Forest Disturbance Thresholds and Cumulative Hydrological Impacts
Abstract
Forest disturbance threshold is defined as a critical disturbance level (e.g., % of forest cover change) in forested landscapes above which significant hydrological impacts are detected. Determining disturbance thresholds is critically important for supporting forest management to ensure the sustaining of ecological and hydrological functions. However, there are very few quantitative evaluations of forest disturbance thresholds globally. In this study, we applied a well-tested methodology (the modified double mass curve) to derive the long-term, continuous hydrological response curves and then to quantify forest disturbance thresholds on annual streamflow in 42 forested watersheds in British Columbia, Canada. The results show that forest disturbance thresholds for significant and cumulative hydrological impacts vary from 7% to 52% of cumulative equivalent clear-cut area with an average of 17% or from 8% to 52% of disturbed area with an average of 19%. Climate (inter-annual and intra-annual) and watershed properties exert critical controls on forest disturbance thresholds. Watersheds with greater snowfall proportions (more annual precipitation falling as snow), more desynchronizations (temporal mismatching) of energy demand and water supply at the intra-annual scale, less diverse ecosystems, and larger watershed sizes have lower forest disturbance thresholds. Given the present forest disturbance levels in the central interior of British Columbia, about half (53%) of the forested watersheds have already crossed the average disturbance threshold. These results highlight that watershed planning and management using forest disturbance thresholds must carefully consider local climate and watershed properties. The methodology can be effectively and robustly extended elsewhere around the globe.
Key Points
-
A new approach with the hydrological response curves is used to quantify forest disturbance thresholds
-
Forest disturbance thresholds for cumulative impacts on annual streamflow vary from 7% to 52% of cumulative equivalent clear-cut area with the average of 17%
-
Climate and watershed properties exert critical controls on the thresholds
Plain Language Summary
This first-ever study uses the hydrological response curves to quantify forest disturbance thresholds on annual streamflow in 42 forested watersheds in the interior of British Columbia, Canada. We find that forest disturbance thresholds range from 7% to 52% of cumulative equivalent clear-cut area with the average of 17%. Climate and watershed properties play critical roles in controlling disturbance thresholds. Watersheds characterized by more annual precipitation falling as snow, more mismatching of energy and water at the intra-annual scale, larger watershed size, and less diverse ecosystems exhibit lower disturbance thresholds. Management implications and future research needs for preserving hydrological functions are discussed.
1 Introduction
Forest disturbance can significantly alter hydrological processes and functions from local watersheds to the globe (Creed et al., 2019; Zhang & Wei, 2021). Global forests are experiencing considerable anthropogenic and natural disturbances. Since the 1990s, global forests have experienced large-scale anthropogenic disturbance (e.g., timber harvesting, land use change) with the most activities and forest loss in the Amazon and Africa (FAO, 2020; Keenan et al., 2015). With climate change impacts, more than 4 million wildfires occurred each year from 2004 to 2019 and the most severe wildfires burned large-scale forests in Australia, Brazil, Canada, and the USA from 2018 to 2021 (Mansoor et al., 2022). In addition, there are large-scale outbreaks of insect infestation in North America (Biederman et al., 2015; Hou et al., 2022; Ren et al., 2021). These anthropogenic and natural disturbances show diverse patterns, severities, and persistence over space and time, interactively and collectively affecting forest dynamics and thus hydrological processes (Wei et al., 2022; Zhang et al., 2022). Once forest disturbance cumulatively reaches a critical level or threshold, significant hydrological impacts can be produced and detected, leading to a dramatic shift to a non-equilibrium state (Julian & Gardner, 2014; Wei et al., 2021). Thus, determining forest disturbance thresholds is critically needed to support forest management to ensure ecological and hydrological functions and services under the context of increasing forest disturbance and climate change impacts.
For relatively large landscapes, forest disturbance can have cumulative impacts on hydrology over space and time. Cumulative hydrological impacts refer to changes in hydrological processes caused by the combined effects of human and natural disturbances in the past and present (Clark, 1994). While a series of individual forest disturbances may not significantly affect hydrological processes in a watershed, the cumulative impact of multiple disturbances can be significant (Reid, 1993). For example, a single and small clear-cut in a large, forested watershed is unlikely to yield a measurable effect on hydrology. However, with the increasing area of harvesting or other types of forest disturbance in the same watershed over a relatively longer period, a detectable change would likely occur (MacDonald, 2000). From a practical perspective, when significant and cumulative hydrological impacts reach, such impacts on flow variables can be detected by statistical evaluations. For example, timber harvesting alone did not significantly alter annual runoff in two boreal forested watersheds in Canada, but when both timber harvesting and subsequent mountain pine beetle infestation were included, significant changes in annual runoff occurred (Hou et al., 2022). Furthermore, the cumulative impacts of forest disturbance on hydrology could be more severe and greater than the sum of individual disturbance impacts (Canter & Ross, 2010). Although the effects of individual forest disturbance types on hydrology have been extensively studied, research assessing cumulative hydrological impacts across large watersheds or landscapes is limited. As planning and management often operate across larger geographic landscapes, such research is critically needed.
Forest disturbance threshold is defined as a critical disturbance level where, once reached or exceeded, significant and cumulative hydrological impacts are detected (Wei et al., 2021). Forest disturbance thresholds for significant and cumulative hydrological impacts are not the same as ecological tipping points which are critical points or zones linking two alternative states of an ecosystem (Andersen et al., 2009; Holling, 1973; Kelly et al., 2015). Once reaching the ecological tipping point, the shift between two ecological states or regimes is normally irreversible and the resultant impacts can be catastrophic (Scheffer et al., 2001). In contrast, with the passing of the forest disturbance threshold, statistically significant hydrological impacts can be caused. However, these hydrological effects can be gradually diminished with recovery from forest regeneration (Coble et al., 2020; Feng et al., 2016). A forest disturbance threshold provides a practical guide for the designing of management strategies to avoid significant and cumulative changes in hydrological processes and functions.
Stednick (1996) recommended a forest disturbance threshold, that is, 20% of forest cover loss from paired watershed experiments in the USA. However, paired watershed experiments often involve abrupt and single event-based forest treatments, and they are not designed for determining forest disturbance thresholds. Besides, forest disturbance thresholds are likely dependent upon hydrological variables of interest (e.g., annual streamflow, high flows, and low flows). To scientifically determine a forest disturbance threshold in an individual watershed, Wei et al. (2021) recommended that a long-term and continuous response curve of hydrological effects caused by forest change (e.g., cumulative forest disturbance) for a selected hydrological variable is required. From the hydrological response curve, hydrological impacts of climate variability over time are removed, and the corresponding disturbance level with the significant breakpoint identified in the curve is the forest disturbance threshold.
