1. Introduction
Land managers addressing forest ecosystem sustainability face complex challenges of integrating societal and ecological objectives for myriad natural resources that are prioritized as ecosystem services, inclusive of wood, water, species, and cultural factors [
1,
2,
3]. Historically, disturbances managed relative to forest ecosystems have been related primarily to resource extraction approaches that retain designated priorities such as ecological structures, processes or functions (e.g., Pacific Northwest moist coniferous forests [
4]). Contemporary forest landscape planners are managing these factors in addition to considering documented effects of climate variability and projections of climate change on forest resources and ecosystems for the development of climate-smart forest-management designs [
5]. Although new information is now accumulating regarding how spatio-temporal variability in climate metrics intersect with various individual forest ecosystem services [
6], integration of studies across forest taxonomic and ecosystem boundaries is needed to address combined management and ecological efficiencies. In particular, integration of forest ecosystem-management approaches for both aquatic- and upland terrestrial systems relative to the known effects of climate variability and potential futures of climate change is of paramount importance.
A division between aquatic- and upland-management planning in forests has been evident, as natural resource specialists have addressed their areas of expertise separately (e.g., fish biology and aquatic ecology, wildlife biology, forest-stand management), before assembling their individual pieces into a larger project or landscape plan. This approach can be seen in the development of best management practices (BMPs) in the United States that have targeted different legal requirements for each category of resource, such as streamside buffer zones to address water-quality standards (e.g., the US Clean Water Act), habitat set-asides for sensitive-species protections (e.g., the US Endangered Species Act), and multi-resource environmental assessments conducted to support forestry practices (e.g., the US National Forest Management Act). However, adaptive management of these processes to better integrate the separate pieces as well as incorporating new knowledge as it becomes available is an essential piece of forest-landscape management. For example, the recent update to the US Forest Service’s Planning Rule [
7] includes new language about management of broader ecosystem services as well as addressing integration of forest landscape issues such as habitat connectivity, watershed protection, wildlife conservation, and resilience to climate change.
Furthermore, renewed recognition that both aquatic-riparian areas and uplands in forests are heterogeneous systems over space and time has emerged from US Pacific Northwest forest syntheses [
8,
9,
10], adding complexity to broader forest landscape sustainability goals, integration across ecosystems therein, and management of climate-smart forests [
5,
6]. Spatiotemporal complexity can be generated by gradients in conditions (e.g., with geographic, geologic and hydrologic conditions, also with latitude, distance from coastal mesic influence, rain-shadows) as well as repeated pulse disturbances, some of which have their roots in extreme climate events (e.g., wind-throw, ice damage, drought, flood) and can be aggravated by human activities (e.g., fire and past fire suppression; past timber-harvest practices including tree cutting and road and culvert construction affecting erosion and debris-flow events). Dynamic forested ecosystems are especially evident from our increasing knowledge of unique species and species-assemblages that result from disturbances, species interactions, and species’ life-history variation including dispersal limitations. In the US Pacific Northwest, this is exemplified by the number of rare or little known late-successional forest species that are afforded site-specific protection under the Northwest Forest Plan survey-and-manage provision [
11,
12,
13].
