Changes in suspended‐sediment yields under divergent land‐cover disturbance histories: A comparison of two large watersheds, Olympic Mountains, USA

Improvements in timber harvest practices and reductions in harvest volumes over the past half century are commonly presumed to have reduced sediment loads in many western US rivers. However, direct assessments in larger watersheds are relatively sparse. Here, we compare 2019–21 sediment concentrations against those of the late 1970s in the Bogachiel and Calawah River watersheds, adjacent and similarly sized (~300 km2) basins in the western Olympic Mountains of Washington State. The Calawah River watershed has experienced significant land‐cover disturbance, including a large 1951 fire, extensive post‐fire salvage logging, and relatively high rates of timber harvest through the 1990s. In contrast, the Bogachiel River watershed did not burn, and experienced only modest timber harvest that largely post‐dated 1970s sediment monitoring. Channel‐width trends suggest the Calawah River was still recovering from 1950s disturbances in the late 1970s. We found that 2019–21 suspended‐sediment loads in the Calawah River were 2.3–2.6 times lower than would have been expected based on 1970s sediment rating curves, while recent loads in the Bogachiel River were a factor of 1.4 ± 1.0 lower. We consider the plausibility and possible explanations of declining concentrations in the less‐disturbed Bogachiel River. Suspended‐sediment yields in the Bogachiel River were two times higher than yields in the Calawah River, which is attributed to a combination of modestly higher precipitation, more efficient runoff generation, and more extensive and erodible Quaternary valley fills in the Bogachiel River. Regional shifts in flood hydrology have also influenced suspended‐sediment loads in both watersheds. Our results then document a significant decline in suspended‐sediment concentrations in the Calawah River over the past half century. Reduced land‐cover disturbance provides the simplest and most likely explanation for this decline, though the wide range of possible concentration changes in the Bogachiel River leaves open possibilities that other processes (human, natural, or methodologic) could be a factor.