In addition, forest disturbance thresholds vary among watersheds due to the heterogeneity in climate and watershed properties. For example, Stednick (1996) indicated that the thresholds required for measurable changes in annual water yield ranged from 15% (Rock Mountain Inland) to 50% (Central Plain) of forest harvested area in the USA. In general, more hydrologically resistant watersheds tend to have higher forest disturbance thresholds as their more diverse tree species and flatter topography can buffer hydrological responses to forest disturbance (Creed et al., 2014; Peña-Arancibia, Bruijnzeel, Mulligan, & van Dijk, 2019; Zhang et al., 2017; Zhou et al., 2015). Thus, understanding the roles of environmental factors in regulating forest disturbance thresholds can help us identify vulnerable areas, avoid crossing thresholds, and predict forest disturbance thresholds.
In snow-dominated watersheds, snowpack accumulation and ablation processes play a critical role in water supply, vegetation dynamics, and many other processes (Berghuijs et al., 2014; Jenicek & Ledvinka, 2020; Rixen et al., 2022; Wieder et al., 2022). Snowpack accumulation and ablation partially determine streamflow magnitude, seasonality, and hydrological regime (e.g., snow-dominated) of a watershed. For example, Li et al. (2017) showed that 53% of the total runoff originated from 37% of snowfall in the western United States. Forest disturbance in snow-dominated watersheds could affect snowpack distribution and accumulation, and consequently, hydrological processes. Because of the hydrological implications of snowpack in snow-dominated watersheds, it is logical to assume that snow conditions could potentially influence cumulative hydrological responses to forest disturbance, and consequently forest disturbance thresholds.
In spite of the growing interest in forest disturbance thresholds, there are very few quantitative studies specifically conducted on this topic. To fill this critical gap, we applied a well-tested method (the modified double mass curve, MDMC) (Giles-Hansen et al., 2019; Li et al., 2018; Liu et al., 2015; Wei & Zhang, 2010; Zhang & Wei, 2012) to derive hydrological response curves for quantitatively determining forest disturbance thresholds for significant and cumulative hydrological impacts on annual streamflow in 42 watersheds in forested landscapes in British Columbia (BC), Canada. The contributing factors to forest disturbance thresholds were also evaluated. The MDMC was initially designed to assess cumulative hydrological impacts on annual streamflow caused by forest disturbance in any individual forested watersheds (Wei & Zhang, 2010), and it can be scientifically used to determine forest disturbance thresholds (Wei et al., 2021) as it effectively produces a long-term and continuous response curve of hydrological effects.
In this study, we aimed to address these three scientific questions: (a) What are forest disturbance thresholds for significant and cumulative hydrological impacts on annual streamflow in BC? (b) How do forest disturbance thresholds vary among climate conditions (inter-annual and intra-annual) in the snow-dominated environment? And (c) how do watershed properties affect forest disturbance thresholds?
2 Materials and Methods
2.1 Study Area
British Columbia (BC) with abundant forest resources (>60% of the provincial landscape) is situated in the west of Canada (Figure 1). There are 424 active hydrometric stations operated in BC by Water Survey of Canada (WSC). Study watersheds were selected based on the following criteria: (a) long-term streamflow records are longer than 40 years from active hydrometric stations; (b) watershed size is less than 10,000 km2; and (c) forest disturbance level is greater than 10%. This resulted in a selection of 42 watersheds (Figure 1 and Table S1 in Supporting Information S1). The 42 selected forested watersheds vary in size from 3.7 to 9,930 km2 and have cumulative forest disturbance due to timber harvesting, wildfire, and mountain pine beetle (MPB) infestation. The region has a cold climate with considerable snowfall (48% of precipitation falls as snow). The annual air temperature for the selected watersheds is 2.2°C and the annual precipitation is 880 mm/yr. Coniferous forests are the dominant forest type. Watersheds in the central interior have relatively flat topography while those in the southeastern and southwestern interior have mountainous terrains (Figure 1). The elevation ranges from 3,500 above sea level (asl. m) down to 350 asl. m in the valley bottoms. The average watershed slope is 9° (Table S1 in Supporting Information S1).
The spatial distribution of the study watersheds.
2.2 Data
In this study, forest disturbance, climate, hydrological, and topographic data were used. Forest disturbance history was derived and calculated from two provincial-level databases, the Vegetation Resource Inventory (VRI) and the Consolidated Cutblocks databases (Version 2019), published by the British Columbia Ministry of Forests. Natural disturbance (i.e., wildfire and insect infestation) information, including tree characteristics, disturbance type, year, and severity, was recorded in the VRI database until 2018, while the detailed logging information (e.g., period, size, and location) until 2018 was summarized in the Consolidated Cutblocks database. We generated the long-term history of forest disturbance for each study watershed from the combined database. Besides, the VRI database also provides Biogeoclimatic Ecosystem Classifications (BEC) for the entire province. BEC combines biological (vegetation), geological (soil), and climatic concepts of a site, indicating the vegetation potential and ecosystem diversity on a site (Meidinger & Pojar, 1991). A detailed description of BEC can be found in Section S2.1 in Supporting Information S1.
Long-term daily discharges and watershed boundaries are openly available from the WSC HYDAT database, Environment and Climate Change Canada. Runoff depths (unit: mm) were calculated from discharges (unit: m3/s) and drainage areas, and annual streamflow (Q) was summarized from daily flows for a given year.
Historical climate variables, including mean air temperature (Tave), precipitation (PPT), and snowfall amount at the monthly scale from 1950 to 2020 were derived from ERA5-Land reanalysis data set, which was released by the European Centre for Medium-Range Weather Forecasts (Muñoz-Sabater, 2019). Different from satellite-based climate products, ERA5-Land derives from data assimilation algorithms combining forecasts with available observations to estimate different climatic variables (Muñoz-Sabater, 2019). Previous studies have indicated that the gridded ERA5-Land reanalysis data set has better performance than satellite-based precipitation products in comparison with in situ data and representation of precipitation and evapotranspiration (ET) seasonality (Hernandez et al., 2022; Jiang et al., 2021; Tarek et al., 2020; Teo et al., 2022). Watershed-scale data were summarized as the average value of all grids based on watershed boundaries.
We also derived the Köppen-Geiger climate classification map for the present day (1980–2016) from the GloH2O website and summarized climate classifications for the entire study area (Section S2.2 in Supporting Information S1). The globe is classified into 30 sub-climate types and five major systems (tropical, arid, temperate, continental, and polar) according to the threshold and seasonality of monthly air temperature and precipitation in the Köppen-Geiger climate classification (Beck et al., 2018). British Columbia generally has a cold, continental climate. Köppen-Geiger Dfc (cold with no dry seasons) and Dsc (cold with dry summer) climate classes are dominated in the study watersheds (Section S2.2 in Supporting Information S1).