Forested headwaters have been proposed as ideal places for the intersection of aquatic-riparian and upslope management for ecological sustainability ideals such as biodiversity management, and can also be useful to address dynamic forest ecosystems under climate change [
14,
15,
16]. In forests of montane landscapes, topographic relief is responsible for numerous headwater streams extending toward ridgelines throughout watersheds, with the spatial extent of headwater streams encompassing 50–80% of forested watersheds [
17,
18,
19,
20]. This affects the riparian-upland forest areas designated to buffer stream values, such as stream-buffer provisions extending to ~70- to 140-m away from streams on federal lands managed under the United States’ federal Northwest Forest Plan [
21], resulting in an increased focus on the ecological values of headwaters and a need to better understand management practices for sustainability of natural resources for ecosystem integrity and commodity production. Headwater values include distinct floral and faunal composition or structure [
22,
23,
24,
25,
26,
27,
28,
29,
30,
31] and headwater connectivity to downstream aquatic-riparian networks for the delivery of down wood [
32,
33,
34] and sediment [
34,
35,
36,
37] for habitat structure, contributions of cool water for cold-adapted northwest fauna such as economically-important salmonid fishes [
19,
38,
39], and prey for downstream fish communities [
40]. Headwaters offer upland connectivity opportunities as well, as they often are the closest areas to ridgelines, and over-ridge habitat connectivity to adjacent watersheds [
15,
16]. For the conservation principle of redundancy of protections, an especially important tenet when systems are dynamic in nature, the multitude of headwaters provide an opportunity for managing redundant headwater stream reaches—for persistence of their associated in situ populations per reach or reach-network, as well as for dispersal habitat connections over ridgelines or downstream. For aquatic-riparian dependent species as well as terrestrial species, dispersal routes along riparian corridors can funnel individuals into narrowing headwaters before a relatively short over-ridge distance can be traversed to a novel watershed; for example, aquatic and terrestrial amphibians may move along headwater riparian areas, disperse over forested ridges, and retain gene flow among populations in adjacent watersheds [
16,
41,
42].
In the Pacific Northwest, in particular, assessments of climate variability and projections of climate change in forested headwaters are important owing to the dominance of these drainages in the landscape, and their importance to multiple downstream and upland ecosystem services—yet we are just beginning to understand climate influences on forested headwater streams. Stream-temperature and streamflow changes are attributes often considered in watershed studies of climate change, as these factors affect stream ecological conditions and functions, especially species habitat suitability via myriad ecophysiological constraints [
38,
39,
43]. In the Northwestern US more generally, at larger scales (i.e., larger landscapes and watersheds), climate projections include warmer temperatures [
44,
45,
46,
47] and altered precipitation patterns such as reduced snowpack—especially at low-to-mid elevations [
48,
49], which have many headwater basins in the region. These changes may affect streamflow, in particular, which is also driven by landscape-scale drainage efficiency in converting precipitation to discharge [
50,
51,
52]. A sensitivity map for summer streamflow across Oregon and Washington, USA showed variable patterns with geographic conditions, including areas in the Cascade Range being particularly vulnerable to low summer streamflows owing to volcanic geology with deep groundwater systems [
52]. Runoff- and groundwater-dominated watersheds contrast, with groundwater-dominated streams being especially sensitive to changes in snowmelt amount and timing [
53]. Less focus on understanding variation in headwater streamflow patterns has occurred, as so far the focus has been on examining watersheds at larger scales, with field streamflow measurements more often assessed at downstream tributary junctions. Discontinuous streamflows in headwater stream reaches can translate to logistical challenges for stream gaging methods that are often used for surface-flow pattern assessment.
Interactions of local geographic conditions with forest-vegetation and stream size are also important considerations for streamflow dynamics. Upland forest management can affect streamflows: generally, timber harvest is thought to decrease evapotranspiration, increasing soil moisture storage, runoff, and streamflows, with patterns following intensity of disturbance and catchment area effects [
54,
55,
56,
57,
58,
59,
60]. Yet patterns and processes may vary with site conditions (e.g., climate, soils, topography, vegetation type, harvest type, time since harvest, season [
54,
59,
61], and general forest ecosystem type: (1) in mesic systems, increased streamflow may result from reduced canopy interception of precipitation and evapotranspiration, leading to more water draining towards stream channels, especially in wet seasons; (2) in arid systems, there may be less change in flow in all but the wettest systems owing to different processes of rainfall interception, soil saturation, and vaporization; and (3) in subalpine systems with arid summers, there may be little response in summer but increased flows during snowmelt [
62]. Relative to streams in small watersheds in the Pacific Northwest moist coniferous zone, a pattern of increased streamflow following harvest has been supported [
61,
63,
64]. For example, in one study [
63], previously intermittent streams became perennial, and summer streamflow enhancement was attributed to increased zones of deep perennial saturation. Conversely, another study reported lower streamflow in small streams in 34- to 43-year-old plantations, compared to older forests (150–500 years old) [
65], suggesting an uncertain amount of variation may be occurring at local scales and perhaps with past site histories. At this time we have limited knowledge regarding the comparative magnitudes of timber-harvest and climate-variation effects on streamflows, or their interactions. However, a study of streamflows in Arizona, USA forests reported forest changes were associated with streamflow variation to an extent similar to that of climate variability [
66].
Understanding how climate variation and forest management interact with streamflow patterns in basins of different sizes and geographic contexts, and consequent effects on other stream resources such as biota, involves a complex web of interactions that is gaining research and management attention. With increasing risks to natural resources from joint disturbances including forest-management practices and climate change projections, studies of the dynamic contexts and consequences of multiple-threat interactions are becoming more important [
67]. Aquatic-dependent biota, in particular, are fast-declining and have been considered a ‘weak link’ in forest ecosystem integrity [
8]; they deserve heightened attention for the interactive threats of habitat loss from anthropogenic disturbance and climate change projections, two of their most important risk factors today. Forested headwater streams and their associated riparian areas provide important reproductive, foraging, and dispersal habitat for unique assemblages of forest-dependent species, especially endemic amphibian species in Pacific Northwest moist coniferous forests [
25,
68]. Previous studies have shown that the composition of aquatic and riparian animal communities in northwest forests is associated with habitat characteristics of headwater reaches, with unique assemblages associated with both perennial-continuous and intermittent-discontinuous portions of headwater stream networks, e.g., Western Oregon: [
24,
25,
68,
69]. However, relationships between headwater streamflows, climate, forest management, and geographic contexts is not well examined by empirical studies.
To advance our understanding of the joint effects of climate and forest management on stream surface flows in small headwater basins, we conducted a retrospective study with summer low-flow data over 16 years, field-collected from headwater streams with discontinuous water flow in forest stands subjected to an experimental harvest regime in Western Oregon, USA. We hypothesized that summer dry-season headwater surface-flow stream lengths would be reduced in past years with reduced precipitation and warmer temperatures. But we also predicted that there could be a signature of the upland forest-harvest regime on summer headwater streamflows, with potentially increased flows in headwater streams having the narrowest streamside riparian buffers (three buffer widths assessed, and unharvested reference streams) in areas with upland thinning and hence relatively more upland soil saturation and perennial runoff. We projected that interactions of buffers and climate might occur, but did not predict an outcome because we had little a priori basis for understanding relative magnitudes of these dual effects for headwater stream drainages in our study area. We also considered headwater basin area per stream reach as a geographic variable affecting streamflow, with larger basins likely collecting more precipitation and having more discharge and surface streamflows. We then used our empirical findings from field-collected surface streamflow patterns collected over this 16-year time interval, along with annual climate variability, to parameterize models of landscape-scale climate change to project effects into the future. Future climate projections for a case-study forested landscape area were then computed to assess potential cumulative effects of changes to headwater surface streamflow patterns, especially relevant to aquatic-dependent forest species with headwater associations. We used the range of the Cascade torrent salamanders (
Rhyacotriton cascadae Good and Wake), a headwater-associated species found in discontinuous streams in the Cascade Range that is proposed for listing under the USA Endangered Species Act, to exemplify the potential consequences of headwater surface streamflow changes. We used this salamander to represent the larger assemblage of biota in the uppermost headwater habitats of these stream systems in young forests at low to mid-elevations. For example, faunal assemblages at headwater streams at our study sites also have been previously described as unique due to increased density and biomass of Diptera (flies) [
24].
