Fires increasingly impact forested watersheds, with uncertain water resources impacts. While research has revealed higher peak flows, longer-term yields may increase or decrease following fire, and the mechanisms regulating post-fire streamflow are little explored. Hydrologic response to disturbance is poorly understood in the Lower Colorado River Basin (LCRB), where snowmelt often occurs before the growing season. Here, we quantify annual streamflow changes following what have been, before 2020, two of the largest wildfires in the modern history of the contiguous United States. We evaluate nine nested watersheds with >50 years records within the Salt River Basin to evaluate fire impact over ranges of elevation, climate, vegetation, burned area, and spatial scale. We employ double-mass comparison of paired watersheds, pre- and post-fire runoff ratio comparison, multiple linear regression of climate and fire, and time-trend analysis. Precipitation and streamflow are decoupled during dry periods; therefore we conduct separate change detection for wet and dry periods. Post-fire summer streamflow increased by 24%–38% at all elevations. While winter/spring streamflow remained constant in the highest, coldest headwaters, winter flows declined in lower-elevation headwaters. As a result, basin annual streamflow declined. These results support emerging understanding that warm semiarid watersheds respond differently to disturbance than well-studied, colder watersheds. Asynchrony between winter snowmelt and summer evaporative demand is likely important when considering long-term impacts of forest management and disturbance on water supply in the LCRB.

Key Points

Summer streamflow increased in headwaters at all elevations following fire
Dominant winter/spring streamflow was unchanged in higher/colder headwaters but decreased in lower/warmer headwaters
Climatological asynchrony of snowmelt and transpiration in warmer watersheds may reduce streamflow benefits of fire

Plain Language Summary

Wildfire is increasingly common and severe in many of the forested watersheds important for water supplies. Following fire, there is an increased risk of short-term flooding. However, we do not understand how wildfire changes the amount of water flowing out of a watershed over multiple years. Although wildfire leaves fewer trees to take up water, it also destroys the shade from sun and wind which protects snowpack and soil moisture from evaporation. Here, we made side-by-side comparisons and before-after comparisons to determine wildfire impacts on the multiyear streamflow from nine watersheds of the Salt River Basin in Arizona. We found that streamflow increased in summer. While the much larger winter/spring streamflow did not change much at high elevations, it declined in lower-elevation watersheds following fire. One reason for this difference might be that at high elevation, the snow melts at the start of the summer growing season, when trees are likely to take up the water. Wildfire reduces trees and thereby increases streamflow. At lower elevations, snow melts much earlier in the year, when trees are not active, making the water savings from burned forests less important. These results suggest that lower, warmer forested watersheds may produce less streamflow following wildfire.

1 Introduction

Forested mountain headwaters are of critical importance to water supplies globally, particularly in water-limited regions, where precipitation is concentrated in the mountains (Viviroli et al., 2007). In the semiarid Southwest region of the United States (US), 65% of the water supply for municipal, industrial and agricultural use originates on forested land (T. C. Brown et al., 2008), where climate and widespread vegetation disturbances impact hydrologic partitioning between evapotranspiration (ET) and streamflow (Fazel et al., 2017; Milly & Dunne, 2020; Teuling et al., 2019; Woodhouse et al., 2016). Forecasted drier, warmer conditions are also linked to increasing forest disturbance, like wildfire, that can supersede the hydrologic impacts of climatic variability (Abatzoglou & Williams, 2016; Adams et al., 2017; Dennison et al., 2014; Robinne et al., 2021; Westerling et al., 2006). Land managers are grappling with cost-benefit analyses of forest restoration treatments which reduce high-severity fire and may conserve streamflow (O’Donnell et al., 2018; Robles et al., 2014). Stem density, large fires, and human-caused ignitions are all increasing in the US Southwest due to human population increases and the associated legacy of wildfire exclusion (Balch et al., 2017; Marlon et al., 2012; Zou et al., 2008). However, the impact of large fires on hydrologic partitioning in the US Southwest as compared to more well-studied, wetter, colder basins remains a key knowledge gap. Importantly, a national-scale assessment of wildland fire impacts on streamflow found the most significant, positive streamflow responses to fire in the semiarid Lower Colorado region (Hallema et al., 2018). Meanwhile, water managers in the Salt River, an economically vital basin of the Lower Colorado, have been puzzled at the lack of streamflow increases in the 10–15 years following high-severity fires affecting as much as two-thirds of watershed area (B. Svoma, Salt River Project, pers. comm. 2019). Therefore, this study quantifies impacts of large, severe wildfires on long-term streamflow in forested watersheds within the Lower Colorado River Basin (LCRB) of the US Southwest and examines evidence for biophysical controls of disturbance response in warm, semiarid regions.

Studies of disturbance effects on long-term, large-scale streamflow have shown a mix of increases and decreases, and the mechanisms responsible for post-fire streamflow remain poorly explored. Therefore, expectations for long-term streamflow are often based on outcomes from other disturbances, which vary widely (Goeking & Tarboton, 2020). In many environments, wildfire increases streamflow (Benavides-Solorio & MacDonald, 2001; Boisramé et al., 2017; Campbell et al., 1977; Driscoll & Carter, 2004; Helvey, 1980; Wine, Cadol, & Makhnin, 2018), especially over a few years or less (Bart, 2016). Other forest disturbances such as harvest or insect infestation also typically contribute to increasing streamflow (Bethlahmy, 1974; Bosch & Hewlett, 1982; A. E. Brown et al., 2005; Livneh et al., 2015). There are, however, examples of severe disturbance resulting in no streamflow changes or even declines, especially in warmer, semiarid watersheds (Biederman et al., 2015; Guardiola-Claramonte et al., 2011; Stednick, 1996). Recent evidence suggests that disturbance may enhance streamflow sensitivity to both positive and negative anomalies in annual climate and that more arid watersheds are more likely to have streamflow reduction in dry years post-disturbance (Ren et al., 2021). Disturbance effects can be modified by anomalous climate conditions or masked at larger spatial scales due to dilution (Hallema, Sun, Bladon, et al., 2017; Wine & Cadol, 2016). Furthermore, many disturbance studies rely on the runoff ratio or other metrics that may not accurately capture threshold dynamics and/or nonzero intercepts in the precipitation-streamflow relationship (Biederman et al., 2015; Flerchinger & Cooley, 2000), especially during dry seasons or years.

Changes in the biophysical controls of the hillslope water balance collectively regulate the streamflow response to forest disturbance. Biophysical changes unique to wildfire include reduced below-canopy surface roughness and infiltration capacity (DeBano et al., 1996; Neary et al., 2011). While these fire impacts on soils contribute to higher peak streamflow, especially during intense summer storms, the annual contribution of summer streamflow tends to be small (<25%) in many western watersheds (Biederman et al., 2015; Robles et al., 2017). In western North America, snowmelt volumes can be strongly influenced by disturbance through reduced canopy interception accompanying loss of forest shelter from sun and wind (Broxton et al., 2015; Golding & Swanson, 1986; Krogh et al., 2020; Moeser et al., 2020; Sexstone et al., 2018; Svoma, 2017; Veatch et al., 2009). In some cases, peak snow water equivalent (SWE) may be unchanged or even decline following severe wildfire, harvest, or insect infestation, especially on warmer, south-facing hillslopes (Baker, 1986; Biederman, Brooks, et al., 2014; Harpold et al., 2014). Forest disturbance has the greatest influence on snow partitioning in colder, wetter watersheds, where the residence time of intercepted snow in the canopy is greater and the snow-covered season is longer, both of which increase the importance of forest canopy in controlling sublimation losses, which can account for up to 30% of winter precipitation in US Southwest forests (Sexstone et al., 2018; Svoma, 2017). Measurements of hillslope-scale ET by eddy covariance methods integrating sublimation, transpiration, and surface evaporation have likewise shown a range of increases and decreases following disturbance (Biederman, Gochis, et al., 2014; Dore et al., 2010; Frank et al., 2014). Hypotheses for the divergent response of hillslope ET to disturbance include variation in climatic setting, tree species, and the degree and pace of disturbance, but further comparisons across these biophysical gradients are needed to develop generalized expectations (Adams et al., 2012). The timing of snowmelt in a given watershed may also play a critical role in regulating the streamflow response to disturbance. In warm watersheds with earlier snowmelt, streamflow may be decoupled from the vegetation activity and energy availability of the growing season (Knighton et al., 2020; Robles et al., 2021), reducing the water “savings” to be realized from lower transpiration. Because of the differing biophysical controls regulating streamflow response to disturbance in forested mountain watersheds, here we separately consider streamflow arising from winter precipitation (streamflow occurring November–June) and summer precipitation (July–October).