This digital elevation model (DEM) for BC province was extracted from Canadian Digital Elevation Data (CDED) published by Natural Resources Canada with a spatial resolution of ∼20 m. DEMs were incorporated with the VRI-Cutblocks database to estimate disturbance levels and used to calculate watersheds' topography (e.g., slope and elevation).
2.3 The Framework for Determining Forest Disturbance Thresholds
In this study, we tested and applied a framework to determine forest disturbance thresholds for significant and cumulative hydrological impacts on annual streamflow in forested watersheds in BC (Figure 2a; Wei et al., 2021). From the framework, forest disturbance levels were first quantified (Step 1), and then long-term and continuous hydrological response curves were derived with the effects of climate variability being removed so that the hydrological impacts of forest disturbance can be quantified (Step 2). Forest disturbance thresholds were determined from the constructed hydrological response curves (Step 3). In the hydrological response curve, the corresponding forest disturbance level of a breakpoint that produces significant hydrological change is viewed as the disturbance threshold. This framework can be used in any individual watershed to determine the disturbance threshold caused by forest disturbance. We applied the framework combining forest disturbance history (represented by disturbed area and cumulative equivalent clear-cut area, CECA; Figure 2b) and the modified double mass curve (MDMC; Figure 2c) to determine forest disturbance thresholds for cumulative hydrological impacts on annual streamflow in 42 forested watersheds in BC.
(a) The framework to determine forest disturbance thresholds for cumulative and significant hydrological impacts. (b) An example of forest disturbance history represented by cumulative equivalent clear-cut area, CECA. MPB denotes mountain pine beetle infestation. (c) An example of the hydrological response curve of annual streamflow produced by the modified double mass curve, MDMC.
First, we used disturbed area (%) and cumulative equivalent clear-cut area (CECA) as proxies to quantify forest disturbance levels. The disturbed area is a static disturbance proxy, which is calculated as the percentage of the total disturbed area (km2) over the watershed area (km2) without considering vegetation dynamics caused by different disturbance types and post-disturbance vegetation recovery. In contrast, CECA is a dynamic indicator that accounts for disturbance severity caused by different disturbance types, post-disturbance recoveries due to forest regeneration, and the effects of disturbance above the given hypsometric line on hydrological processes (Giles-Hansen & Wei, 2021; Hou et al., 2022; Lewis & Huggard, 2010; Winkler & Boon, 2017).
For each disturbance type (e.g., logging, wildfire, and insect infestation), annual equivalent clear-cut area (ECA) is calculated as the disturbed proportion (disturbed area over the watershed area) multiply a coefficient (ECA = disturbed area × coefficient) that captures forest dynamics (Winkler & Boon, 2017). Since stand-replacing disturbance (e.g., logging and wildfire) generally removes the entire forest stand while non-stand-replacing disturbance (e.g., insect infestation) produces partial mortality and gradual changes in the overstory, coefficients are assigned based on the type of disturbance to capture differences in forest dynamics (Hou et al., 2022). For example, to estimate ECA for logging and wildfire (stand-replacing disturbance), 100% coefficients are initially assigned when the disturbance occurs, and they gradually decrease with forest regeneration in the subsequent years. For trees that are attacked by non-stand-replacing disturbance (e.g., insect infestation), coefficients increase in the first few years due to the gradual changes in stands and decrease with regeneration. Coefficients also vary depending on site index, BEC zone, and tree mortality (Lewis & Huggard, 2010). Once the annual ECA for each disturbance type is estimated, the values are summed to be the CECA of the watershed (Figure 2b). CECA is a good indicator to present gradual and cumulative forest disturbance in BC (Vore et al., 2020). The calculations of CECA were conducted in R based on the combined VRI-Cutblocks-DEM database. More details of CECA calculations can be found in Section S4 and Figure S2 in Supporting Information S1.
Second, we applied the MDMC to construct the long-term and continuous hydrological response curve with the effect of climate variability on annual streamflow being removed. The MDMC applies the effective precipitation (i.e., the difference between annual precipitation and predicted evapotranspiration from the Budyko framework) to remove climate variability impacts on annual streamflow (Wei & Zhang, 2010; Zhang & Wei, 2012). The MDMC assumes that accumulated climate variability has a near-linear effect on accumulated annual streamflow if forest disturbance plays a limited role. When forest disturbance gradually increases but is still below its disturbance threshold, the MDMC follows an initial pattern between accumulated annual effective precipitation and accumulated annual streamflow (Figure 2c). This indicates that the initial equilibrium state between forests, climate, and hydrology is not disturbed. In such cases, no significant hydrological impacts of cumulative forest disturbance would be examined. However, when the disturbance level continues increasing and exceeds the threshold, a breakpoint occurs, the initial relationship is disrupted, and significant and cumulative hydrological impact takes place (Figure 2c). The disturbance level that corresponds to the breakpoint is defined as the disturbance threshold. The MDMC is developed based on the water balance equation at the annual scale (i.e., P = ET + Q), thus it cannot be used to construct the hydrological response curve to determine forest disturbance threshold for sub-annual low flows or high flows.
Finally, statistical examinations, that is, the Pettitt's and distribution-free cumulative sum (CUSUM) tests, were employed to confirm the statistical significance of breakpoints in the hydrological response curves due to forest disturbance with the slope series of the curve as examined targets. The non-parametric Pettitt's test has been widely used to determine the most likely occurrence of a breakpoint in hydrological and meteorological studies (Collar et al., 2022; Pettitt, 1979). The distribution-free CUSUM (cumulative sum) test determines the step change year in a time series (McGilchrist & Woodyer, 1975). For the slope series, the potential breakpoint occurs at the maximum value of the cumulative sum and the significance is confirmed if the maximum value of the cumulative sum is greater than the critical value at a given confidence interval (Grayson et al., 1996). These two tests have no specific requirements for the distribution of data series. The R packages “trend” and “trendchange” were used to conduct Pettitt's and distribution-free CUSUM tests, respectively (R Core Team, 2020). For Pettitt's test, the significance of the breakpoint was α = 0.05 while that was identified at the 95% confidence interval for the distribution-free CUSUM test.