4. Discussion
In a retrospective study, we assessed the variability in forested headwater-stream surface flow at 65 stream reaches from 13 managed forest study sites across 16-years, and examined streamflow associations with basin size, climate metrics, and timber harvest treatments with alternative riparian buffer widths in a before-after-control-impact forest harvest experiment. We focused on small streams with discontinuous flow during spring wet seasons, as this was the dominant stream-reach hydrotype occurring at our study sites [
68]. We hypothesized that there would be reduced summer dry-season stream lengths in past years with reduced precipitation and warmer temperatures. We predicted that there could be an interacting signature of the forest harvest regime (i.e., stream-buffer width with upland thinning) on summer headwater stream length, but timber-harvest effects in headwater streamflows have not been well studied, and our context included upland thinning with streamside riparian-buffer zones. In larger stream systems, in contrast, there can be increased streamflow with upland overstory canopy reduction resulting from increased discharge [
54,
55,
56,
57,
58,
59,
60,
61,
62,
63,
64]. Using our study results, we analyzed future climate projections in the study area to assess potential changes to headwater streams at the landscape scale.
Among 27 stream-reach habitat and flow metrics calculated, the single stream metric that best characterized spatial and temporal variation in surface flow among discontinuous stream reaches and time periods in our study was the percent dry length (% dry length) of reaches. As all study streams had spatially intermittent (discontinuous) streamflow during spring wet seasons, the % dry length represented dynamics in surface water occurrence across reaches and time periods. This metric was used in additional analyses of patterns with regard to climate metrics, basin area, and harvest treatments.
When analyzed separately, % dry length was associated with nine of 22 climate factors (
p < 0.05;
Figure 6,
Table 5) and headwater basin area of stream reaches in our sample (
p = 0.002,
Figure 7), but not riparian-forest buffer treatment after upland thinning. In our final multivariate model of % dry length relative to all factors, two climate metrics (summer heat: moisture index, SHM; mean minimum summer temperature, Tmin sm) and basin area retained significance (
Table 6). There were reduced surface streamflows in our discontinuous stream reaches during warm-dry summers and in basins less than 12.6-ha in area. This supports our initial prediction of reduced headwater streamflow during drier, warmer years, with smaller basin areas having an additional effect on surface flows.
Although our analyzed harvest-and-buffer treatments were not associated with % dry length in our univariate or multivariate analyses, further assessments are warranted. We had implementation constraints on deployment of wide buffer treatments in small basins, which may have reduced the power of our analyses. Simply, we were unable to deploy wider buffers in the smallest headwater basins while retaining sufficient area between the upslope buffer edge and the ridgeline to implement a thinning treatment. In our study design, we aimed to have at least 60 m between the upslope edge of the buffer and the ridgeline for the thinning to occur [
77], and such a distance was not always possible if a wide buffer was used in a small watershed. Despite our attempts to do a random treatment design, this type of geographic constraint limited our options in headwaters. Additionally, our heterogeneous upland thinning regimes may have exerted differential effects on streamflow, possibly interacting with different buffer widths. Potential confounding effects of drainage area, buffer width, and upslope harvest regimes could be further examined in other studies. However, if larger drainages and streams were addressed, different surface flow metrics likely would need to be analyzed. Nevertheless, our variable implementation of buffers and thinning regimes in headwater basins is likely operationally representative of many forest landscape management practices in our region. In particular, forest managers have guidance to use narrower riparian buffers along non-fish bearing discontinuous streams, and our study had ample replication to examine effects of narrow buffers (6-m, ≥15 m) on streamflow in those types of aquatic systems, and no buffer effect on streamflow was observed.
We found that the effects of future climate change scenarios on SHM and Tmin sm resulted in significant increases in headwater stream dry-channel lengths (i.e., reduced stream-reach surface flows). Modeled discontinuous streams in headwater forests were projected to continue shrinking, with the effect varying across three future time periods, dependent upon climate change scenario analyzed (
Figure 8). In year 2055, the range of the three scenarios resulted in a 4.5%–6.4% loss of streamflow, resulting in additional dry channels summing to 805–1145 km of first-order stream lengths and 597–849 km of streams in basin sizes of 12.6 ha or less in the modeled area. By 2085, 7.1%–11.5% losses of headwater stream reaches were projected, a reduction of 1270–2058 km of first-order streams and a loss of 941–1525 km of surface streamflow in reaches in 12.6-ha basins. It should be noted that headwater systems can account for 50%–80% of watersheds [
17,
18,
19,
20], especially within Pacific Northwest landscapes with steep montane topographies like those we studied at our 13 sites in the Coast Range and Cascade Range of Western Oregon. This abundance of small streams is evident as we summed projected reductions in streamflow across the case-study landscape of the western Cascade Range.
Our landscape projection was aimed to assess what shrinking headwater stream habitats could mean for endemic, sensitive biota tied to this portion of the forested landscape. Focusing on a vertebrate species with sensitive status that is known to be highly associated with discontinuous streams
Rhyacotriton cascadae: [
68,
69], the headwater stream lengths projected to be lost across the species range were extensive, with potentially dramatic predictions for altered surface-water habitat conditions and likely over-ridge habitat connectivity. The 2085 climate projections of 7.1%–11.5% streamflow loss by first-order stream channel length translated to additional dry channels summing to 1270–2058 km in length in the modeled landscape of the range of the Cascade torrent salamander. Because of the possible mismatch in stream surface-flow hydrotypes between modeled first-order streams (possibly perennial streams) and our study streams (discontinuous streams), we also calculated effects of shrinking headwater streams in small basins, which were likely a better-matched subset to our discontinuous streamflow hydrotype. In headwater basins of <12.6 ha, a more conservative streamflow loss was projected in 2085: 941–1525 km of headwater streams would no longer have surface flow. Furthermore, most of this projected surface streamflow loss occurred in lower-elevation bands of the modeled Cascade Range (
Figure 4), which had greater modeled stream lengths, and coincided with most of the known sites of the target salamander species considered.
There are several potential consequences of shrinking headwaters for this target species, sister taxa within its species complex (
Rhyacotriton spp.), and other associated taxa of headwater stream species. First, dry surface-water units along stream channels do not appear to be occupied by this
Rhyacotriton [
68]. It is likely that torrent salamanders undertake vertical migration during summer drought periods, and can move down into substrates to retain moisture. They may also be able to track moisture patterns and move to nearby surface-water units or moisture refuges in the riparian zone. However, downstream perennial reaches may be more inhospitable owing to their occupancy by potential predators including giant salamanders (
Dicamptodon spp.), coastal cutthroat trout (
Oncorhynchus clarkii clarkii Richardson), and sculpins (Cottidae) [
68,
87,
88]. Downstream retention of surface water with upstream drying may cause increased trophic interactions, and result in reduced surface activities or increased microhabitat cover use by prey species [
88]. Alternatively, it is possible that current small perennial streams will become discontinuous in the future, and this may result in a shift in the predator-species complex further downstream, and thus possibly increased occurrences of torrent salamanders in those reaches. An interacting concern is altered water temperatures with drying or warming climates, as this is a coldwater-associated species complex (
R. variegatus Stebbins and Lowe occupied habitats, 6.5–15 °C; thermal stress at 17.2 °C) [
89]. At one of our study sites, headwater stream thermal regimes were variable across the headwater reach network, especially in summer, likely because of thermal exchanges at local within-site scales [
90]. Although the effects of surface-water drying and possible warming on torrent salamander reproductive success or survival can only be conjectured at this time, it is likely that foraging habitat at the ground surface would be altered, and the active season of the animal likely would be reduced due to lack of moisture or interacting factors such as predators; each of these factors could reduce survival and reproduction.
Upland forest detections of
Rhyacotritron spp. allude to their complex life history, with both aquatic and terrestrial life forms [
91], and the importance of both stream-riparian and upland forest habitats for the animals. Although little is known about their breeding habits,
Rhyacotriton spp. eggs have been discovered in low-flow headwater streams, and some have been known to incubate there through the spring and summer, hatching after up to 290 days; larvae may be restricted to headwater streams for up to five years [
91]. Such reliance on headwater-stream habitats could pose survival issues for drying and warming conditions. After metamorphosis and gill absorption, adults can move to forested uplands, where their longevity is unknown. Other studies have detected members of the
Rhyacotriton species complex 30 m from streams using time-constrained searches [
92], and 200 m from streams using pitfall traps [
93]. To date, mark-recapture studies to track torrent salamander movements have not been reported. Upland habitat associations also are little known, but these are highly moisture-dependent species occurring in the temperate rain forests of the US Pacific Northwest Coast and Cascade Ranges. At one of our study sites, 11 upland captures of
R. variegatus were reported at artificial cover-boards, suggesting they may use down wood on the forest floor as a microhabitat refuge [
41]; down wood can retain cool, moist conditions [
94]. A landscape genetics study across the Oregon Coast Range found forest cover was a top predictor of
R. variegatus and
R. kezeri Good and Wake gene flow, likely owing to associated moist microhabitats [
42]. Consequences of potentially hundreds of km of future drying of portions of the uppermost headwater streams occupied by this species complex could place ecophysiological constraints on inter-stream dispersal and reduced habitat connectivity among populations. If reliable surface water is located lower in basins in the future, then it is likely that populations may shift downstream and over-ridge dispersal would require moving longer distances—either perpendicular to streams or linearly up drying channels. Longer overland dispersal could expose these salamanders to prolonged dry conditions and terrestrial predators such as shrews. Longer distances between adjacent headwater streams can reduce net movements among adjacent drainages, as torrent salamander leg length is only ~1–2 cm, and they are not known to be highly mobile. Increased population fragmentation is likely. Two of the four
Rhyacotriton species are proposed for listing under the US Endangered Species Act, including the Cascade torrent salamander. Effects of our climate and landscape projections on discontinuous stream surface-flow changes are additional threats to consider in the context of management strategies to help retain their long-term persistence.
Retention of headwater surface flows and their link to shorter-distance dispersal routes over ridgelines likely would aid persistence of these species and other biota in their assemblages. Some stream-riparian restoration and management activities may promote these conditions in concept see Table 6 in [
14]; also [
15,
16,
41], although few empirical data from headwater forest management are available at this time, suggesting that management approaches may need to be revisited as knowledge develops [
95]. Prevention of streambank erosion and channel infilling from management activities would help retain headwater stream surface flow, and this could be ensured by equipment exclusion zones along streams and restricting yarding operations that drag logs across streams. This type of protection is a primary aim of riparian buffers. In a landscape that is highly managed for timber commodities, set-asides for rare species habitats and habitat connectivity can reduce other interacting anthropogenic disturbances that could affect forest habitat conditions by further reducing surface-water availability, such as activities linked to landslides further from stream channels. A network of headwater sub-drainage reserves has been proposed for Pacific Northwest forests to address sensitive headwater species and habitats in some portions of a larger watershed, where other portions could have timber-management priorities [
14,
15,
16,
41,
96]. Our landscape model suggests that there are benefits for locating these reserves in lower-elevation portions of the western Cascade Range landscapes to overlap a large number of headwater sub-drainages and sensitive species ranges. These headwater-retention areas could be linked by riparian corridors to downstream reaches and overland to neighboring basins by managing ground structures and vegetation conditions [
14,
15,
16,
41]. Additional landscape-scale modeling to examine likely distributions of hill-shaded headwater streams, cold-water basins, and persistent snow cover for late-season runoff could further assist climate-smart management [
5,
6,
16], where overlap of multiple compatible factors also could result in provisions to provide other natural resource benefits, including habitats for other sensitive species such as Pacific salmonid fishes [
38,
39] and clean water for growing urban and rural human communities.