| INTRODUCTION
Excess suspended sediment in streams and rivers degrades water quality and aquatic habitat (Arismendi et al., 2017;Bilotta & Brazier, 2008;Greig et al., 2005;Kemp et al., 2011), and remains an ongoing concern in the Pacific Northwest Region of the United States for economically and culturally important populations of salmonids (Jensen et al., 2009). It is well established that mid-20th century forest harvest practices in the United States led to significant increases in suspended-sediment delivery (Beschta, 1978;Brown & Krygier, 1971;Grant & Wolff, 1991;Lieberman & Hoover, 1948;Richardson et al., 2018). However, over the last several decades, implementation of best management practices-including application of vegetated buffers along streams, improved logging road construction, and installation of sediment traps in ditches and along culverts, among others-have become widespread (Cristan et al., 2016;Croke & Hairsine, 2011;Ice et al., 2004). These improved practices have generally been associated with reduced, though not always absent, post-logging impacts, including mass wasting associated with poor road construction, elevated coarse and fine sediment delivery to rivers, and reduced water quality (Cristan et al., 2016;Fraser et al., 2012;Goodman et al., 2023;Hatten et al., 2018;Klein et al., 2012;Lewis, 1998;Madej et al., 2012;McBroom et al., 2008).
Most detailed studies of sediment response following timber harvest in the western United States have focused on relatively small watersheds (e.g. Gomi et al., 2005;Goodman et al., 2023;Hatten et al., 2018), which provide more opportunity for rigorous control/ treatment study designs, intensive monitoring, and identification of explicit process linkages (e.g. Lewis et al., 2001). However, understanding channel and sediment regime responses to such land-cover disturbances in larger rivers is important, as downstream alluvial regions of the network provide essential functions, including channel conveyance during floods (Wyżga et al., 2016) and aquatic migratory and spawning habitat (Geist & Dauble, 1998). Further, understanding the evolution of channel change to historical disturbances offers a broader perspective and potentially more informed baseline (McMillan et al., 2022) that can be used in management decisions going forward and provide insight on the sensitivity to potential future disturbances, including those associated with climate change (Khan & Fryirs, 2020). However, drawing definitive conclusions between channel response and basin-scale management activities in larger rivers remains a challenge, in part because of difficulties of controlled experiments to target cause and effect (Al-Chokhachy et al., 2016;Croke & Hairsine, 2011).
There is good reason to expect that forestry practices would have measurable effects in larger (>100 km 2 ) watersheds. In northern California, the combination of logging impacts and major mid-20th century floods resulted in significant increases in average sediment concentrations and overall sediment transport in multiple large rivers, with decreasing suspended-sediment concentration (SSC) in subsequent decades associated with improved logging practices and less active flood hydrology (Warrick et al., 2013). Similarly, lake cores from a 240 km 2 catchment in western Oregon's Coast Range identified elevated sediment accumulation in the mid-20th century, attributed to intensive forestry practices coupled with large floods (Richardson et al., 2018). Mid-20th century accumulates exceed both background accumulation rates of prior centuries and subsequent modern rates following passing of forestry practices regulation in the 1970s and a transition to a less active flood hydrology. In the western Olympic Mountains, Washington, significant positive correlations between logging road density and fine sediment intrusion in spawning gravels indicated that elevated sediment delivery in the 1970s was impacting both small and large watersheds (Cederholm & Salo, 1979;Cederholm et al., 1981). In the Deschutes River of Washington, 30 years of turbidity monitoring (an established surrogate for suspended sediment; Gray & Gartner, 2009) showed a progressive improvement in water clarity starting in the 1980s, attributed to improved road construction and best management practices implementation (Reiter et al., 2009). It then seems reasonable to presume that suspended-sediment loads in many western Washington rivers with managed watersheds were elevated as a result of mid-century logging practices, and that those loads have declined as harvest practices improved and harvest rates generally declined (Cristan et al., 2016;Croke & Hairsine, 2011;Ice et al., 2004). While direct assessments of changing SSC in western Washington rivers have so far not identified clear changes attributable to changing land-use patterns (e.g. Gray, 2018), most studies have focused on rivers draining glaciated stratovolcanoes (Anderson & Konrad, 2019;Curran, Grossman, Magirl, & Foreman, 2016;Curran, Grossman, Mastin, & Huffman, 2016), which likely overshadow sediment inputs from other sources (e.g. Czuba et al., 2011).
Here, we assess how SSC in the modern era compares against those of the late 1970s (Nelson, 1982) in two adjacent and similarly sized ($300 km 2 ) watersheds in the western Olympic Mountains of Washington State. We treat this as an imperfect paired-watershed study, with the Calawah River acting as the treatment basin and the Bogachiel River acting as the control; the Calawah River watershed has experienced significant land-cover disturbance, including a large 1951 fire, extensive post-fire salvage logging, and relatively high rates of clearcutting through the 1990s (O'Connor & Cundy, 1993). In contrast, the Bogachiel River watershed is largely within Olympic National Park and has experienced only modest timber harvest, limited to the lower portion of the study area basin and largely post-dating the 1970s sediment sampling. Based on existing literature, we would expect to see significant declines in suspended-sediment yields in the Calawah River watershed as a result of reduced timber harvest activity and decades of recovery from fire and subsequent post-fire salvage logging disturbances. We do not expect changes in sediment availability in the Bogachiel River watershed, which has largely been unchanged over the study period. We also used historical aerial imagery between 1939 and 2017 to quantify the historical extent of land-cover disturbance (fire or logging) in the two watersheds and evaluated channel width changes to provide additional context of watershed conditions and disturbance response.  (Gerstel & Lingley, 2003). Steep hillslopes in the headwater tributaries of both watersheds are prone to landslides, a common characteristic of the steep, forested, mountain system (Dieu & Shelmerdine, 1996;O'Connor & Cundy, 1993;Smith & Wegmann, 2018).
Overall relief is greater in the Bogachiel River (434 m), with maximum elevations that extend higher in the Olympic Mountains relative to the Calawah River basin (455 msl vs. 308 msl) (Table S1) through a wide valley that ranges between 1 and 3 km in width ( Figure 1).
Land-cover in both the Calawah River and Bogachiel River study area basins is forested with Sitka spruce and western hemlock dominating along the river valley (Franklin & Dyrness, 1973 In 1951, the Great Forks Fire (approximately 154 km 2 ) swept through the Calawah River basin but did not extend to the Bogachiel River basin (Figure 1). Fire is a natural, but catastrophic, disturbance in the wetter forest regions of the coast region of the Pacific Northwest (Halofsky et al., 2018;Littell et al., 2010), with multi-century recurrence intervals for stand-replacing fires (Huff, 1995). Intensive salvage logging and associated road building occurred in years following the fire in the North Fork Calawah River and portions of the South Fork Calawah River, including sidecast roads often associated with hillslope failures and shallow landslides (Cederholm & Salo, 1979;Dieu & Shelmerdine, 1996;O'Connor & Cundy, 1993  Total aerial extents of forest clearing were summarized for the contributing areas of the Calawah River and Bogachiel River stream gauges for each photo year. The cleared area in a given photoset aggregates clearing that occurred over some number of years preceding the nominal photo year, though that integration window is not exactly known. The active channel was delineated as the wetted channel area and proximal unvegetated fluvial surfaces (Cadol et al., 2011;East et al., 2017;O'Connor et al., 2003). The river channel centreline was defined as the centre of the wetted area. When multiple channels were present, the centreline of the largest channel was traced.
Mean active channel width was calculated as active channel area divided by centreline length. Channel delineation extended from the