2.4 Estimating the Roles of Environmental Factors
Forest disturbance thresholds should vary among environments, which might be affected by climate and watershed properties (e.g., vegetation characteristics and topography). We examined the roles of different environmental variables in the variations of forest disturbance thresholds (Table 1). To represent inter- and intra-annual climate, Köppen-Geiger climate classification, dryness index (DI = PET/PPT), and snow fraction (snowfall/PPT, the proportion of precipitation falling as snow) were used (PET calculation can be found in Section S3.1 in Supporting Information S1). Watersheds were classified into energy-limited (DI < 0.76) and equitant (0.76 < DI < 1.35) environments according to the long-term DI (Zhang et al., 2017). Water availability (precipitation) is sufficient for forest growth while forest growth is constrained by energy (e.g., temperature, solar radiation) in energy-limited environments. This study tested if constrained energy availability at the inter-annual scale could affect disturbance thresholds. In BC, snow processes (e.g., snow melting) play a considerable role in hydrology, such as determining the distribution of water throughout the year. Thus, snow fraction was used to estimate the effect of snow on forest disturbance thresholds. According to the Köppen-Geiger climate classification (Beck et al., 2018), the study watersheds can be classified into Dfc (cold continental with no dry season) and Dsc (cold continental with dry summer) climates. The dry summer condition in the Dsc climate implies that precipitation is limited in summer, but the temperature is high, suggesting the desynchronized energy demand and water supply at the intra-annual scale. It is supposed that watersheds with seasonal dry conditions tend to have lower disturbance thresholds, which was also tested in this study.
| Category | Factor | Description |
|---|---|---|
| Climate | Dryness index | DI = PET/PPT. PET and PPT denote potential evapotranspiration and precipitation, respectively. It was used to represent annual water and energy availability in watersheds. |
| Snow fraction | Snowfall/PPT. It indicates the proportion of precipitation falling as snow and was used to estimate the effect of snow on forest disturbance thresholds. | |
| Köppen-Geiger classification | Dfc climate denotes the cold continental climate with no dry season and Dsc represents the cold continental climate with dry summer. It suggests the desynchronized or synchronized energy demand and water supply at the intra-annual scale. | |
| Ecosystem diversity | Number of dominant BEC zones | More BEC zones indicate more diverse ecosystem compositions. |
| Topography | Watershed size | It is potentially related to the buffering capacity. |
| Water retention index | The water retention index (IR) is a function of the watershed's average slope and range in elevation (i.e., basin relief), and represents topographic retention capacity (Hou et al., 2023). |
Then, we focused on the role of ecosystem diversity. Watersheds with more diverse ecosystems were supposed to be hydrologically resistant and consequently, had higher disturbance thresholds. This study applied the number of dominant BEC zones for the study watersheds to indicate ecosystem diversity. More BEC zones indicated more diverse ecosystem compositions. Because some BEC zones only accounted for a small proportion of the watershed area and their roles would be minor, we only counted BEC zones when their proportions were greater than 10% of the watershed area.
Finally, hydrologically resistant environments are hypothesized to have greater buffering capacity, indicating the ability of watersheds to attenuate negative hydrological impacts or higher retention capacity, and thus higher disturbance thresholds. It is commonly believed that larger watersheds have greater hydrological buffering (Zhou et al., 2015), and thus watershed size was selected as a factor for testing. Additionally, water retention index (IR) was applied to represent topographic retention capacity (Hou et al., 2023). The average water retention index is calculated as a function of the watershed's average slope and range in elevation (i.e., basin relief) (Hou et al., 2023). Watersheds with steep slopes and significant elevation differences (i.e., relief) tend to have lower water retention capacities with shorter water residence times and shorter flow paths (Jencso & McGlynn, 2011). Watersheds with a size smaller than 500 km2 were classified as smaller watersheds while those with a size larger than 500 km2 were large watersheds.
The significant correlations between environmental variables and forest disturbance thresholds were tested by univariate and multivariate linear regressions, while the significant differences in forest disturbance thresholds between environmental categories were tested by the non-parametric Mann-Whitney U test (Mann & Whitney, 1947). The Mann-Whitney U test has been widely used in hydrological studies to detect significant differences in variables between two groups or periods because the test is insensitive to the distribution of data samples (Aryal & Zhu, 2020). The statistical analyses were performed in R (R Core Team, 2020).
3 Results
3.1 Forest Disturbance Thresholds
Forest disturbance thresholds for significant and cumulative hydrological impacts on annual streamflow varied from 7% to 52% of CECA with an average of 17% in 42 forested watersheds (Figure 3a and Table 2). They were equivalent to 8%–52% of disturbed area with an average of 19% without considering forest dynamics and hydrological recoveries (Figure 3a and Table 2). Large-scale, stand-replacing disturbances removed a large proportion of forests in three watersheds experiencing treatment-based logging (W15 and W16) and catastrophic wildfire (W9) in a disturbed year, and produced significant hydrological impacts in the subsequent years, thus considerably larger thresholds were estimated (Figure 3b). The quantified cumulative forest disturbance levels represented by CECA (Figure S2 in Supporting Information S1), established hydrological response curves (MDMCs; Figure S3 in Supporting Information S1), and determined the forest disturbance thresholds for the study watersheds (Tables S3 and S4) can be found in Supporting Information S1.
(a) The determined forest disturbance thresholds for significant hydrological impacts on annual streamflow represented by cumulative equivalent clear-cut area, CECA and disturbed area. Dots are forest disturbance thresholds for the study watersheds. (b) A comparison of the determined forest disturbance thresholds represented by CECA with gradual and abrupt (W9, W15, and W16) disturbances.
| Thresholds | CECA (%) | Disturbed area (%) |
|---|---|---|
| Sample size (N) | 42 | 42 |
| Mean | 17 | 19 |
| Median | 15 | 16 |
| Max | 52 | 52 |
| Min | 7 | 8 |
| Standard deviation (SD) | 10.6 | 11.0 |
This study also evaluated the relationship between disturbance thresholds and hydrological sensitivities, indicating the response intensity (%) of annual streamflow per unit of CECA change (%) (Section S3.2, Table S4, and Figure S4 in Supporting Information S1). There was a significant and negative relationship between forest disturbance thresholds and hydrological sensitivities (p < 0.01) (Figure S4 in Supporting Information S1), suggesting more hydrologically sensitive watersheds had significantly lower disturbance thresholds on annual streamflow. An increase of 1 in hydrological sensitivity would result in a decrease of 7% of the disturbance threshold.
3.2 Effects of Climate
Forest disturbance thresholds significantly decreased with increasing snow fractions (p < 0.01, Figure 4a), indicating that watersheds with more annual precipitation falling as snow have smaller forest disturbance thresholds. The multiple regression model (Table 3) also showed and confirmed the significant and negative relationship between snow fractions and disturbance thresholds. Also, forest disturbance thresholds were significantly larger in the equitant system (ranging from 10% to 52% of CECA with an average of 24%) than in the energy-limited system (ranging from 7% to 25% of CECA with an average of 12%) (p < 0.01; Figure 4b). This implies that watersheds with sufficient annual precipitation but more constrained energy availability (e.g., light, temperature, solar radiation) have lower disturbance thresholds.