There is a relative paucity of disturbance hydrology studies in warm, semiarid watersheds such as those dominating all but the highest elevations of the LCRB. In contrast to the more well-studied Upper Colorado River Basin (UCRB), characteristic features of the lower basin include a bimodal (winter/summer) precipitation distribution, greater variability of precipitation and snowpack seasonally, annually, and with elevation (Ellis & Sauter, 2017; Hallema, Sun, Bladon, et al., 2017; Svoma, 2011), and total evaporation in excess of 80% in many watersheds (i.e., runoff ratios <20%; Robles et al., 2017, 2021). Surface water resources in both the Upper and Lower Basins are dominated by winter precipitation (Barnett et al., 2004; Molotch et al., 2004). However, peak monthly streamflow in the LCRB, resulting from a combination of winter rains and snowmelt, typically occurs in February–March, well before peak evaporative demand and vegetation activity, whereas UCRB snowmelt typically occurs during April–June, under higher evaporative demand and vegetation activity (Robles et al., 2021). This temporal decoupling of moisture inputs from the summer growing season may alter the relationship between forest water uptake and streamflow response to disturbance in the LCRB (Knighton et al., 2020).

The Upper/Lower Colorado Basin contrast can be viewed through the lens of studies proposing that disturbance will have detectable (positive) impacts on streamflow only when annual precipitation exceeds a ∼500 mm threshold (Adams et al., 2012; Bosch & Hewlett, 1982; Stednick, 1996), presumably because below 500 mm, most precipitation is evaporated regardless of forest condition (Hibbert, 1979). Tree-harvesting studies generally support increased annual streamflow for a minimum of several years following disturbance, but most of those studies have been conducted at locations where mean annual precipitation exceeds ∼500 mm (Baker, 1986; Bosch & Hewlett, 1982; A. E. Brown et al., 2005; H. Brown et al., 1974; Stednick, 1996; Zou et al., 2010). With mean annual precipitation ranging from <300 mm to >1,000 mm, LCRB watersheds span the ∼500 mm threshold spatially. Temporally, many LCRB watersheds also cross this threshold due to high interannual precipitation variability, which increases with aridity (Biederman et al., 2017). Even semiarid watersheds with annual precipitation below the 500 mm threshold may show streamflow response to disturbance if the watershed is snowmelt dominated (Adams et al., 2012). Snowpack storage may delay moisture inputs until the growing season, increasing the importance of any changes in forest water uptake post-disturbance. In the present study, the wide range of watershed elevation, temperature and precipitation provide an excellent opportunity to compare hydrologic response to disturbance in warmer, drier watersheds versus colder, wetter ones.

Here, we leverage two of the largest fires in western US history to investigate the long-term impacts on streamflow across nine gauged Salt River watersheds in the LCRB. The Rodeo-Chediski Fire (2002) and Wallow Fire (2011) collectively affected ∼4,000 km², including 20% of the total area and ∼50% of the forested area of the entire Salt River Basin, which supplies hydropower and the largest in-state renewable surface water supply to the metropolitan area of Phoenix, Arizona. Although these two fires collectively impacted just under half of all forested highlands in the Salt River Basin, long-term annual streamflow did not increase, in contrast with expectations for warm, semiarid watersheds (Hallema et al., 2018; Wine & Cadol, 2016). Here, we address three hypotheses for the lack of increase in annual, basin-scale streamflow for 1–15 years following wildfire within a process-based framework:

H1 – Hydrologic partitioning in headwaters watersheds following large wildfire varies with elevation-dependent interactions between vegetation and hydrologic regime;
H2 – Long-term hydrological sensitivity to fire is moderated by counteracting seasonal responses; and
H3 – Wildfire will only have detectable impacts on streamflow in wet years (above ∼500 mm precipitation) because below 500 mm, annual precipitation and streamflow become decoupled.

Prior regional-to-national-scale studies which attributed increased streamflow to fire in the LCRB have generally filtered for watersheds designated by USGS as reference gauges (least-affected by human impacts; Hallema, Sun, Bladon, et al., 2017; Hallema et al., 2018; Wine & Cadol, 2016). To build upon these prior studies and enable more specific biophysical inferences for warm, semiarid watersheds, here we include all nine nested watersheds within the Salt River Basin. Some of these have human impacts up to several percent of mean annual streamflow, and we accordingly provide corrections and/or evaluations of these human impacts on our results. Additional novel aspects of this study include separate treatment of winter and summer flows, consideration of dry years in which streamflow and precipitation become decoupled, and comparison of two adjacent headwater regions which are warmer/lower and higher/colder and have different snowmelt climatology.

2 Study Watersheds

The Salt River basin spans 11,097 km² in central Arizona, with a large elevation range (640–3,476 m; Figure 1). Vegetation is comprised of mixed conifer forests at the highest elevations, ponderosa pine forests and grasslands at mid elevations, and pinyon-juniper woodlands, chaparral and mesquite shrublands at lower elevations. More than 80% of mean annual streamflow originates in the forested high-elevations and mid-elevations most impacted by fire (O’Donnell et al., 2018). The average annual flow of the Salt River Basin above Roosevelt, AZ is 700 million m³, or 600,000 ac-ft. Precipitation is bimodal due to the influence of the summer monsoon season; however, the majority of annual streamflow (>75%) occurs following winter precipitation and spring snowmelt (Figure 2), hereafter termed winter streamflow.

Details are in the caption following the image — **Figure 1**
Open in figure viewer PowerPoint

Study areas maps including (a) the Salt River Basin in Arizona, United States (inset) and its constituent watersheds, (b) The location of unburned (light green) forests and burned (dark green) forests as well as non-forested burned areas (gray), (c) mean annual precipitation and (d) mean annual air temperature. The Salt River Basin is divided into two subregions (a) where the White Mountains subregion is characterized by higher elevation, precipitation, snowpack, streamflow, and runoff ratios, as well as later snowmelt runoff, and the Mogollon Rim subregion is characterized by lower precipitation, snowpack, streamflow and runoff ratios (Table 1) with earlier snowmelt runoff. Two large wildfires outlined in red in (b) are Rodeo-Chediski (2002) in Mogollon Rim subregion and Wallow (2011) in White Mountains subregion.

We analyzed the nine watersheds (Table 1) within the Salt River basin above Roosevelt, AZ that had long-term records of streamflow (>50 years) and wildfire activity (1985–2017) and minimal and/or well-quantified human disturbance. To test our first two hypotheses, we grouped watersheds into two subregions with distinct climatology and vegetation. These are the higher/colder White Mountain region in the eastern portion of basin, which supplies ∼67% of the basin annual average flow, and the lower/warmer Mogollon Rim region in the western portion, which supplies ∼33% of basin flow. These subregions are nested within the Salt River basin, which has two gages on the main stem (SAC and SAR). Watersheds in the White Mountains have greater forest cover, precipitation, runoff and runoff ratios, lower air temperatures, and later spring snowmelt than those in the Mogollon Rim (Table 1 and Figure 2).

Table 1. Physical and Climatic Characteristics of Watersheds in the Salt River Basin and Its Mogollon Rim and White Mountains Subregions

Watershed	Study code	USGS gauge	Area (km²)	Mean elevation (m)	Forest cover (%)	Fire years	Burn area (%)	^aHigh severity (%)	^bMAT (°C)	^cMAP (mm)	^dMAP winter (%)	Mean annual ^eRR	Mean annual Q (mm)	Q winter (%)	Data period
Mogollon Rim Subregion
Carrizo Creek	CAS	9496500	1,142	1,928	62%	2002	63%	28%	11.2	510	53%	0.06	30	86%	1955–1960, 1968–1974
Carrizo Creek	CAS	9496500	1,142	1,928	62%	2002	63%	28%	11.2	510	53%	0.06	30	86%	1978–2017
Cibecue Creek	CIC	9497800	748	1,746	25%	2002	33%	13%	12.5	513	55%	0.10	50	76%	1960–2008, 2011–2017
Cherry Creek	CHG^C	9497980	517	1,689	19%	–	–	–	13.0	601	58%	0.09	54	87%	1966–1978, 1980–2017
White Mountains Subregion
Black River Point of Pines	BLP	9489500	1439	2,456	82%	2011	71%	26%	7.4	654	53%	0.20	131	84%	1955–2017
Black River Fort Apache	BLF	9490500	3,165	2,201	57%	2011	32%	12%	9.2	602	51%	0.18	106	86%	1958–2017
East Fork White River	EFW^C	9492400	185	2,494	96%	–	–	–	8.0	862	54%	0.19	161	78%	1957–2017
White River Fort Apache	WHF^C	9494000	1,629	2,207	66%	–	–	–	9.3	700	53%	0.14	101	85%	1958–2017
^fSalt River Main Stem
Salt River at Chrysotile	SAC	9497500	7,327	2,058	51%	2002 and 2011	10% + 14%	4% 5%	10.3	589	52%	0.13	77	84%	1955–2017
Salt River at Roosevelt	SAR	9498500	11,097	1,884	39%	2002 and 2011	11% + 9%	5% 3%	11.6	567	53%	0.12	66	84%	1955–2017

^a Burn severity data only available for the 2002 Rodeo Chediski and 2011 Wallow Fires.
^b MAT = Mean annual air temperature from TopoWX 1949–2017.
^c MAP = Mean annual precipitation from PRISM 1949–2017.
^d Winter is November–June, Summer is July–October.
^e RR = Runoff ratio = streamflow (Q)/precipitation (P).
^f Burn area as a percent of watershed area shown are for the Rodeo-Chediski and Wallow fires, the only fires to impact at least 20% of any long-term gauged basin within the period of record. The Salt River Main Stem gauges (SAR and SAC) reached 20% total area affected by the sum of the two landscape-scale fires in 2011. For change detection methods involving a threshold fire year, we evaluated both 2002 and 2011 separately for these main stem gauges.

3 Data Sets

3.1 Streamflow Data

This study used several public data sets. Daily streamflow for each watershed was obtained from the USGS National Water Information System (http://waterdata.usgs.gov/nwis) for hydrologic years 1955–2017. In all watersheds, 95%–99% of reported daily streamflow values were classified as “observed” and “approved” by USGS, meaning suitable for publication. Daily values were processed to monthly and annual streamflow (Q) using a modified hydrologic year appropriate to the bimodal precipitation of the region (November 1–October 31). The hydrologic year was further divided into winter (November–June) and summer (July–October). Streamflow values were normalized by catchment area (i.e., specific streamflow) and expressed as a depth of water (mm) for consistency with precipitation. Some hydrologic change detection studies rely uniquely on reference gauges, identified by USGS as a subset of watersheds with the lowest human impacts in each region. Here, inclusion of all available subwatersheds supported the evaluation of the impacts of varying climate, vegetation, and spatial scales of watersheds on post fire streamflow (H1). Fortunately, human impacts are very low in all watersheds studied here (Robles et al., 2017) and consist of small irrigation withdrawals (Table S5 in Supporting Information S1) and two gauged diversions described in the Annual Water Data Report (http://wdr.water.usgs.gov). Daily gauge records were used to correct a diversion from the Black River (Figure S12 in Supporting Information S1; USGS Gauge 9,445,000) and a diversion into Carrizo Creek (USGS Gauge 949,500). For a detailed evaluation of human impacts on the study watersheds, please see the Supporting Information for Data Section 3.1 Streamflow Data: Potential Impacts of Human Water Use.

3.2 Climate Data

We evaluated three potential sources of information for water inputs. First, spatially averaged monthly precipitation (P) was calculated from the PRISM gridded data set (www.prismclimate.org) at 4-km resolution. Second, we obtained monthly P from SNOTEL stations in the study area with continuous records beginning in 1981 or earlier (http://www.wcc.nrcs.usda.gov/snow). Third, peak SWE at each SNOTEL was evaluated as a predictor of streamflow, because winter flows dominate annual streamflow in the Salt River Basin (>75%), and because peak SWE is the water available for snowmelt after sublimation losses. We evaluated SNOTEL-based P and peak SWE assigned to each watershed using (a) the SNOTEL station nearest to the catchment centroid and (b) a mean value based on the three nearest SNOTEL sites. For the period 1981–2017 during which all three data sources were available, we found that PRISM P was the best overall predictor of seasonal and annual streamflow (correlation with annual streamflow r = 0.82), and PRISM P was therefore used for analyses. For air temperature (T), PRISM and SNOTEL records are known to have artificial trends most likely related to sensor changes. We used TopoWx data with minimum and maximum air temperature averaged for each day, then aggregated monthly.

3.3 Wildfire Data

The Rodeo-Chediski and Wallow fires represent overwhelming step changes in the known fire history of the Salt River Basin (Figure 3). Although numerous small fires have been recorded in the basin's history dating back to 1916, wildfire impacted an average of 0%–1% of the basin annually before the 1990s (Robles et al., 2017). Perimeters of 128 wildfires were obtained from the Monitoring Trends in Burn Severity (MTBS) data website (https://www.mtbs.gov/) for 1984–2016 and from the Geospatial Multi-Agency Coordination data website (https://www.geomac.gov/) for 2017. These were used to calculate fire-impacted (i.e., burned) area. We quantified high-severity burned area using a relativized differenced normalized burn ratio (RdNBR) threshold of 643, corresponding to the midpoint between moderate and high severity classes in a study of 1197 Southwest Composite Burn Index (CBI) field plots (Singleton et al., 2019). In the case where fire perimeters from different years overlapped, we only accounted for the area burned in the first fire year to avoid double counting. This “double burn” issue was minimal due to dominance of the Rodeo-Chediski and Wallow Fires.

Between June 18 and July 7, 2002, the Rodeo-Chediski fire burned 1,900 km², much of this (∼1,200 km²) within the Salt River Basin, including ∼ 900 km² of conifer forest in the higher elevations of the Mogollon Rim subregion (Figure 1). Considering the study watersheds for the current analysis, Cibecue Creek (CIC) was 33% burned (13% at high severity) and Carrizo Creek (CAS) was 63% burned (28% at high severity; Figures 1 and 3 and Table 1). From May 29 to July 8, 2011, the Wallow Fire burned 2,100 km², about half of which (∼1,000 km²) was within the Salt River Basin and ∼800 km² in the conifer forests of the Black River watershed in the White Mountains subregion (Figure 1). The Black River is 81% forested above the Point of Pines gauge (BLP; Table 1 and Figure 1). Most of that forest was affected by the fire (71% of total watershed area and 86% of forest). The remaining watersheds were not impacted by the two large fires (<1%) and remained <20% impacted by all fires accumulated since 1985 (Figure 3).

4 Methods

We used four empirical analysis methods following Biederman et al. (2015). These are presented in order of increasing complexity and data requirements: (a) double-mass streamflow analysis of paired watersheds, (b) runoff ratio comparison, (c) multiple linear regression with annually variable burn area and climate as predictor variables (i.e., watershed burn area may change each year), and (d) time-trend analysis comparing the residuals of a climate-driven linear streamflow model before and after fire. The threshold year required by methods 1, 2, and 4 was determined by two alternative approaches. In the first approach, a watershed was classified as burned beginning the year after cumulative fire impact exceeded 20% of watershed area (Figure 3), a threshold commonly reported as a minimum disturbance area for detecting streamflow response (Adams et al., 2012; A. E. Brown et al., 2005; Hallema et al., 2018; Stednick, 1996). Alternatively, given the overwhelming dominance of the Rodeo-Chediski and Wallow fires in basin history, we used the years of those fires (2002 or 2011) as the thresholds for the relevant watersheds (Table 1). In all watersheds except for the Salt River main stem gauges SAC and SAR, these alternative approaches indicated the same year (i.e., the two biggest fires represent large step changes in cumulative burn area percentage). For these two main stem gauges with contributing area from both large fires, results are reported applying either 2002 or 2011 as threshold years. Statistical comparisons were made at significance levels of 0.05 and 0.10.

4.1 Double-Mass Paired Watershed Comparison

In double-mass analysis (Searcy & Hardison, 1960), the slope of the linear relationship between cumulative seasonal or annual streamflow of a burned watershed is compared against that of a control watershed with similar size and elevation (no suitable control exists for the entire Salt River Basin). We used covariance analysis to test whether the relationship changed after the fire threshold year at the 95% confidence level. Double-mass analysis has been employed in paired watershed studies of disturbances by bark beetles (Bethlahmy, 1974; Biederman et al., 2015; Love, 1955; Potts, 1984) and of timber harvest and related management practices (Leaf, 1975; Zhang & Wei, 2012; Zhao et al., 2010). The streamflow effect of fire was calculated as the difference between the cumulative flow measured in the burned watershed and the flow that would have been predicted in that watershed had it not burned, that is, based on the pre-fire relationship with the control watershed.

4.2 Classifying Wet and Dry Years

Exploratory plots of annual (or winter/summer) Q versus P showed a pronounced breakpoint in many watersheds as opposed to a continuous linear or nonlinear relationship. In dry years or seasons with less than a given threshold P, Q was low but relatively constant (i.e., independent of P), possibly due to groundwater discharge (Garner et al., 2013). Meanwhile, wetter years or seasons when P was above the threshold showed an increasing linear relationship between P and Q. We illustrate this important characteristic using one sample watershed from each subregion (Figure 4, for remaining watersheds see Figure S1 in Supporting Information S1). The threshold P for each watershed and time period (annual, winter, and summer) was determined to minimize the squared streamflow residuals simultaneously across the wet and dry years. We then tested whether the resulting linear model segments were significantly different using analysis of covariance. Threshold behavior was thereby identified in all Mogollon Rim (lower, warmer, and drier) watersheds annually and seasonally. In the higher, cooler, wetter White Mountains, there was evidence of threshold behavior in summer, but not during the winter or annually, when the long-term P–Q relationship could be fit by a single linear model (Figures 4 and S1), similar to basins of the UCRB (e.g., Biederman et al., 2015). Hereafter, years and seasons showing a single, increasing linear relationship of Q with P are termed “wet,” while those below the threshold P and showing Q independent of P are termed “dry.”

4.3 Comparisons of Streamflow, Runoff Ratio, and Climate

We analyzed seasonal and annual streamflow changes following wildfire threshold years using a statistical comparison of pre-fire and post-fire streamflow, where significant changes in streamflow or runoff ratio (RR = Q/P) were investigated for fire effects. For example, a significant increase in post-fire Q without an accompanying P increase would suggest that fire increased Q. Not all streamflow and climate observations were normally distributed (Lilliefors, 1967); therefore, mean climate and pre-fire and post-fire runoff ratio values were compared using the nonparametric Mann–Whitney U-test (Mann & Whitney, 1947), which makes no assumption about sample distribution. Post-fire changes in RR were multiplied by post-fire P to quantify effects on streamflow.

4.4 Linear Regression Modeling With Annually Variable Fire Impacts

Linear regression modeling was used to test for wildfire effects on seasonal and annual streamflow. Model form for each seasonal period (year, winter, and summer) was evaluated before 2002, the date of the first major fire in the modern historical record, by stepwise linear regression. Due to data availability of the predictors, regression modeling started in 1984 for models containing burned area (Section 3.3), in 1981 for models containing SNOTEL peak SWE (Section 3.2), and to 1955 for models based on streamflow and/or temperature (Section 3.2). Models were evaluated for all pre-fire years and seasons classified as wet (i.e., where a relationship existed between P and Q, see Section 4.2). Across the study watersheds, annual Q was best predicted by precipitation, separated into its winter and summer components (average R² = 0.79; average RMSE = 26 mm; Table S1 in Supporting Information S1):

$urn:x-wiley:00431397:media:wrcr25851:wrcr25851-math-0001$ (1)

where a, b, and c are fitted coefficients and the subscripts “w” and “s” denote summer and winter.

Winter streamflow was best predicted using P_w alone (P_s from the preceding summer was not a significant predictor; average R² = 0.83; average RMSE = 24 mm; Table S2 in Supporting Information S1):

$urn:x-wiley:00431397:media:wrcr25851:wrcr25851-math-0002$ (2)

The best model of summer streamflow included P_s as well as the preceding P_w, consistent with a previous analysis in the Salt River Basin showing that winter precipitation was a significant predictor of summer flows (Robles et al., 2017; average R² = 0.56; average RMSE = 8 mm; Table S3 in Supporting Information S1):

$urn:x-wiley:00431397:media:wrcr25851:wrcr25851-math-0003$ (3)

An additional term to account for annual wildfire data was added to streamflow models (Equations 1-3) to test for effects of fires in multiple ways. Fire impacts were evaluated over two spatial domains including the burned percentage of the entire watershed or, alternatively, the burned percentage of forested area. Fire impacts were then quantified using total burned area within the spatial domain, high-severity burned area, or their combination. Each fire metric was evaluated over look-back periods of 2, 5, 10 or all available years since 1985, to investigate variability of the time period over which a fire has significant effects on streamflow, which may be transient. Model results were not improved through inclusion of high-severity fire in models, consistent with the strong correlation between high-severity and total fire (r = 0.88–0.93, depending on the look-back period). Furthermore, total percent burned area data (A_b) are generally more widely available than high-severity burned area and are available here for more years. Therefore, total percent burned area (A_b) was used as the fire metric for multiple linear regression (Equation 4). Models were more likely to show significant fire effects when calculated as a percentage of the entire watershed rather than just the forested portion.

$urn:x-wiley:00431397:media:wrcr25851:wrcr25851-math-0004$ (4)

Air temperature was expected to negatively impact streamflow due to its exponential relationship to vapor pressure deficit, itself a primary control on evaporative losses (Shuttleworth, 2012). However, seasonal or annual air temperature did not improve model performance (Figures S3–S5 in Supporting Information S1). While annual and seasonal temperatures are rising broadly across the study region and were therefore higher post-fire than pre-fire, it is expected that the lack of predictive power of temperature for streamflow here resulted from negative correlation between temperature and precipitation for winter, summer, and annually, making precipitation alone a sufficient predictor for climatic differences.

4.5 Time-Trend Analysis

Time-trend analysis may provide a more robust assessment of ecosystem response to a threshold event, such as a single, large fire (Biederman et al., 2015; Bosch & Hewlett, 1982; Guardiola-Claramonte et al., 2011; Zhao et al., 2010). First, empirical models of streamflow and climate (Equations 1-3) are created for each watershed with a period of pre-fire years reserved for independent model evaluation. Next, the models are used to predict streamflow for both pre-fire evaluation and post-fire periods, and fire effects are indicated by changes in the structure of residuals between the pre-fire and post-fire periods. As in the double-mass analysis (Section 4.1), a watershed was classified as “post-fire” beginning the year after total fire area exceeded 20% of the watershed area or, alternatively, following the year of the relevant landscape-scale fire(s). Time-trend analysis was exclusively performed on years and seasons classified as “wet” (Section 4.2).

Each model was calibrated to the pre-fire period from the beginning of streamflow records (Table 1) until 5 years before fire in each watershed. Following calibration, residual histograms and a Lilliefors test of normality (Lilliefors, 1967) confirmed normal distribution of residuals. The 5-year period between calibration and fire was reserved as an evaluation period to test the empirical model's ability to provide accurate pre-fire streamflow predictions. A one-sided t-test (p < 0.05) was used to evaluate the expectation that the pre-fire mean residual would not be different from zero. Similarly, a t-test was used to detect non-zero post-fire residuals, indicating fire effects on annual or seasonal streamflow. When significant, the mean post-fire residual directly estimates the long-term annual fire effect on streamflow.

4.6 Dry Years and Seasons Analysis

Dry years and seasons (see Section 4.2) do not show a relationship of Q with P and should not be evaluated using multiple linear regression or time-trend analysis. Instead, we compared mean streamflow during dry years and seasons before and after fire using the Mann–Whitney U-test.

5 Results

5.1 Double-Mass Paired Watershed Comparison

Following the 2002 Rodeo-Chediski fire, a greater double-mass slope suggests summer flows increased in the extensively burned CAS watershed (percent burned area A_b = 63%) relative to the control CHG^C (Figure 5), although the slope returned to the pre-fire value from the sixth summer (e.g., 2008) onward, suggesting hydrologic recovery. However, reduced flows in winter drove net reductions in annual streamflow (Figures 5a and 5b), because winter flows are ∼4 times larger than summer flows (Table 1). In the less-extensively burned CIC (A_b = 33%), summer flows increased without detectable impact on winter or annual streamflow (Figures 5d–5f). Following the 2011 Wallow Fire, a significant slope increase indicates that summer flows increased in the extensively burned BLP watershed (A_b = 71%) relative to the control EFW^C, but that winter and annual flows remained unchanged (Figures 5h–5j). At a larger spatial scale, comparison of the Black River (A_b = 32%) and White River (control) at their confluence (WHF^C vs. BLF) also demonstrated increased summer streamflow from the burned watershed, but to a lesser extent than the smaller sub-watershed (BLP), consistent with dilution of impacts at increasing spatial scale.

5.2 Climate and Streamflow Comparisons

The annual air temperature (T) variability was greater across watersheds (∼6°C) than among years (∼3°C, Figure S2 in Supporting Information S1). In contrast, annual P was more variable over time (∼500 mm) than among watersheds, where it averaged ∼150 mm except at the high-elevation EFW^C, which typically received at least 150 mm greater P than any other watershed. Both specific Q and RR varied temporally with P in all watersheds, and P was significantly correlated with both Q (r = 0.83) and RR (r = 0.70). Air temperature was negatively correlated with Q (r = −0.43) and RR (r = −0.44), and a similar negative relationship between P and T (r = −0.35) suggests that wetter years are likely to be colder and vice versa.

Air temperature was higher and precipitation was lower in the years following wildfire than before (Table 2), consistent with prevailing hot drought during the 21st century in the US Southwest (Udall & Overpeck, 2017). Accordingly, annual and winter Q declined in all watersheds, both burned and control. The annual RR was significantly less following the 2002 Rodeo-Chediski fire in one burned watershed (CIC), and following the 2011 Wallow fire in two control watersheds (EFW^C and WHF^C), but no burned watersheds (Table 2). At the basin scale, post-fire RR decreased in both the SAC and SAR watersheds. In the heavily burned CAS watershed, streamflow was above the expected value during each of the first six summers following fire, despite all six summers being dry, with precipitation below the breakpoint (see Section 4.2; Figure 6). No corresponding post-fire changes were found in CAS during the winter or annually (Figures S10 and S11 in Supporting Information S1).

Table 2. Differences Between Mean Pre-Fire and Post-Fire Climate and Streamflow^a

Site	Burn year	Air temp. ΔT_mean (°C)	Annual precip. ΔP (mm)	Annual streamflow ΔQ (mm)	Winter^b ΔQ_w (mm)	Summer ΔQs (mm)	Runoff ratio ΔRR (−)	Estimated fire effect^c (mm)
Mogollon Rim Headwaters – Rodeo-Chediski Fire
CAS	2002	0.9**	−60**	−16*	−16*	0	−0.02	−10
CIC	2002	1.0**	−72**	−21**	−18**	−3*	−0.02**	−12**
Mogollon Rim Headwaters – Control
CHG^C	^d2002	0.9**	−78**	−17*	−12	−4**	−0.02	−10
White Mountains Headwaters – Wallow Fire
BLP	2011	1.3**	−87*	−39	−46	6	−0.03	−15
BLF	2011	1.2**	−84*	−48	−47	−1	−0.05	−26
White Mountains Headwaters – Control
EFW^C	^d2011	1.2**	−134**	−56**	−49*	−7	−0.04*	−26*
WHF^C	^d2011	1.1**	−90**	−56**	−50**	−6*	−0.06**	−37**
Salt River Basin Main Stem – Both Fires
SAC	2002	1.0**	−67**	−28*	−27*	−1.6	−0.03	−17
SAC	2011	1.2**	−67*	−37**	−35	−2	−0.04*	−23*
SAR	2002	1.0**	−68**	−24*	−22	−2.2	−0.02	−10
SAR	2011	1.2**	−65*	−33*	−31	−3	−0.04*	−21*

Note. Pre-fire and post-fire period observations significantly different at **p < 0.05 or *p < 0.10 using the nonparametric Mann–Whitney U-test. Significant results are also bold. Burn threshold years 2002, the year of the Rodeo-Chediski Fire, and 2011, the year of the Wallow Fire, were each evaluated for mainstem gauges SAC and SAR affected by both fires.
^a Post-fire mean minus pre-fire mean (from 1955 to the last year prior to fire, see Table 1).
^b ΔQ_w = winter period streamflow, November-June, ΔQ_s = summer, July–October. All other values are annual.
^c The product of post-fire mean P with change in RR.
^d To facilitate comparisons, artificial burn years were imposed for control watersheds (CHG^C = 2002; EFW^C = 2011; WHF^C = 2011).

5.3 Regression Modeling

Multiple linear regression (MLR; Equations 1-4) for the wet years detected very few streamflow impacts of fire (Table S4 in Supporting Information S1). Adding annually updated burned area to the climate-only streamflow model improved model performance only for summer and only in the extensively burned BLP watershed (A_b = 71% in the 2011 Wallow Fire); meanwhile, MLR analysis detected no significant effects of fire on the much larger annual and winter flows (not shown). These mostly null fire impacts detected by regression modeling were relatively insensitive to fire look-back periods ranging from 1 year to all years since 1985 (Figure S6 in Supporting Information S1). A 2-year look-back period resulted in the best model improvement (Equation 4 relative to Equation 3) based on the coefficients of determination, mean absolute error, and detectability of fire impacts in a greater number of watersheds (Table S4 in Supporting Information S1).

5.4 Time-Trend Analysis

Time-trend analysis differs from the regression analysis by pooling the residuals of a climate-only regression model and testing for a collective post-fire change in streamflow. In all watersheds, models of streamflow based on seasonal precipitation (Equations 1 and 2) were significant (p < 0.001) predictors of annual or winter streamflow during wet years (Tables S1 and S2 in Supporting Information S1). During the summer (Equation 3 and Table S3 in Supporting Information S1), significant models were obtained for all watersheds except CAS and CHG^C, where the small numbers of wet summers may have precluded significant linear relationships between P with Q (Figure S1 in Supporting Information S1). Otherwise, models described the observations well and accounted for 65%–88% of observed annual Q variability during the calibration period (Figure S7 in Supporting Information S1), 70%–90% of winter Q variability (Figure S8 in Supporting Information S1), and 82%–90% of summer Q variability (Figure S9 in Supporting Information S1). Model skill, assessed by mean absolute error (MAE, range 11–28 mm yr⁻¹), was similar during the calibration and evaluation periods (Table 3). Post-fire mean residuals of annual streamflow in wet years were significantly less than zero (i.e., streamflow was less than predicted from climate variables) in four watersheds (CAS, CIC, WHF^C, and SAR) and greater than zero (increased annual Q) in one watershed (BLP). Time-trend analysis did not detect any significant fire effects on seasonal flows (results not shown).

Table 3. Time-Trend Linear Regression Model With Fire as a Binary Threshold: Coefficients^a, Skill, Statistics, and Estimated Fire Effects on Annual Streamflow (Q) During Wet Years

	a	b	c	MAE	MAE		Mean residual	Std. Error
Site						R²	Post-fire streamflow change ΔQ
	Intercept	Winter P	Summer P	Calibration	Evaluation
	(mm)	(−)	(−)	(mm)	(mm)		(mm)	(mm)
CAS	−29	0.26	−0.02	21	21	0.65	−15**	2
CIC	−87	0.36	0.12	19	14	0.66	−11**	1
CHG^C	−56	0.28	0.06	23	16	0.68	−5	4
BLP	−119	0.6	0.14	28	18	0.85	12*	2
BLF	−123	0.58	0.16	25	18	0.85	1	2
EFW^C	−45	0.38	0.08	23	13	0.88	−10	2
WHF^C	−76	0.4	0.08	18	18	0.87	−19**	2
SAC⁰²	−79	0.44	0.10	19	8	0.83	−7	3
SAC¹¹	−85	0.44	0.11	18	18	0.85	−11	2
SAR⁰²	−77	0.41	0.09	19	15	0.81	−15**	1
SAR¹¹	−83	0.41	0.11	18	11	0.81	−16*	1

Note. **denotes significant effect at p < 0.05. *denotes significant effect at p < 0.10. Significant results are also bold. 02 and 11 signify the use of 2002, the year of the Rodeo Chediski Fire, and 2011, the year of the Wallow Fire, as burn threshold years for mainstem gauges SAC and SAR affected by both fires.
^a Parameter estimates of Equation 1: Q = a + bP_w + cP_s.

6 Discussion

6.1 Evaluation of Hypotheses

Large wildfires generally increased summer streamflow at all elevations, but responses of winter and annual streamflow differed in adjacent headwaters with differing elevation; higher, colder headwaters showed little change, while lower, warmer headwaters showed decreased winter and annual streamflow (Figure 7). This result supports H1 – Hydrologic partitioning following large wildfire varies with elevation-dependent interactions of vegetation and hydrologic regime. While post-fire summer streamflow increased by 24%–38%, especially in the more heavily burned headwaters of both climatic subregions (CAS and BLP; Figure 7), annual flows were dominated by winter flow (Table 1), which remained unchanged in the higher/colder White Mountains but declined by 47% in a heavily burned headwater of the lower/warmer Mogollon Rim (CAS, A_b = 63%). Different streamflow responses in winter versus summer support H2 – Long-term hydrological sensitivity to fire is moderated by counteracting seasonal responses, especially in the lower, warmer headwaters of the Mogollon Rim that are characteristic of many portions of the LCRB. Time trend analysis showed no significant wildfire impacts on streamflow on the main stem at SAC, which integrates streamflow mostly from the higher, colder White Mountains (Figure 1), whereas annual streamflow for the entire basin at SAR (effectively the streamflow at SAC plus streamflow generated in headwaters of the lower-elevation Mogollon Rim) showed a decline of ∼15 mm/yr following the 2002 Rodeo-Chediski Fire in the Mogollon Rim (Table 3). This basin-scale result reinforces the interpretation that fire did not impact long-term annual streamflow from higher/colder watersheds but reduced streamflow in the lower/warmer forested portions of the Salt River Basin. During dry years, there was little evidence of fire impacts on streamflow, likely because the majority of precipitation was partitioned to ET, regardless of vegetation status (Hibbert, 1979), supporting H3 – Wildfire will only have detectable impacts on streamflow in wet years (above ∼500 mm precipitation) because below 500 mm, annual precipitation and streamflow become decoupled. Accordingly, climatically dry conditions likely obscured streamflow responses in either direction (Table 2). Below, we evaluate what these results imply about biophysical controls on hydrologic response and what we have learned about change detection following two large wildfires in these warm, semiarid watersheds.

6.2 Biophysical Processes Affecting Hydrologic Response to Large Wildfire

Inferences drawn from the contrasting hydrologic responses to these two fires in adjacent headwaters of the White Mountains and Mogollon Rim (Figure 7) can shed light upon how expected hydrologic response to disturbance may differ between colder/wetter and warmer/drier basins.

Snow sublimation in US Southwest forests can vary from 10% to 30% of winter precipitation and is more important in higher, colder watersheds due to longer residence time in the canopy and a longer snow season (Sexstone et al., 2018; Svoma, 2017). Forest disturbance reduces interception but exposes snowpack to sun and wind, increasing snowpack sublimation (Biederman, Gochis, et al., 2014; Goeking & Tarboton, 2020; Musselman et al., 2008). Therefore, peak snowpack may be optimized at intermediate forest density (Veatch et al., 2009). In the Salt River Basin, Broxton et al. (2020) previously showed peak SWE is optimized at 30%–50% forest cover. While adding high-severity fire to regression models did not improve streamflow prediction (Section 4.4), the post-fire decline in winter streamflow is consistent with the idea that widespread, severe fire reduced canopy cover well below 30% in many places (Table 2), increasing net snow sublimation and reducing snowmelt volumes at all elevations. Because increased snow sublimation is expected with canopy reduction in the high-elevation White Mountains (Broxton et al., 2020), the lack of winter streamflow reduction in the White Mountains headwaters suggests a counteracting biophysical response that differs with elevation (H1).

A characteristic feature of the LCRB, typified by the lower-elevation Mogollon Rim headwaters, is the decoupling of energy and water availability because shallower snowpack melts early in the year, well before peak evaporative demand and vegetation activity (Barnhart et al., 2020; Robles et al., 2021). We suggest this may diminish the importance of reduced transpiration to the hydrologic disturbance response in warm, semiarid watersheds such as the Mogollon Rim (Knighton et al., 2020; Robles et al., 2021), explaining the lack of increased streamflow (Figure 7; Adams et al., 2012; Guardiola-Claramonte et al., 2011). Meanwhile, the snowmelt pulse in the higher, colder White Mountains headwaters typically occurs between April and May (Figure 2), close to the summer peak in ET (Dore et al., 2010; Knowles et al., 2020). We therefore propose the following working hypotheses to explain the small increases in streamflow post-fire for high/cold watersheds and decreases for low/warm watersheds: (a) low post-fire forest density (Table 2) resulted in net increased sublimation at all elevations; (b) transpiration reductions were large enough to balance or outweigh increased sublimation in the White Mountains, where snowmelt coincides with the growing season, resulting in increased streamflow; however, the smaller transpiration “savings” in the Mogollon Rim headwaters was not enough to compensate for greater snow sublimation, and winter/spring streamflow declined. We postulate that these mechanisms can help explain the contrast between increased streamflow often reported post-disturbance for wetter, colder basins and decreased streamflow in drier, warmer basins (Bethlahmy, 1974; Biederman et al., 2015; Buma & Livneh, 2017; Goeking & Tarboton, 2020; Guardiola-Claramonte et al., 2011; Zhang & Wei, 2012). Furthermore, lower-elevation headwaters are more often subject to at least some minor human impacts and therefore excluded from studies filtering for USGS reference designation, possibly helping resolve the discrepancy between prior studies attributing increased streamflow to fire in the LCRB (Hallema et al., 2018; Wine & Cadol, 2016) and the declines observed by managers in the Salt River Basin. This study also found increased or unchanged winter and annual streamflow in the highest, least disturbed watershed (BLP) but showed that basin-scale trends were influenced by a different response in lower-elevation headwaters excluded from prior analyses due human impacts that turned out to be quite small (Figure S12 in Supporting Information S1 and Table S5 in Supporting Information S1).

6.3 Increased Summer Streamflow Following Wildfire

Here, we confirm that these two large wildfires increased summer streamflow in headwaters at both high and low elevations in the Salt River Basin, despite a marginal or even negative effect on annual streamflow (Figures 5 and 7). Interestingly, we observed above-average summer streamflow in the heavily burned CAS sub-watershed for six climatically dry summers following the Rodeo-Chediski Fire (Figure 6), despite the typically insensitive P–Q relationship during dry years (Figure 3), consistent with prior studies showing soil glazing and/or decreased surface roughness can overwhelm the streamflow generation signal following large wildfire (Ffolliott et al., 2011). No such changes were measured for winter or annual flows (Figures S10 and S11 in Supporting Information S1). Decreased root water uptake may also factor into this result by enhancing summer low flows (Bearup et al., 2014; Kinoshita & Hogue, 2015). In this study, summer streamflow prediction was improved significantly by addition of the prior P_w (Equation 2), consistent with winter baseflow recharge and seasonal carry-over of stored moisture (Wyatt et al., 2015). Given that winter precipitation and associated baseflow have declined in the LCRB over recent decades (Garner et al., 2013; Zhang et al., 2021) the magnitude and sustainability of increased post-fire peak flow will depend on the degree to which surface versus groundwater processes control summer streamflow generation.

6.4 Methodological Comparison, Alignment, and Limitations

Of the four methods, our strongest inferences derive from the double-mass and time-trend results. Double-mass watershed comparison is robust across seasons, applicable in both wet and dry years (Figure 5), and integrates the effects of non-stationarity including temperature and vegetation. Double-mass analysis does not assume a single form of the P–Q relationship for a given watershed, but rather a linear relationship Q₁–Q₂ between streamflow in treatment and control watersheds (Figure 5) that is robust through a range of climate variability including wet and dry years. Here, declining precipitation, streamflow, and runoff ratio in the control watershed WHF^C during post-fire years (Tables 2 and 3) possibly affected the double-mass comparison with the burned BLF watershed. However, under the assumption that paired watersheds experience similar relative changes in climate, a non-stationary control watershed likely represents the best predictor of streamflow in the burned watershed, had it not burned.

Time-trend analysis is applicable to all watersheds but limited to wet years and subject to assumptions of stationarity (Table 3). Time-trend showed reduced annual streamflow for the two Mogollon Rim watersheds (CAS and CIC; Table 3 and Figure 7) but comparatively little impact in the White Mountains (BLP and BLF; Table 3 and Figure 7). Time-trend results were consistent with the double-mass results (Figures 5 and 7), supporting H2 and highlighting a potentially different response of lower, warmer semiarid watersheds in the Mogollon Rim from more well-studied higher, colder watersheds such as in the White Mountains and many portions of the UCRB. A principal limitation of the time-trend method is the assumption of stationarity in the P–Q relationship, particularly with respect to observed trends of increased air temperature, decreased precipitation, and shifts from snowfall to rainfall (McCabe et al., 2017; Robles et al., 2021; Woodhouse et al., 2016). Overall, CAS was the most difficult watershed to model, because it showed the weakest relationship between P and Q of all study watersheds (Table 3 and Figure S1 in Supporting Information S1); however, the negative effect of fire on annual flow at CAS determined by time-trend analysis was consistent with double-mass analysis (Figure 7). No significant effects of fire on streamflow were detected using the MLR and climatic analyses, which are not recommended for application to seasonally dry catchments due to the potential for non-linearity (Wine, Cadol, & Makhnin, 2018; Wine, Makhnin, & Cadol, 2018) in the P–Q relationship and/or threshold behavior (Figure 4).

6.5 Challenges to Detecting Hydrologic Change in the LCRB

The current results support our third hypothesis and show that in semiarid watersheds with high interannual precipitation variability, it is critical to separate wet and dry years (or seasons) in any type of empirical model utilizing the relationship between precipitation and streamflow (Figure 4). Because the watersheds of the Salt River Basin do not exhibit continuous relationships between streamflow and precipitation in dry years (Section 4.2, Figures 4 and S1 in Supporting Information S1), forcing a single regression through all years, transformed or otherwise, will under-predict streamflow in dry years when P and Q are uncorrelated. For example, a single linear model fit to all years in the CIC sub-watershed (Figure 3a) would under-predict observed streamflow in 10 of the 13 dry years (not shown) and incorrectly imply that fire increased annual streamflow. This is critical in the US Southwest, where the majority of years since 1999 have been drier than the historical average (Udall & Overpeck, 2017).

The runoff ratio remains one of the most frequently used metrics of watershed disturbance detection (e.g., Buma & Livneh, 2017). While we present runoff ratio for comparability with prior studies, we note its critical limitations, especially in semiarid watersheds with wet/dry year threshold behavior (Figure 4). The runoff ratio incorrectly implies passage of a P–Q relationship through the origin (i.e., intercept = 0), whereas watershed measurements usually show that precipitation must exceed a minimum amount of ET (i.e., the horizontal axis intercept of a P–Q plot such as wet years in Figure 4) before streamflow may occur (Biederman et al., 2015; Flerchinger & Cooley, 2000; Stednick, 1996). The present study suggests that semiarid watersheds of the Salt River Basin are all characterized by intrinsic annual precipitation thresholds of ∼400–450 mm (Figures 4 and S1), which is similar to watersheds in the UCRB (Biederman et al., 2015) and to the precipitation threshold of ∼500 mm proposed for disturbance detection (Adams et al., 2012; Bosch & Hewlett, 1982). Semiarid watersheds with high interannual variability may show hydrologic response to disturbance only in wet years above such threshold, and therefore change detection should focus on these years. Additionally, before/after fire comparisons based on the runoff ratio (Table 2) cannot adequately disentangle effects of chronic changes in climate from those of disturbance, and we consider this method relatively weak.

As with any study of long-term streamflow change detection, we expect our results may be influenced by chronic rises in temperature, which drives increased vapor pressure deficit and potentially increases sublimation, evaporation, and transpiration. Among the empirical analysis methods employed here, double-mass analysis is likely the most appropriate for detecting fire impacts in the context of non-stationary climate (see Section 6.1). Estimates of process-level responses attributable to climate change versus forest disturbance are better suited to studies applying mechanistic models or making process measurements (Biederman, Gochis, et al., 2014; Boisramé et al., 2019). We acknowledge that warmer, drier conditions over recent decades could have negated otherwise observable streamflow increases (Hallema et al., 2018; Hallema, Sun, Caldwell, et al., 2017) due to reduced interception and transpiration. However, given widespread predictions of ongoing climate change, it is valuable to know that the net effect of climatic and biophysical changes did not increase streamflow in the first two decades following these landscape-scale fires. Similarly, several of the large, gauged watersheds in this study have small, ungauged diversions for irrigation of between 1,500 and 4000 acres (6 and 16 km²) of agricultural fields. While the best available information suggests these withdrawals have negligible impact on our analysis (Supporting Information for Data Section 3.1 Streamflow Data: Potential Impacts of Human Water Use, Figure S12 in Supporting Information S1 and Table S5 in Supporting Information S1), it is possible that unknown increases in usage of these small irrigation withdrawals could at least partially obscure increased streamflow due to fire.

A final challenge to note is that despite the landscape scale and relatively high intensity of the Rodeo-Chediski and Wallow Fires (Figure 1 and Table 1), the total watershed area burned at high intensity remained below 30%, a threshold of basal area reduction previously identified for mechanical thinning to produce streamflow increases in the nearby Beaver Creek watershed (Baker, 1986).

7 Conclusions

Southwestern US wildfires are often associated with images of catastrophic floods in the initial high-intensity summer monsoon storms following disturbance. Although the current results support increased post-fire summer streamflow, counteracting negative effects of fire on the much-larger winter streamflow dominated the signal in net annual water resources, which declined. Importantly, the elevation gradient of watersheds studied here revealed contrasting impacts of fire on winter streamflow: higher/colder watersheds showed minimal changes, similar to results seen in cold/wet regions and usually attributed to offsetting changes in interception and transpiration, whereas the lower/warmer watersheds showed decreased winter streamflow. We hypothesize this different response to disturbance in lower, warmer watersheds could be due to the shorter winter season minimizing the importance of interception changes and because the main snowmelt pulse occurs well before the growing season, reducing the impacts of vegetation change on streamflow. Further process-based field studies, observations of other disturbances, and modeling are needed to test this hypothesis.

Acknowledgments

The authors thank four reviewers and two editors for their investment in earlier versions of this manuscript. USDA is an equal-opportunity employer. The views expressed herein are those of the authors and do not necessarily reflect USDA views or policies.

Open Research

Data Availability Statement

All of the data sets used herein were downloaded from publicly available sources as cited in Section 3. The authors wish to thank those individuals and organizations who collected and made these data available. MATLAB codes for statistics and figure generation are available upon request from the corresponding author.

Supporting Information

References

Abatzoglou, J. T., & Williams, A. P. (2016). Impact of anthropogenic climate change on wildfire across western US forests. Proceedings of the National Academy of Sciences of the United States of America, 113, 201607171. https://doi.org/10.1073/pnas.1607171113
10.1073/pnas.1607171113
Web of Science®Google Scholar
Adams, H. D., Barron-Gafford, G. A., Minor, R. L., Gardea, A. A., Bentley, L. P., Law, D. J., et al. (2017). Temperature response surfaces for mortality risk of tree species with future drought. Environmental Research Letters, 12(11), 115014. https://doi.org/10.1088/1748-9326/aa93be
10.1088/1748-9326/aa93be
ADSWeb of Science®Google Scholar
Adams, H. D., Luce, C. H., Breshears, D. D., Allen, C. D., Weiler, M., Hale, V. C., et al. (2012). Ecohydrological consequences of drought- and infestation- triggered tree die-off: Insights and hypotheses. Ecohydrology, 5(2), 145–159. https://doi.org/10.1002/eco.233
10.1002/eco.233
PubMedWeb of Science®Google Scholar
Baker, M. B. (1986). Effects of ponderosa pine treatments on water yield in Arizona. Water Resources Research, 22(1), 67–73. https://doi.org/10.1029/wr022i001p00067
10.1029/wr022i001p00067
ADSWeb of Science®Google Scholar
Balch, J. K., Bradley, B. A., Abatzoglou, J. T., Nagy, R. C., Fusco, E. J., & Mahood, A. L. (2017). Human-started wildfires expand the fire niche across the United States. Proceedings of the National Academy of Sciences of the United States of America, 114(11), 2946–2951. https://doi.org/10.1073/pnas.1617394114
10.1073/pnas.1617394114
CASADSPubMedWeb of Science®Google Scholar
Barnett, T., Malone, R., Pennell, W., Stammer, D., Semtner, B., & Washington, W. (2004). The effects of climate change on water resources in the west: Introduction and overview. Climatic Change, 62(1), 1–11. https://doi.org/10.1023/B:CLIM.0000013695.21726.b8
10.1023/B:CLIM.0000013695.21726.b8
Web of Science®Google Scholar
Barnhart, T. B., Tague, C. L., & Molotch, N. P. (2020). The counteracting effects of snowmelt rate and timing on runoff. Water Resources Research, 56(8), e2019WR026634. https://doi.org/10.1029/2019WR026634
10.1029/2019WR026634
ADSWeb of Science®Google Scholar
Bart, R. R. (2016). A regional estimate of postfire streamflow change in California. Water Resources Research, 52(2), 1465–1478. https://doi.org/10.1002/2014WR016553
10.1002/2014WR016553
ADSWeb of Science®Google Scholar
Bearup, L. A., Maxwell, R. M., Clow, D. W., & McCray, J. E. (2014). Hydrological effects of forest transpiration loss in bark beetle-impacted watersheds. Nature Climate Change, 4(6), 481–486. https://doi.org/10.1038/nclimate2198
10.1038/nclimate2198
ADSWeb of Science®Google Scholar
Benavides-Solorio, J., & MacDonald, L. H. (2001). Post-fire runoff and erosion from simulated rainfall on small plots, Colorado Front Range. Hydrological Processes, 15(15), 2931–2952. https://doi.org/10.1002/hyp.383
10.1002/hyp.383
ADSWeb of Science®Google Scholar
Bethlahmy, N. (1974). More streamflow after a bark-beetle epidemic. Journal of Hydrology, 23(3–4), 185–189. https://doi.org/10.1016/0022-1694(74)90001-8
10.1016/0022-1694(74)90001-8
ADSWeb of Science®Google Scholar
Biederman, J. A., Brooks, P. D., Harpold, A. A., Gochis, D. J., Gutmann, E., Reed, D. E., et al. (2014). Multiscale observations of snow accumulation and peak snowpack following widespread, insect-induced lodgepole pine mortality. Ecohydrology, 7(1), 150–162. https://doi.org/10.1002/eco.1342
10.1002/eco.1342
Web of Science®Google Scholar
Biederman, J. A., Gochis, D. J., Reed, D. E., Papuga, S. A., Brooks, P. D., Harpold, A. A., & Brooks, P. D. (2014). Increased evaporation following widespread tree mortality limits streamflow response. Water Resources Research, 50(7), 5395–5409. https://doi.org/10.1002/2013WR014994
10.1002/2013WR014994
ADSWeb of Science®Google Scholar
Biederman, J. A., Scott, R. L., Bell, T. W., Bowling, D. R., Dore, S., Garatuza-Payan, J., et al. (2017). CO₂ exchange and evapotranspiration across dryland ecosystems of southwestern North America. Global Change Biology, 23(10), 4204–4221. https://doi.org/10.1111/gcb.13686
10.1111/gcb.13686
ADSPubMedWeb of Science®Google Scholar
Biederman, J. A., Somor, A., Harpold, A. A., Gutmann, E. D., Breshears, D. D., Troch, P. A., et al. (2015). Recent tree die-off has little effect on streamflow in contrast to expected increases from historical studies. Water Resources Research, 51(12), 9775–9789. https://doi.org/10.1002/2015WR017401
10.1002/2015WR017401
ADSWeb of Science®Google Scholar
Boisramé, G. F. S., Thompson, S. E., Collins, B., & Stephens, S. L. (2017). Managed wildfire effects on forest resilience and water in the Sierra Nevada. Ecosystems, 20(4), 717–732. https://doi.org/10.1007/s10021-016-0048-1
10.1007/s10021-016-0048-1
Web of Science®Google Scholar
Boisramé, G. F. S., Thompson, S. E., Tague, C. (N.), & Stephens, S. L. (2019). Restoring a natural fire regime alters the water balance of a Sierra Nevada catchment. Water Resources Research, 55(7), 5751–5769. https://doi.org/10.1029/2018WR024098
10.1029/2018WR024098
ADSWeb of Science®Google Scholar
Bosch, J. M., & Hewlett, J. D. (1982). A review of catchment experiments to determine the effect of vegetation changes on water yield and evapo-transpiration. Journal of Hydrology, 55(1–4), 3–23. https://doi.org/10.1016/0022-1694(82)90117-2
10.1016/0022-1694(82)90117-2
ADSWeb of Science®Google Scholar
Brown, A. E., Zhang, L., McMahon, T. A., Western, A. W., & Vertessy, R. A. (2005). A review of paired catchment studies for determining changes in water yield resulting from alterations in vegetation. Journal of Hydrology, 310(1–4), 28–61. https://doi.org/10.1016/j.jhydrol.2004.12.010
10.1016/j.jhydrol.2004.12.010
ADSWeb of Science®Google Scholar
Brown, H., Baker, M. B., Jr, Rogers, J. J., Clary, W. P., Kovner, J. L., Larson, F. R., et al. (1974). Opportunities for increasing water yield and other multiple use values on ponderosa pine forest lands (USDA Forest Service Research Paper RM-129, p. 36). U.S. Department of Agriculture, Forest Service, Rocky Mountain Forest and Range Experiment Station.
Google Scholar
Brown, T. C., Hobbins, M. T., & Ramirez, J. A. (2008). Spatial distribution of water supply in the Coterminous United States1. Journal of the American Water Resources Association, 44(6), 1474–1487. https://doi.org/10.1111/j.1752-1688.2008.00252.x
10.1111/j.1752-1688.2008.00252.x
ADSWeb of Science®Google Scholar
Broxton, P. D., Harpold, A. A., Biederman, J. A., Troch, P. A., Molotch, N. P., & Brooks, P. D. (2015). Quantifying the effects of vegetation structure on snow accumulation and ablation in mixed-conifer forests. Ecohydrology, 8(6), 1073–1094. https://doi.org/10.1002/eco.1565
10.1002/eco.1565
Web of Science®Google Scholar
Broxton, P. D., van Leeuwen, W. J. D., & Biederman, J. A. (2020). Forest cover and topography regulate the thin, ephemeral snowpacks of the semiarid Southwest United States. Ecohydrology, 13(4), e2202. https://doi.org/10.1002/eco.2202
10.1002/eco.2202
Web of Science®Google Scholar
Buma, B., & Livneh, B. (2017). Key landscape and biotic indicators of watersheds sensitivity to forest disturbance identified using remote sensing and historical hydrography data. Environmental Research Letters, 12(7), 074028. https://doi.org/10.1088/1748-9326/aa7091
10.1088/1748-9326/aa7091
Web of Science®Google Scholar
Campbell, R. E., Baker, M. B., Ffolliott, P. F., Larson, F. R., & Avery, C. (1977). Wildfire effects on a ponderosa pine ecosystem: An Arizona case study(Research Paper RM-191). USDA Forest Service Rocky Mountain Forest and Range Experiment Station. Retrieved from http://www.fs.fed.us/rm/pubs_rm/rm_rp191.pdf
Google Scholar
DeBano, L. F., Ffolllot, P. F., & