and larger (1.05) than mean Pg (Table 6). Sensitivity of the very humid QC1 natural forest was greater for large events (1.07 against 1.24) in comparison with the semi-humid QC2 and Mediterranean QC3 forests (Table 6).

Figure 3. Mean throughfall (TF) sensitivity coefcients for the measurement sites. See Table 1 for the tree species represented by the location codes.

Table 6. Throughfall (TF) sensitivity coefcients classifed for storm magnitudes (Pg) smaller and larger than mean Pg recorded in the measurement sites. See Table 1 for the tree species represented by the location codes.

Projected changes in annual precipitation and storm magnitude

During the 2020-2050 period, the low emissions (RCP 2.6) scenario predicted increased annual precipitation on average in all climate types, excluding the semi-humid climate, yet the high emissions (RCP 8.5) scenario predicted decreased annual precipitation on average in all climates excluding the arid climate (Table 7). Projected changes to mean Pg from the low emission scenario was +16.0%, +3.2%, 0%, and -1.8% for Mediterranean, very humid, arid, and semi-humid forests, respectively (Tables 5 and 7). The high emissions scenario predicts that mean Pg will change by -5% in very humid forests and -10% in arid forests, but mean Pg will not change for forests in the Mediterranean and semi-humid climates (Tables 5 and 7).

Table 7. Projections of annual precipitation, and mean storm magnitude (Pg) during the period of 2020-2050 predicted by the GCM under RCP 2.6 (low emission), and RCP 8.5 (high emission) scenarios.


Storm magnitude and throughfall

We collected a wide range of storms across sites hosting common forests in all of Iran’s climate types (Tables 1-2; Fig. 1). Stands for which comparable TF data exist (e.g., FO and QC1) are comparable to those reported elsewhere, for example: FO versus Fagus sylvatica (e.g., Staelens et al., 2008) and QC versus Quercus serrata and Quercus acutissima (e.g., Toba & Ohta, 2005). Relationships between Pg and TF for individual forest often indicate differences in the canopy-meteorological interactions that control TF production by plants (Zhang et at., 2016). In this study, as in others, Pg accounted for most (nearly all) of TF variability regardless the species, yet the slope and intercept of the Pg-TF correlation varied across the forests of differing species-climate combination (Table 4). Relative TF:Pg in our study, 39.3-74.5%, appears to be strongly tied to forest structure (e.g., density, seasonal change, vegetation area index, gap fraction, and canopy storage capacity), and climate conditions (Pg and, perhaps storm intensity and wind conditions) as shown by many others (Crockford & Richardson, 2000; 2011; Staelens et al., 2008; Muzylo et al., 2012; Motahari et al., 2013). Results indicated that, when the same species experiences different climate conditions, they can differentially partition Pg into TF—something particularly evident QC in three different climates (Table 4).

Throughfall sensitivity to storm amount

Large fluctuations in TF sensitivity were observed between different species, climate types, and storm sizes (greater or less than mean annual Pg) common to Iran (Table 6). For forest managers, it is apparent that different species selected for planting in different climates will exhibit differing TF sensitivities to shifting Pg (Fig. 3). Practically, managers can use these sensitivity coefficients (Fig. 3 or Table 6) and projected changes to Pg (Tables 5 versus 7) and estimate the potential shift in TF supply to the forest surface. For example, PE’s sensitivity coefficient (2.35) indicates that a 10% decrease in Pg could approximately reduce TF by 23%. Although we identified little change in the total yearly precipitation in the arid climate for the most recent decade (where these PE plantations are generally situated), an approximate estimation of 8% decrease in the storm size observed in the recent decade can induce an 18% decrease in TF in these plantations. Greater sensitivity coefficients for PE (and CA) plantations may be a result of (1) these species having larger storage capacities than other tree species in this study (Sadeghi et al., 2015), (2) their arid climate allowing their canopies to dry more efficiently between storms, and (3) most-to-nearly all the storms being small (Table 6). The higher sensitivity of these species to Pg has implications for forest managers dealing with climate change, since both are the most widely-used species for afforestation in the arid and semiarid regions of Asia (e.g., Iran, Lebanon, Syria, Pakistan, Iraq, and Afghanistan) and are frequently selected due to their greater tolerance to drought, high/low temperature extremes, and being faster-growing than native broad- or needle-leaved tree species (Jazirei, 2009). Under a changing climate, water resource management in arid regions is therefore complicated by afforestation initiatives and forest managers may be able to use TF sensitivity as a variable in their decision to choose a species with the least sensitivity to expected shifts in Pg.

The remaining species located in the Caspian forests of northern Iran with very humid, semi-humid and Mediterranean climates exhibited roughly the same sensitivity (mean = 1.13; Fig. 3). However, according to our results, the replacement of QC3 natural forest with CS man-made in the Mediterranean areas of the eastern Caspian region will likely increase TF sensitivity by 0.31 (Fig. 3). This increased TF sensitivity with converting Mediterranean QC forests to CS may be a product of greater tree density of young CS (Table 1) or, more theoretically, greater roughness lengths and zero-plane displacement heights (Rutter et al., 1975; Valente et al., 1997). Moreover, CS’s scale-like needle-leaves have different leaf shedding habits compared to QC3’s broadleaves, which will differentially interact with Pg to alter TF and its sensitivity (Pypker et al., 2011).

Natural broad-leaved forests in the Caspian region (FO and QC1) showed roughly the same sensitivity (Fig. 3). The primary function of the Caspian forests, other than wood production, is conservation of soil and water resources (Sagheb Talebi et al., 2014). Thus, it is expected that the low TF sensitivity of FO and QC1 to shifts in Pg will help buffer the region against climate changes. But, restoration of the Caspian deciduous forest of northern Iran by planting species, like CS (which had a greater TF sensitivity coefficient), were extensively performed by Iran’s Forest, Rangeland, and Watershed organization, and may have considerable effects on ecosystem ecohydrology through altered TF supply. In contrast, introduction of PB in the semi humid climate of the central Caspian region for restoration of degraded forests showed roughly the same, even lower, sensitivity compared to native species (AV and QC2) (Fig. 3). Thus, understanding the relationship between hydrologic cycling and the impact of afforestation projects on these variables can be useful for forest management and selection of suitable species for reforestation of degraded forest ecosystems.

Although the consistently-observed strong corre­lation between stand-scale TF and storm amount across forest types and climates confirms that storm amount is the primary factor affecting TF generation after canopy structures are saturated (Levia et al., 2011), the degree to which stand-scale TF responds to storm amount has been linked to vegetation structure (leaf area index, crown depth, etc.) (Staelens et al., 2008; Toba & Ohta 2005). Consequently, it is advisable to incorporate vegetation characteristics with TF sensitivity in response to changes in Pg responses in future investigations.

Combining throughfall sensitivity and climate projections

Changes in Pg to an area are expected to be compounded by the TF sensitivity of each forest type. An example “rough” estimation using the 16% increase in Pg predicted by RCP 2.6 scenario in the Mediterranean climate results in a 20% and 15% increase in TF generation in CS and QC3 forests, respectively. The projected 3.2% increase (by RCP 2.6) and 5% decrease (by RCP 8.5) in mean Pg, however, in the very humid climate of Iran where the Caspian forests are located (Table 7) is not anticipated to significantly alter TF supply to the forest floor due to the tree species’ low sensitivity (Fig. 3). Despite the highest sensitivity of plantations in the arid climate (~2), very slight changes in mean Pg are predicted by the RCP 2.6 scenario which may not influence TF receipt at the surface. However, under the RCP 8.5 scenario, a 10% decrease in mean Pg in the arid climate could decrease TF by 23% and 15% per event beneath PE and CA plantations, respectively. Fewer impacts may be realized in the semi-humid climate where the dominant tree species showed low TF sensitivity and not significant changes in mean event size are predicted by both RCPs. Clearly, these climate change scenarios (and others) can be a reference for setting minimum and maximum configurations of forest cover in water management and planning associated with adaptation. There will be large uncertainty among GCM models, most GCM models may have different performances across regions, and, as a consequence of lacking large scale observations, downscaling from GCM grid data to local areas is difficult. Moreover, the delta change factor strategy used in this study does not adjust climate projections, but assumes that the signal or changes are reasonably projected by climate models, even though the models are biased (Leng & Tang, 2014). A key feature of this approach is that, because the method uses historical precipitation as its basis and GCM data only to change the magnitude of the historical precipitation, it fails to account for changes in P variability predicted by different GCM models (Leng & Tang, 2014). Still, the projections and downscaling of changes in Pg under the most conservative (low emission) and the highest greenhouse gas emissions pathway (high emission) scenarios (Table 7) underscores the necessity of work to understand TF sensitivity of common forest types in a region.


This national assessment of throughfall beneath common forest types in Iran showed large fluctuations in throughfall sensitivity for species planted in different climates. Arid needle-leaved plantations exhibited generally higher sensitivity responses to changing storm amount, while very humid, natural broad-leaved forests had low sensitivity responses. Projections of mean storm magnitude under a low emission scenario indicated that, compared to historical data, modest changes would be expected for most climates (excluding arid) on average during 2020-2050. However, under a high emission scenario, larger alterations are expected in all climate types except the arid climate. Results indicate that any projected change in storm size will not simply be translated directly to a change in throughfall. Rather, the increase or decrease in throughfall amount may be better predicted as a product of the projected change in storm magnitude and the forest-specific throughfall sensitivity. These findings have implications for selection of the most appropriate and adapted species for afforestation under climate change.


Abbasian P, Attarod P, Sadeghi SMM, Van Stan II JT, Hodjati SM, 2015. Throughfall nutrients in a degraded indigenous Fagus orientalis forest and a Picea abies plantation in North of Iran. Forest Syst 24(3) e035.

Attarod P, Kheirkhah F, Khalighi Sigaroodi Sh, Sadeghi SMM, 2015a. Sensitivity of reference evapotranspiration to global warming in the Caspian region, North of Iran. J Agr Sci Tech-Iran 17: 869-883.

Attarod P, Sadeghi SMM, Pypker TG, Bagheri H, Bagheri M, Bayramzadeh V, 2015b. Needle-leaved trees impacts on rainfall interception and canopy storage capacity in an arid environment. New Forest 46: 339-355.

Baltas E, 2007. Spatial distribution of climate indices in northern Greece. Meteorol Appl 14: 69–78.

Carlyle-Moses DE, Gash JHC, 2011. Precipitation interception loss by forest canopies. In: Levia DF, Carlyle-Moses D, Tanaka T (eds) Forest hydrology and biogeochemistry: synthesis of past research and future directions. Ecolog studies 216: 407–424.

Chung SO, Nkomozepi T, 2012. Uncertainty of paddy irrigation requirement estimated from climate change projections in the Geumho River basin, Korea. Paddy Water Environ 10: 175–185.

Collins WJ, Bellouin N, Doutriaux-Boucher M, Gedney N, Halloran P, Hinton T, Hughes J, Jones CD, Joshi M, Liddicoat S, et al., 2011. Development and evaluation of an Earth-system model HadGEM2. Geosci Model Dev 4: 997–1062.

Crockford RH, Richardson DP, 2000. Partitioning of rainfall into throughfall, stem-flow and interception: effect of forest type, ground cover and climate. Hydrol Process 14: 2903–2920. 3.0.CO;2-6" target="_blank"><2903::AID-HYP126>3.0.CO;2-6

Davies-Barnard T, Valdes PJ, Jones CD, Singarayer JS, 2014. Sensitivity of a coupled climate model to canopy interception capacity. Clim Dynam 42(7-8): 1715-1732.

Fischer G, Tubiello FN, van Velthuizen H, Wiberg DA, 2007. Climate change impacts on irrigation water requirements: Effects of mitigation, 1990–2080. Technol Forecast Soc 74: 1083–1107.

Ford ED, Deans JD, 1978. The effects of canopy structure on stemflow, throughfall and interception loss in a young sitka spruce plantation. J Appl Ecol 15(3): 905-917.

FRWO (Forest, Range, and Watershed Organization of Iran), 2012. Project of Participatory Afforestation in the Zagros Region.

Friesen J, Lundquist J, Van Stan II JT, 2015. Evolution of forest precipitation water storage measurement methods. Hydrol Process 29: 2504–2520.

Goyal RK, 2004. Sensitivity of evapotranspiration to global warming: a case study of arid zone of Rajasthan (India). Agr Water Manage 69: 1–11.

Huntingford C, Zelazowski P, Galbraith D, Mercado LM, Sitch S, Fisher R, Lomas M, Walker A, Jones C, Booth, B, et al., 2013. Simulated resilience of tropical rainforests to CO2-induced climate change. Nature Geoscience 6: 268–273.

Hempel S, Frieler K, Warszawski L, Schewe J, Piontek F, 2013. A trend-preserving bias correction - the ISI-MIP approach. Earth Syst Dynam 4(2): 219-236.

Huntington TG, 2006. Evidence for intensification of the global water cycle: review and synthesis. J Hydrol 319: 83-95.

Hupet F, Vanclooster M, 2001. Effect of the sampling frequency of meteorological variables on the estimation of the reference evapotranspiration. J Hydrol 243: 192–204.

Jazirei MH, 2009. Dryland Afforestation. University of Tehran Press, Iran. pp 560. ISBN 978-964-03-6122-1.

Jones CD, Hughes JK, Bellouin N, Hardiman SC, Jones GS, Knight J, Liddicoat S, O'Connor1 FM, Andres RJ, Bell C, et al., 2011. The HadGEM2-ES implementation of CMIP5 centennial simulations. Geoscie Model Dev 4(3): 543-570.

Leng G, Tang Q, 2014. Modeling the impacts of future climate change on irrigation over China: sensitivity to adjusted projections. J Hydrometeorol 15(5): 2085-2103.

Levia DF, Frost EE, 2006. Variability of throughfall volume and solute inputs in wooded ecosystems. Prog Phys Geog 30: 605–632.

Levia DF, Keim RF, Carlyle-Moses DE, Frost EE, 2011. Throughfall and stemflow in wooded ecosystems, in Forest Hydrology and Biogeochemistry: Synthesis of Past Research and Future Directions, Ecol Studr Ser 216: 425–443.

Liu J, Song M, Horton RM, Hu Y, 2013. Reducing spread in climate model projections of a September ice-free arctic. P.12571-6 in Proc. of Conf. of National Academy of Sciences of the United States USA 110 (31).

McCuen RH, 1974. A sensitivity and error analysis of procedures used for estimating evaporation. Water Resour Bull 10 (3): 486–498.

McGovern M, Pasher J, 2016. Canadian urban tree canopy cover and carbon sequestration status and change 1990–2012. Urban For Urban Gree 20: 227-232.

Miao C, Duan Q, Sun Q, Huang Y, Kong D, Yang T, Ye A, Di Zh, Gong W, 2014. Assessment of CMIP5 climate models and projected temperature changes over Northern Eurasia. Environ Res Lett.

Miralles DG, Gash JH, Holmes TRH, de Jeu RAM, Dolman AJ, 2010. Global canopy interception from satellite observations. J Geophys Res 115.

Moore LD, Van Stan II JT, Gay TE, Rosier C, Wu T, 2016. Alteration of soil chitinolytic bacterial and ammonia oxidizing archaeal community diversity by rainwater redistribution in an epiphyte-laden Quercus virginiana canopy. Soil Biol Biochem 100: 33-41.

Moss RH, 2010. The next generation of scenarios for climate change research and assessment. Nat 463: 747–56.

Motahari M, Attarod P, Pypker TG, Etemad V, Shirvany A, 2013. Rainfall interception and canopy storage capacity of a Pinus eldarica plantation in a semi-arid climate zone: an application of the Gash model. J Agr Sci Tech-Iran 15: 981–994.

Muzylo A, Llorens P, Domingo F, 2012. Precipitation partitioning in a deciduous forest plot in leafed and leafless periods. Ecohydrology 5: 759–767.

Návar J, 2017. Fitting rainfall interception models to forest ecosystems of Mexico. J Hydrol 548: 458–470.

Pypker TG, Levia DF, Staelens J, Van Stan II JT, 2011. Canopy structure in relation to hydrological and biogeochemical fluxes. XVII. In: Levia, D.F., Carlyle-Moses, D.E., Tanaka, T. (Eds.), Forest Hydrology and Biogeochemistry: Synthesis of Past Research and Future Directions. Ecological Studies Series, Springer-Verlag, Heidelberg 216: 371–388.

Raat KJ, Draaijers GPJ, Schaap MG, Tietema A, Verstraten, JM, 2002. Spatial variability of throughfall water and chemistry and forest floor water content in a Douglas fir forest stand. Hydrol Earth Syst Sci 6: 363–374.

Rosier CL, Van Stan II JT, Moore LD, Schrom JOS, Wu T, Reichard JS, Kan J, 2015. Forest canopy structural controls over throughfall affect soil microbial community structure in an epiphyte-laden maritime oak stand. Ecohydrology.

Rutter AJ, Morton AJ, Robins PC, 1975. A predictive model of rainfall interception in forests II. Generalization of the model and comparison with observations in some coniferous and hardwood stands. J Appl Ecol 12: 367–380.

Sadeghi SMM, Attarod P, Van Stan II JT, Pypker TG, Dunkerley D, 2015. Efficiency of the reformulated Gash's interception model in semiarid afforestations. Agr Forest Meteorol 201: 76-85.

Sadeghi SMM, Attarod P, Van Stan II JT, Pypker TG, 2016. The importance of considering rainfall partitioning in afforestation initiatives in semiarid climates: A comparison of common planted tree species in Tehran, Iran. Sci Total Environ 568: 845-855.

Sagheb Talebi K, Sajedi T, Pourhashemi M, 2014. Forests of Iran, A treasure from the past, a hope for the future. Plant Vegetation 10: 149.

Shahid S, 2011. Impacts of climate change on irrigation water demand in northwestern Bangladesh. Climatic Change 105: 433–453.

Savenije HHG, 2004. The importance of interception and why we should delete the term evapotranspiration from our vocabulary. Hydrol Process 18(8): 1507-1511.

Singh VP, Xu C-Y, 1997. Sensitivity of mass transfer-based evaporation equations to errors in daily and monthly input data. Hydrol Process 11: 1465–1473. 3.0.CO;2-X" target="_blank"><1465::AID-HYP452>3.0.CO;2-X

Staelens J, Schrijver AD, Verheyen K, Verhoest NEC, 2008. Precipitation partitioning into throughfall, stemflow, and interception within a single beech (Fagus sylvatica L.) canopy: influence of foliation, rain event characteristics, and meteorology. Hydrol Process 2: 33-45.

Sun G, Zhou GY, Zhang ZQ, Wei XH, McNulty SG, Vose JM, 2006. Potential water yield reduction due to forestation across China. J Hydrol 328: 548–558.

Taylor KE, Stouffer RJ, Meehl GA, 2012. An overview of Cmip5 and the experiment design B. B Am Meteorol Soc 93: 485–98.

Toba T, Ohta T, 2005. An observational study of the factors that influence interception loss in boreal and temperate forests. J Hydrol 313: 208–220.

Tollefson J, 2016. Global warming already driving increases in precipitation extremes. Nature.

Ul Hassan M, Qureshi AS, Heydari N, 2007. A proposed framework for irrigation management transfer in Iran: Lessons from Asia and Iran. Colombo, Sri Lanka: International Water Management Institute. 37 pp. (IWMI Working Paper 118)

Valente F, David JS, Gash JHC, 1997. Modelling interception loss for two sparse eucalypt and pine forests in central Portugal using reformulated Rutter and Gash analytical models. J Hydrol 190: 141–162.

Van Stan II JT, Godon DA, 2018. Mini-Review: Stemflow as a Resource Limitation to Near-Stem Soils. Front Plant Sci 9(248): 1-9.

Vose S, Zimmermann B, Zimmermann A, 2016. Detecting spatial structures in throughfall data: The effect of extent, sample size, sampling design, and variogram estimation method. J Hydrol 540: 527-537.

Wallace J, Macfarlane C, McJannet D, Ellis T, Grigg A, Van Djik A, 2013. Evaluation of forest interception estimation in the continental scale Australian water resources assessment-landscape (AWRA-L) model. J Hydrol 499: 210–223.

Wang YH, Yu P, Feger KH, Wei X, Sun G, Bonell M, Xiong, W, Zhang S, Xu L, 2011. Annual runoff and evapotranspiration of forestlands and non-forestlands in selected basins of the Loess Plateau of China. Ecohydrology 4: 277–287.

Zhang Z, Zhao Y, Li X, Huang L, Tan H, 2016. Gross rainfall amount and maximum rainfall intensity in 60-minute influence on interception loss of shrubs: a 10-year observation in the Tengger Desert. Sci Rep-UK 6: 26030.