(a) The relationship between disturbance thresholds and snow fractions. (b) A comparison of the disturbance thresholds (represented by CECA) between EL (energy-limited) and EQ (equitant) environments. Bar denotes standard deviation. *** denotes statistically significant with a p-value less than 0.01.
| Category | Constant | Index | Coefficient with p-value | R2 | p-value |
|---|---|---|---|---|---|
| Climate | 42.41 | Snow fraction | −28.44 [0.100] | 0.59 | <0.001 |
| Ecosystem diversity | Number of BEC zones | 3.40 [0.037] | |||
| Topography | [Log10(Area)] | −8.91 [<0.001] | |||
| IR | 13.89 [0.263] |
- Note. IR denotes water retention index. Numbers in the square brackets denote the p-values of individual factors.
Forest disturbance thresholds in watersheds with dry summer (Dsc) were significantly lower than those in watersheds with no dry season (Dfc) (p < 0.10; Figure 5), indicating that summer water shortage could lead to lower disturbance thresholds. Specifically, forest disturbance thresholds varied from 7% to 18% of CECA with an average of 12% in Dsc and 8%–52% of CECA with an average of 19% in Dfc watersheds, respectively.
A comparison of forest disturbance thresholds (represented by CECA) between Dsc and Dfc environments. Bars denote standard deviation. * denotes statistically significant with a p-value less than 0.10.
3.3 Effects of Ecosystem Diversity
The relationship between forest disturbance thresholds and the number of dominant BEC zones in watersheds with gradual disturbances (three watersheds with considerable abrupt change were removed) was evaluated (Figure 6). As suggested by linear regression (Figure 6), the determined forest disturbance thresholds were significantly and positively correlated to the number of dominant BEC zones in watersheds (p < 0.05), suggesting that watersheds with more diverse ecosystems and vegetation types have higher disturbance thresholds. Additionally, this study also constructed a multiple regression model to test the relationship between forest disturbance thresholds and controlling variables (Table 3). The result also shows and confirms a significant and positive relationship between forest disturbance thresholds and the number of dominant BEC zones.
The relationship between disturbance thresholds (represented by CECA) and the number of BEC (Biogeoclimatic Ecosystem Classification) zones.
3.4 Roles of Topography
As suggested by univariate linear regressions (Figures 7a and 7c), forest disturbance thresholds were significantly affected by watershed size and water retention index. Forest disturbance thresholds significantly decreased with increasing watershed size (Figure 7a), and significantly increased with water retention index (p < 0.10) (Figure 7c). We also found that large watersheds (>500 km2) had significantly smaller disturbance thresholds than small watersheds (<500 km2) (Figure 7b). For large watersheds, forest disturbance thresholds ranged from 7% to 25% of CECA (average = 14%), while 7%–52% of CECA (average = 22%) in small watersheds. The multiple linear regression model confirmed the significant and negative role of watershed size in influencing disturbance thresholds, but the positive role of water retention index was non-significant (Table 3). Thus, both univariate and multivariant analyses consistently confirmed the significant role of watershed size in forest disturbance thresholds, while the role of water retention index was inconclusive due to inconsistent results from the statistical tests.
(a) The relationship between disturbance thresholds and watershed sizes. (b) A comparison of disturbance thresholds (represented by CECA) between large and small watersheds with standard deviation (bar). * denotes statistically significant with a p-value less than 0.10. (c) The relationships between disturbance thresholds and water retention index.
4 Discussion
4.1 Forest Disturbance Thresholds and the Proposed Framework
Our study applied and tested the proposed framework combining long-term forest disturbance level (i.e., disturbed area and cumulative equivalent clear-cut area, CECA) and the modified double mass curve (MDMC) for deriving the hydrological response curve to systematically determine disturbance thresholds for cumulative and significant hydrological impacts on annual streamflow in forested landscapes in British Columbia. We found that forest disturbance thresholds for cumulative hydrological impacts on annual streamflow ranged from 7% to 52% of CECA with an average of 17%, which is lower than the widely recommended threshold, that is, 20% of forest cover loss summarized from paired watershed experiments (Bosch & Hewlett, 1982; Stednick, 1996). When only disturbed area is considered (ignoring hydrological recoveries from forest regeneration), disturbance thresholds varied from 8% to 52% of disturbed area with an average of 19%. As Wei et al. (2021) indicated, the widely recommended disturbance threshold (20% of forest cover loss) might be problematic because it was not determined from the hydrological response curve, and forest treatments in paired watershed experiments are normally severe (e.g., more than 20% of forest cover change). For example, forest treatments being greater than 20% of the watershed areas account for 95.7% of the watersheds (90 of 94) in Bosch and Hewlett (1982) and 87.4% of the watersheds (83 of 95) in Stednick (1996). The small sample size of the watersheds that have less than 20% of forest disturbance levels as well as the difficulty of constructing long-term and continuous hydrological response curves make paired watershed experiments not suitable for the quantification of forest disturbance thresholds.
In contrast, the forest disturbance thresholds determined with our framework are scientifically defensible. The hydrological response curves through MDMCs can capture forest disturbance thresholds for significant hydrological responses of annual streamflow to gradual and cumulative forest disturbance. Forested watersheds, particularly large ones normally experience multiple disturbance types over space and time (Scherer, 2011), resulting in a cumulative and gradual increase in forest disturbance levels. These gradual changes in forest disturbance levels provide scenarios ranging from no or limited disturbance to severe disturbance. As the level of forest disturbance increases and reaches a critical point (i.e., forest disturbance threshold), the relationship between forest and water is significantly altered, and a new state of hydrological responses occurs, which is captured by the modified double mass curve (the hydrological response curve). The successful applications on these 42 watersheds demonstrate that the modified double mass curve is an effective and robust hydrological response curve for quantifying forest disturbance thresholds in any single watershed where long-term data are available.
The results from the MDMCs are well validated by experimental and observation-based studies (Figure S3 in Supporting Information S1), suggesting that our method is robust. The significant breakpoints in hydrological response curves of W9 (Fishtrap), W15 (241 Creek), and W16 (Dennis Creek) were caused by wildfires and treatment-based logging in 2003, 2007, and 2002, respectively. The unpredictable wildfire and logging activities produced large jumps of CECA from 18% to 44% in W9, 30% to 48% in W15, and 42% to 52% in W16, and significant hydrological changes were caused in the subsequent years. The significant hydrological responses in W15 and W16 captured by the hydrological response curves in this study are consistent with the results from paired watershed experiments by Winkler et al. (2021), which revealed significant changes in hydrological responses when 47% and 52% of forests in W15 and W16 were removed. In this study, the forest disturbance thresholds determined from the hydrological response curves were 48% of CECA or 48% of disturbed area in W15 and 52% of CECA or 52% of disturbed area in W16 (Figure S2 and Table S4 in Supporting Information S1). In addition, the study by Owens et al. (2013) concluded total annual streamflow from 2004 to 2010 increased after the wildfire in 2003 in W9. These three watersheds with higher disturbance thresholds are hydrologically resistant with smaller hydrological sensitivities of 0.73 in W9, 0.79 in W15, and 0.54 in W16 (Table S4 in Supporting Information S1). Another experimental-based example in W10 can also support the estimation from our framework. Moore and Scott (2005) found that annual water yield had a recovery toward pre-harvest conditions beginning in the mid-1990s, while this study found hydrological recovery following forest disturbance recovery after 1992. Although the studies in W9 and W10 did not determine forest disturbance thresholds but suggested increased annual streamflow and hydrological recovery, which are also captured by the MDMCs in this study. These findings between paired watershed experiments (Moore & Scott, 2005; Winkler et al., 2021), measured results (Owens et al., 2013), and this study clearly indicate the robustness of the applied MDMC for estimating significant hydrological responses of annual streamflow to abrupt and event-based stand-replacing disturbance and determining disturbance thresholds. However, the dramatic disturbances might impede us from accurately exploring disturbance thresholds if they may occur prior to dramatic disturbance changes.
The study found a significant and negative relationship between hydrological sensitivities and forest disturbance thresholds (Figure S4 in Supporting Information S1), indicating that more hydrologically sensitive watersheds have lower disturbance thresholds. An increase of 1 in hydrological sensitivity would result in a decrease of 7% of the disturbance threshold. As suggested by Ponce Campos et al. (2013), a hydroclimatic threshold is needed to understand which resilience or resistance will break down to predict the consequences of future climate change and forest disturbance on water yields. The finding from our study can provide such information. Although forest disturbance thresholds and hydrological sensitivities are different concepts, they are both related to watershed resistance. Hydrologically sensitive watersheds are typically characterized by unfavorable climatic conditions and lower water retention capacity (Creed et al., 2014; Zhou et al., 2015), which have less ability to withstand external changes and likely lead to more negative hydrological consequences caused by disturbances. Thus, in hydrologically sensitive watersheds, a lower level of forest disturbance (lower disturbance thresholds) can produce significant hydrological responses (higher hydrological sensitivities).
4.2 Impacts of Climate on Forest Disturbance Thresholds
Our results (Figure 4a and Table 3) show that forest disturbance thresholds decreased with increasing snow fractions, suggesting that watersheds with more annual precipitation falling as snow have smaller forest disturbance thresholds. An increase of 10% in snow fraction would result in a decrease of 6.8% of the disturbance threshold (Figure 4a). This result clearly suggests that forest disturbance thresholds are greatly affected by snow processes in British Columbia. Stednick (1996) also found that the disturbance levels for measurable water yield increases vary among climate conditions owing to hydrological rain, rain and snow, and snow-dominated precipitation patterns. In snow-dominated watersheds, forest disturbance alters snow dynamics, such as peak snow water equivalent (SWE), snowmelt rates, and snow duration (Pomeroy et al., 2012), and the changes in snow processes are more important in heavier snowfall areas or years (Aygün et al., 2022; Boon et al., 2012). These considerable changes in snow processes and the corresponding hydrological processes in heavier snowfall years or areas suggest more hydrologically sensitive systems and thus lower forest disturbance thresholds. In addition, snow responses to forest disturbance in heavier snowfall watersheds (e.g., boreal regions) are more complicated, variable, and less consistent (Burenina et al., 2012; Onuchin et al., 2021; Wei et al., 2022). Forest disturbance in heavier snowfall regions with more complicated snow interactions would be more easily to produce significant and cumulative hydrological responses, which potentially make disturbance thresholds smaller.
This study showed that energy-limited watersheds had significantly lower disturbance thresholds (Figure 4b). Having a closer look at annual snow fractions in energy-limited and equitant watersheds in this study, we found that energy-limited watersheds also had significantly higher snow fractions (Figure S5b in Supporting Information S1). The lower disturbance thresholds in snow-dominated and energy-limited watersheds might be due to the changes in surface energy availability (e.g., albedo, net shortwave radiation, temperature) affected by forest disturbance as it greatly controls evapotranspiration and snow processes (Foster et al., 2016; Stigter et al., 2021). The increasing surface energy availability after stand-replacing forest disturbance, such as timber harvesting and wildfire, can be particularly considerable since it can raise temperature and increase evaporation (De Frenne et al., 2021). In contrast, non-stand-replacing forest disturbance can affect surface energy and hydrological processes differently (Boon, 2009), which may lead to different disturbance thresholds, a subject requiring future research. In spite of limited studies on disturbance thresholds in energy-limited watersheds, Fu et al. (2023) estimated the evapotranspiration sensitivities to vegetation change using an improved Budyko framework and showed a similar finding. They found that in regions where water availability is sufficient and energy is limited (similar to our study watersheds), evapotranspiration changes are more sensitive to vegetation growth.
Interestingly, we found that the synchronization of energy demand and water supply at the intra-annual scale also affects threshold variations. Watersheds with dry summer (Dsc Köppen-Geiger classification) had significantly lower thresholds than watersheds without a dry season (Dfc Köppen-Geiger classification) (Figure 5), resulting from the considerable soil moisture deficit and slow vegetation growth. In Dsc watersheds, the maximum energy timing and water timing are mismatched, and those watersheds with the desynchronized energy demand and water supply tend to be water-limited during summer and energy-limited during winter (Feng et al., 2019). In snow-dominated British Columbia, forest disturbance would further extend the seasonal dry period due to early snow melting (Winkler et al., 2017). Water supply is considerably constrained by limited precipitation while energy demand is still high, which is more likely to produce severe soil moisture deficit in the extended summer growing season (Bower et al., 2005). In this way, vegetation regeneration and growth after disturbance become more constrained due to increased soil moisture deficit (Dodson & Root, 2013), and the slower growth of vegetation plays a limited role in buffering and mitigating the negative hydrological responses. Thus, a smaller level of forest disturbance could lead to significant changes in annual streamflow, and the disturbance thresholds are lower in watersheds with more desynchronized timing of energy demand and water supply.