USGS stream gauge locations on the Bogachiel River and Calawah
River to RKM 53 on the Bogachiel River, RKM 29 on the North Fork Calawah River, and RKM 15 on the South Fork Calawah River ( Figure 1). We compare these values to reported channel width change in the Hoh River, another forested, unregulated river that is the next major river south of our study area (East et al., 2017; Figure 1).
To characterize uncertainties in channel width estimates, three additional individuals re-digitized the active channel area and centrelines in seven 1-2 km sub-reaches along the two rivers (see online Supporting Information for details). Channel widths for subreaches along the two rivers were averaged for each of the four digitizers and year of imagery. We report digitizing error as the standard deviation of the river-averaged channel widths from the four total digitizers for each year of imagery.
T A B L E 1 Historical imagery used in basin land-cover and channel planform change analysis. Imagery type abbreviations: B&W, black and white; SFP, single-frame photo; DOQ, digital orthophoto. Year

| Contemporary suspended-sediment loads
Contemporary suspended-sediment loads were monitored over water years 2019-21 (WY19-21) at USGS stream gauges on the Calawah River and Bogachiel River ( Figure 1). Suspended-sediment monitoring was based on continuous discharge and turbidity records coupled with discrete suspended-sediment measurements (Anderson et al., 2022) used to calibrate SSC-turbidity rating curves, following standard USGS methods (Edwards & Glysson, 1999;Rasmussen et al., 2009). Values of SSC and turbidity were closely related (R 2 = 0.96 on both sites). We used the large set of turbidity-estimated SSC values to develop SSC-discharge rating curves, providing a good characterization of the average SSC-discharge relation over a wide range of flows. This rating curve was used to estimate sediment concentrations during gaps in the turbidity record. Over the WY19-21 period, 81% (Bogachiel River) and 99% (Calawah River) of the integrated suspended-load estimates were turbidity regressions, with the remaining load estimated using discharge regressions. Details of data collection and load estimation, including SSC-turbidity and SSCdischarge regression models, are presented in Anderson et al. (2022).

| Comparison with historical rating curves
SSC data were previously collected across the Calawah and Bogachiel watersheds between 1975 and 1979 (Nelson, 1982), allowing us to compare contemporary and historical sediment conditions in both watersheds. The comparison methods below attempt to isolate the impact of changing SSC-discharge relations on suspended-sediment loads, holding discharge constant, similar to analyses presented in Safeeq et al. (2020).
At the Calawah River, USGS data archives include 137 SSC measurements collected between 1975and 1978, and Nelson (1982 published a two-part power-law regression fit to those data. We used two additional approaches to estimate SSC-discharge rating curves from the original measurements. First, we re-fit a segmented two-part linear relation to the data after log-transforming all variables to achieve homoscedasticity (e.g. a piecewise power-law relation) using the 'segmented' R package (Muggeo, 2008). Second, we took the SSC-discharge rating curve derived from WY19-21 monitoring and identified the scaling factor that provided the highest coefficient of determination (R 2 ) when compared against 1970s SSC measurements.
This approach assumes that the functional form of the SSC-discharge relation has remained constant, and that any change involved simple vertical translations (in log-log space) of that relation. This scaling approach was selected based on the apparent similarity of the functional form of the SSC-discharge relations across the two eras. All three equations were applied to 15-min discharge records over the WY19-21 period, providing a range of expected WY19-21 suspended-sediment loads if 1970s rating curves were still in effect.