4.3 Roles of Watershed Properties in Forest Disturbance Thresholds
Forest disturbance thresholds for hydrological impacts on annual streamflow significantly increased with the number of dominated BEC zones, representing ecosystem diversity (Figure 6 and Table 3). Watersheds with more diverse ecosystems tend to be hydrologically resistant as they have a greater ability to maintain hydrological functions even under higher levels of forest disturbance (Creed et al., 2014; Zhang et al., 2017). Thus, disturbance thresholds are larger in these watersheds. Watersheds with more diverse ecosystems are normally featured with more vegetation types, more diverse landscape patterns, and greater spatial heterogeneity in vegetation composition (Ferraz et al., 2013), suggesting a more resistant ability. Once forest disturbance occurs, these resistant systems show a greater buffering capacity to forest disturbance (Ellison et al., 2017; van Dijk et al., 2012), and consequently higher disturbance thresholds.
Interestingly, we found that forest disturbance thresholds significantly decreased with increasing watershed sizes (Figure 7a and Table 3) and forest disturbance thresholds were significantly smaller in large watersheds (Figure 7b), suggesting greater hydrological responses to forest disturbance in large watersheds. This is contradictory to the general perception that larger watersheds have greater buffering capacities in constraining changes in peak flows (Blöschl et al., 2007; Filoso et al., 2017; Huff et al., 2000; Zhou et al., 2015). However, forest disturbance thresholds in this study were determined based on the separated hydrological effects caused by forest disturbance, which are different from the response of peak flow magnitudes. Nevertheless, our results are consistent with the studies by Li (2018) who detected the amplified effects of forest disturbance on annual streamflow variations with increasing watershed sizes in the southern interior of British Columbia, and by Hou et al. (2023) who identified significantly larger hydrological sensitivities to forestation operations in large watersheds globally. We speculate that the lower forest disturbance thresholds in larger watersheds might be due to more amplified effects from a larger number of tributaries, disturbed patches, and more forest fragmentation (Fischer et al., 2021; Tinker et al., 1998) over space and time. More research is needed to understand possible controlling mechanisms.
Water retention index (IR) can potentially affect forest disturbance thresholds as it indicates water retention capacity to buffer hydrological changes caused by forest disturbance. Our univariant analysis (Figure 7c) showed a significant and positive relationship between water retention index and forest disturbance thresholds. However, the multivariant test suggested a positive but non-significant contribution (Table 3). The inconsistent results from these statistical tests made our inference on the relationship between IR and forest disturbance thresholds inconclusive. Future studies on the role of IR should consider a larger sampling size and include other topographic factors that may better represent water retention capacity.
4.4 Implications for Forest and Watershed Management
Forest disturbance threshold is a critical disturbance level in forested landscapes above which significant hydrological impacts can be detected. Determining disturbance thresholds is challenging in forested watersheds but important for forest and watershed management. Quantitative evaluations of forest disturbance thresholds can greatly help manage negative hydrological impacts to ensure sustainable water resources in a changing environment. Forests provide the most stable and highest quality water supplies among all land uses (Liu et al., 2021) but are facing considerable pressures due to global climate change and anthropogenic activities (FAO, 2020). The loss of forests threatens sustainable water resources for human beings and water-related ecosystem services. Determined forest disturbance thresholds can provide scientific support to protect hydrological functioning. For example, the Okanagan timber harvesting guidelines suggested that ECA should be lower than 20% in community watersheds, 25% in fisheries-sensitive watersheds, and 30% in all other watersheds for potential changes in peak flows following logging in the Okanagan Basin (Winkler & Boon, 2017). Our estimated average threshold is generally lower than those recommended for Okanagan timber harvesting. However, our thresholds are based on annual streamflow, while those recommended are for peak flows.
In 42 forested landscapes in British Columbia, we found that the average forest disturbance threshold is 17% of CECA (Figure 3 and Table 2). According to the estimated CECA in 70 watersheds in the central interior of BC (Giles-Hansen & Wei, 2021), forest disturbances in 52.9% of the total watershed number and 55.8% of the areas have crossed the average threshold (17%). Such a high percentage (more than 50%) suggests a critical need to consider new forest management strategies to cope with forest disturbance and their effects on water and water-related functions (Gronsdahl et al., 2019; Palmer & Ruhi, 2019), particularly under climate change impacts and an increasing forest disturbance context.
Management strategies, for example, reforestation in harvested sites, can help mitigate the increase of disturbance level and could be suggested in watersheds that have lower disturbance thresholds so that significant hydrological impacts would not result, or hydrological functioning can be restored. Due to the unpredictable occurrence and severity of natural disturbances (e.g., wildfire) regionally and around the globe, managing natural disturbances to avoid reaching disturbance thresholds remains a difficult task. Management actions, such as early detection and prescribed fires, can be taken to reduce fuel accumulation, and decrease potential ignitions, thus, reducing the likelihood of severe disturbance (Lindenmayer et al., 2022). Despite the importance of forest disturbance thresholds for forest and watershed management, unfortunately, many watershed management strategies have not incorporated forest disturbance thresholds into their decisions around the globe.
Forest disturbance thresholds vary among climate conditions and watershed properties (Figures 4-7 and Table 3), as well as the choice of hydrological metrics of interest (e.g., annual streamflow, low flows, and high flows). This implies that forest disturbance thresholds are largely watershed-specific. The large variations ranging from 7% to 52% of CECA also clearly suggest that applying forest disturbance thresholds to support forest-water management needs to account for climate (inter-annual and intra-annual) and watershed properties (e.g., ecosystem diversity and watershed size). The one-threshold-fits-all approach is unlikely to reflect the variations in forest disturbance thresholds. Moreover, applying forest disturbance thresholds for significant hydrological impacts needs to consider other important flow variables such as low flows and peak flows, which are beyond the scope of this study.
4.5 Limitations, Uncertainties, and Implications for Future Study
This is the first-ever study that applies a practical and robust framework to determine forest disturbance thresholds for significant and cumulative hydrological impacts on annual streamflow in the 42 forested landscapes in British Columbia. Even though several existing studies mentioned disturbance thresholds, they focused on assessing hydrological responses to forest change (Bathurst et al., 2018; Bosch & Hewlett, 1982; Buma & Livneh, 2017; Hallema et al., 2018; Oda et al., 2018; Stednick, 1996). Our study fills a critical gap that very limited quantitative determinations were done to identify forest disturbance thresholds for cumulative and significant hydrological impacts in any individual watershed. The proposed framework combining dynamic forest disturbance history and hydrological response curve allows us to determine disturbance thresholds when watersheds experience cumulative and gradual disturbance of different types. This framework is applicable to any individual watersheds around the world where long-term data on forest disturbance, climate, and hydrology data are available. Due to increasing forest disturbances and climate change impacts around the globe, more research on this topic is urgently needed in many other regions to support forest and watershed management for healthy hydrological functions in the future.
We suggest future studies need to apply dynamic indicators to better represent forest change levels. The determined forest disturbance thresholds represented by CECA which considers forest dynamics and hydrological recoveries are more reasonable and practical compared to the static indicator (e.g., the disturbed area). Hydrological responses are never static and are affected by the interactions among climate conditions, tree characteristics, disturbance types, disturbance severities, and post-disturbance trajectories (Hou et al., 2022; McEwen et al., 2020; Moore et al., 2004; Ren et al., 2021; Skubel et al., 2015). These critical traits need to be represented by a comprehensive and integrated forest change indicator such as CECA applied in this study. We understand that the static forest disturbance proxy, that is, disturbed area (or forest cover change) is easy to access around the globe, but they hardly represent the dynamic nature of vegetation changes. When long-term vegetation data are not available for deriving indicators such as CECA, some remote sensing-based vegetation indices such as leaf area index (LAI), and normalized difference vegetation index (NDVI) at a larger spatial scale can be used to represent dynamic forest changes.
In this study, we determined disturbance thresholds in 42 forested landscapes in British Columbia, Canada by use of the MDMC as the hydrological response curve. The MDMC assumes climate variability would not disturb the stationarity of hydrological processes (Wei & Zhang, 2010). If non-stationary hydrological processes are caused by climate variability in watersheds, using the MDMC might cause uncertainty and should be with caution. Although Yang et al. (2021) revealed that historical climate change has not led to non-stationarity in annual streamflow in most global watersheds, alternative hydrological response curves that can capture the non-stationary hydrological responses to forest disturbance are needed to determine disturbance thresholds. Besides, the MDMC can well capture significant hydrological responses to dramatic forest disturbance (e.g., logging in W15 and W16 and wildfire in W9) when long-term data are available. However, rapid forest disturbance and their hydrological effects in small watersheds might not be well reflected gradually in the MDMC so the forest disturbance thresholds might not be accurately captured from the curves. In this sense, the MDMC is more suitable for large watersheds where more gradual and cumulative forest disturbance often takes place.
Forest disturbance thresholds are affected by climate and watershed properties. The higher values of forest disturbance thresholds in W9, W15, and W16 with abrupt disturbance might dominate the relationships between disturbance thresholds and environmental factors, and consequently introduce uncertainty. Additional univariate and multivariate analyses have been conducted to compare the results between the inclusion and removal of these three watersheds (Tables S5 and S6 in Supporting Information S1). The comparisons further confirmed the significant and positive relationships between the disturbance thresholds and the number of BEC zones, and the significant and negative relationships between disturbance thresholds and snow fraction and watershed size, which support the robustness of our findings. Another multiple linear regression model of disturbance thresholds and watershed sizes, snow fractions, and the number of BEC zones showed all significant impacts of individual factors, which reinforces our findings (Table S7 in Supporting Information S1). However, more statistical analyses using non-linear techniques (e.g., machine learning) should be conducted to explore non-linear relationships between environmental variables and disturbance thresholds in the future when the sample size is sufficient. In addition, the inconsistent results showing the impact of water retention index on disturbance thresholds suggest that the index may not be a significant factor for explaining the variations in those disturbance thresholds. This calls for consideration of other watershed properties, such as soil, landscape patterns, and geology in future studies.
Additionally, disturbance thresholds are largely subject to the choice of hydrological metrics of interest (e.g., annual streamflow, low flows, and high flows). Hydrological variables at a short time interval, for example, low flows or high flows, link with critical hydrological functions and risks such as floods and droughts, and play an important role in aquatic habitat, fish flows, and riparian ecology (Palmer & Ruhi, 2019; Poff et al., 1997; Poff & Zimmerman, 2010). An understanding of forest disturbance thresholds for significant hydrological impacts on low flows or high flows is critically needed.
5 Conclusions
Forest disturbance threshold is a critical disturbance level in forested landscapes above which significant hydrological impacts can be detected. This study applied and tested a proposed framework combining dynamic forest disturbance history and the hydrological response curve to determine disturbance thresholds and their contributing factors in 42 forested landscapes in British Columbia. Overall, forest disturbance thresholds range from 7% to 52%, and the average forest disturbance threshold (17% of CECA) is lower than the previously suggested value (20% of forest loss). Watersheds with greater snowfall proportions, more desynchronizations of energy demand and water supply, less diverse ecosystems, and larger watershed sizes have lower forest disturbance thresholds. We estimate that more than 50% of watersheds in the central interior of British Columbia have already crossed the average disturbance threshold, which has important implications for forest management for the provision of hydrological functions in the context of increasing forest disturbance and climate change impacts. Determining disturbance thresholds is challenging in forested watersheds but important for forest and water resource sustainability. More research is critically needed to study forest disturbance thresholds for other hydrological variables and in different regions around the globe.
Acknowledgments
We would like to thank the editor, the associated editor, and two anonymous reviewers for their constructive comments and suggestions on this paper. We are grateful to Jinyu Hui from University of British Columbia Okanagan for helping with the calculations of forest disturbance levels. This research was jointly supported by Natural Sciences and Engineering Research Council of Canada, Discovery Grants Program (No. RGPIN-2021-02628), and by the British Columbia Ministry of Forests through a contract with University of British Columbia (Okanagan campus) (No. RE24REM084).
Conflict of Interest
The authors declare no conflicts of interest relevant to this study.
Open Research
Data Availability Statement
The data of watershed characteristics and data used for analysis in this study (Hou, 2023) are available at the University of British Columbia's cross-disciplinary data repository “Scholars Portal Dataverse” via https://doi.org/10.5683/SP3/GDCQZN. ERA5-Land data (Muñoz-Sabater, 2019) are openly available through Copernicus Climate Change Services at https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-land-monthly-means?tab=overview and the Köppen-Geiger climate map (Beck et al., 2018) can be obtained from http://www.gloh2o.org/koppen/. Forest disturbance data are openly available through the Government of British Columbia at https://catalogue.data.gov.bc.ca/dataset/harvested-areas-of-bc-consolidated-cutblocks- and https://catalogue.data.gov.bc.ca/dataset/vri-historical-vegetation-resource-inventory-2002-2022-, respectively. Hydrologic data are openly available through Water Survey of Canada at https://wateroffice.ec.gc.ca/search/historical_e.html. Canadian Digital Elevation Data (CDED) are openly available via Natural Resources Canada at https://maps.canada.ca/czs/index-en.html.