| Relative changes in water yield
Land-cover change can impact suspended-sediment loads directly, through increased sediment delivery associated with hillslope disturbance, or indirectly, in cases where increased discharge results in higher sediment transport capacity and/or accelerated erosion of inchannel sediment sources (Lewis et al., 2001;McEachran et al., 2021;Reid et al., 2010;Safeeq et al., 2020).
We assessed whether the relation of daily mean discharges at the Bogachiel River and Calawah River sites differed between the 1970s monitoring era and recent discharge monitoring. To accomplish this, we first fit a loess (Cleveland & Devlin, 1988) curve to the relation between discharge at the two sites over the post-2018 monitoring period. We then used that loess relation to predict discharges at the Calawah River over the 1970s monitoring period based on concurrent discharge at the Bogachiel River and looked for systematic deviations between observed and predicted discharge. We repeated this process using both untransformed and log-transformed values of daily discharge and using a variety of loess 'span' parameters. Two-sided ttests with unequal variance were used to assess differences. Because we are limited to daily mean discharge values for this analysis, we are unable to assess if there have been changes in hydrograph shape (e.g. flashiness) at the sub-daily level, particularly if there was no corresponding change in overall daily water yield.

| Assessing long-term hydrologic trends
We used discharge records from the Hoh River to assess how longterm changes in flood hydrology may have impacted total suspended sediment loads passing the two monitored sites. We applied static sediment rating curves to the long-term hydrologic record to estimate annual loads. The potential for rating curve changes over time means the resulting load estimates are unlikely to be valid measures of true sediment transport outside the window over which samples were collected (Warrick & Rubin, 2007;Warrick et al., 2013 POR 1976POR -19801984-present), with R 2 = 0.91. This regression was then used to estimate daily discharge in the Calawah River from 1926 to present.
We estimated suspended-sediment loads over this full interval using both our modern SSC-discharge rating curve and a rating curve fit to 1970s measurements. As will be shown, rating curves from the two eras show little change in functional form and can be related by a simple scaling factor. Consequently, load estimates from the two rating curves are likewise essentially scaled versions of the same relative trends in hydrology. We refer to these trends as 'static rating load estimates'.

| Suspended-sediment concentration and loads
Over the 2019-21 monitoring period, the mean annual suspendedsediment yield in the Bogachiel River basin was 620 tons/km 2 /yr ( instantaneous runoff depths in the Bogachiel River were on average about 18% higher than concurrent values in the Calawah River during high-flow events, albeit with substantial event-to-event variability (Figure 5a). In addition to greater runoff depth, SSC tends to be about 20-25% higher in the Bogachiel River than in the Calawah River under comparable runoff conditions (Figure 5b).
The rating curve crossover at $2 m 3 /s/km 2 occurs at or above the upper limit of conditions observed over the monitoring period, making it unclear if the reversal is real or an artifact of extrapolation.

| Comparison to historical sediment yields
At the Calawah River, WY19-21 load estimates based on 1970s SSC-discharge rating curves were 2.3-2.6 times higher than those estimated via turbidity monitoring (Table 2; Figures 6 and 7). This indicates a substantial decline in sediment concentrations for a given discharge relative to conditions 50 years ago. The shift in concentrations appears to be relatively uniform across all flows, with no obvious change in functional form; the 1970s rating curve estimated as a scaled version of the WY19-21 curve is largely indistinguishable from those fitted directly to 1970s SSC measurements ( Figure 6).

| Water yields
Formal statistical tests confirm visual impressions that the discharge relation between the two sites is not detectably different between the two monitoring eras (Figure 8). To the degree that any (non-significant) differences exist, they tend to show that for a given runoff in the Bogachiel River basin, the Calawah River basin yielded slightly less water in the 1970s than in the modern era, counter to expectations that forested land-cover loss would tend to increase water yields.
The absence of detectable changes in relative water yield between the two sites makes it unlikely that indirect hydrologic impacts of forest clearing had a significant impact along the major

| Discharge-driven changes in loads
The application of static SSC-discharge rating curves to the esti- Assuming that the static rating load estimates from the two rating curves are reasonable approximations of the true sediment loads for F I G U R E 5 Representative period of instantaneous runoff (discharge/drainage area) for the Calawah River and Bogachiel River (a) and relation between suspended-sediment concentration (SSC) and runoff depth for the Calawah River and Bogachiel River (b); concentrations are based on SSCdischarge rating curves derived from calibrated turbidity records (Anderson et al., 2022). In the Calawah River watershed, SSC for a given discharge has dropped substantially since the late 1970s, with observed WY19-21 F I G U R E 7 Load ratios for the Calawah River and Bogachiel River, comparing loads based on 1970s sediment information versus information from WY19-21. Ratios for the Calawah River, based on three different 1970s rating curves seen in Figure 6  ( Figure 4). Further, high-impact logging practices and high rates of forest clearing persisted through the 1990s (Figure 3) and were identified as a key driver of fine sediment deposition in the neighbouring Clearwater River during this era (Cederholm & Salo, 1979), making it likely that these practices also continued to elevate loads in the Calawah River. The combination of a recovery from the 1951 Great Forks Fire and post-fire salvage logging, improved logging practices, and reduced logging rates then seems to provide a succinct and plausible explanation for the observed decrease in SSC in the Calawah River. That such a recovery occurred seems corroborated by the significant decrease in channel widths through the 1980s and 1990s, contrary to increasing regional flood activity and widening trends in less-disturbed watersheds nearby (Figure 4).
This straightforward interpretation is modestly complicated by results implying that the Bogachiel River may also have seen declining concentrations over this same interval, albeit of a smaller magnitude and with sizable uncertainty. Given that the Bogachiel River water- show a distinct period of elevated concentrations from 1973 to 1980, relative to measurements before or after (Gray, 2018); holding discharge constant, loads estimated from rating curves during the period of elevated SSC were about 1.6 times higher than those using a 'nonelevated' rating curve. This implies that concentration changes on the order of 60%, similar to our best estimate of relative changes in the Finally, it is plausible that the apparent change in the Bogachiel River is simply a methodological error, and that the true change is closer to zero. The simplest scenario here would be that Nelson's (1982) published annual loads for the Bogachiel River, based on limited (and undocumented) SSC sampling, were biased high. This would leave us with the mass balance-estimated change ratio of 1.4 ± 1.0, which readily includes the possibility of no change. We tend to take the available raw SSC measurement data (including Nelson's (1982) SSC sampling) as unbiased, given that SSC sampling and analysis methods were, by the 1970s, standardized through the Federal Interagency Sedimentation Program (FISP, 1963) and a common element of USGS monitoring. However, given the limited methodological details provided by Nelson (1982) and the absence of raw SSC data for the Bogachiel River, we are unable to definitively rule out methodological error.
We ultimately lack the data to materially constrain these various  (Cederholm & Salo, 1979;O'Connor & Cundy, 1993; Figure 4), we ultimately judge that recovery from 20th-century land-cover distubance is the most likely cause of declining concentrations in the Calawah River, while readily acknowledging the possibility that the exact magnitude of that change, or the magnitude specifically attributable to land-cover disturbance, may be lower than indicated.
More generally, these results also highlight that sediment availability in a watershed is often not stationary. The use of a single SSCdischarge rating curve, or application of a rating curve to discharge records well beyond the period over which sediment samples were collected, may result in substantial errors in estimated loads (Gray, 2018;Warrick et al., 2013).

| Applicability to other watersheds
Given the relatively unique nature of the 1951 Great Forks Fire, it is reasonable to consider whether the observed declines in concentration in the Calawah River are likewise unique within the region. Exploration of 1950s photography (available through Earth Explorer; https://earthexplorer.usgs.gov/) shows that the extensive and total clearing of timber observed in the Calawah River basin in 1955 ( Figure 3a) was common in many watersheds around the Olympic Mountains and western Cascade Range, and reflected typical logging practices as opposed to a unique response to the fire. Sediment sampling in the 1970s also occurred more than 20 years after the major land-cover disturbances of the 1950s, which is near the upper end of the time window over which land-cover-related sediment disturbances have typically been observed (Gomi et al., 2005;Warrick et al., 2013). It then seems likely that higher concentrations in the 1970s are at least partially attributable to ongoing disturbance from contemporaneous forest practices in the Calawah River basin, which were common across the region. Based on all of the above, we believe that the sediment dynamics we have observed in the Calawah River basin are likely qualitatively similar to nearby managed watersheds. However, differences in landscape characteristics (gradient, position relative to streams, geology) likely modulate the impact of forest clearing (Bywater-Reyes et al., 2017;Goodman et al., 2023), making direct quantiative extrapolation to other watersheds difficult.

| Cross-basin differences in suspendedsediment yields
Over the WY19-21 monitoring period, the less-disturbed Bogachiel River watershed had nearly twofold higher suspended-sediment yields than the Calawah River watershed ( Table 2). The higher yields from the Bogachiel River watershed were the result of the watershed both yielding more runoff per unit area and each unit of water containing, on average, higher sediment concentrations ( Figure 5). We further explore possible underlying physical causes for these differences.
The simplest explanation for higher instantaneous and mean runoff depths in the Bogachiel River is higher annual precipitation rates; basin average annual precipitation for the Bogachiel River basin is 13% higher than in the Calawah River basin (Table S1). Higher precipitation in the Bogachiel River basin is primarily associated with high rates of orographic precipitation within the core of the Olympic Range, which the Calawah River basin does not drain (Figure 1).
Much of the main-stem and North Fork Calawah River are underlain by highly permeable continental glacier material (Figure 1) Cundy, 1993). No such features are observed in the Bogachiel River.
Higher losses to groundwater in the Calawah River basin may therefore contribute to an overall lower runoff depth for the basin.
Even holding runoff depths constant, concentrations in the Bogachiel River are consistently higher than in the Calawah River.
Channel planform comparisons also suggest higher relative sediment transport in the Bogachiel River, which is broad and braided with extensive bare gravel deposits, in contrast to the relatively confined and single-threaded Calawah River ( Figure 10). One plausible could also contribute to higher concentrations in the Bogachiel River.

| CONCLUSIONS
We compared channel width and SSC trends for two adjacent and similarly sized forested mountain watersheds with different fire and timber harvest histories. We observed a significant increase in channel width in the Calawah River following a 1951 fire and associated salvage logging, with subsequent channel narrowing through the 1980s and 1990s. These trends were distinct from climate-driven width trends observed in the less-disturbed Bogachiel River and Hoh River, indicating a unique period of land-cover-related disturbance and recovery in the Calawah River. This was corroborated by comparisons of concentration-discharge relations in the Calawah River from 2019 to 2021 against those from the late 1970s, which show that concentrations for a given discharge were 2.3-2.6 times lower in the modern era than in the 1970s. Attribution of declining SSC-discharge relations in the Calawah River solely to recovery from mid-20th century land-cover disturbance is complicated by ambiguous (1.4 ± 1.0) relative concentration declines in the minimally disturbed Bogachiel River basin, leaving open the possibility that factors other than land-cover disturbance may also have contributed to declining concentrations in both watersheds. Regardless, given the magnitude of the decline in the Calawah River, and multiple lines of evidence of land-cover disturbance causing elevated loads during the 1970s, we judge it most likely that recovery from those 20th-century land-cover disturbances was a major cause of declining concentrations in the Calawah River. We were unable to detect any changes in runoff efficiency in the Calawah River basin associated with 1970s land-cover disturbance. However, regional increases in flood activity in the late 1970s likely contributed to elevated sediment loads in both watersheds through the early 2000s; flood activity has generally been trending downwards in recent decades. Modern suspended-sediment yields in the Bogachiel River were twice as high as those in the Calawah River, reflecting a combination of higher precipitation rates, more efficient runoff generation, and higher SSC. We speculatively link the latter two factors to the extent and character of valley-filling unconsolidated glacial deposits. Our results largely confirm expectations that improved timber harvest and reduced harvest rates since the 1970s have substantially reduced SSC and overall yields, while highlighting the large role of geology and precipitation in spatial variations of sediment yields.

DATA AVAILABILITY STATEMENT
Data presented in this study are openly available in NWIS (2022) or as ScienceBase Data Releases (Anderson et al., 2022;Landstedt et al., 2023).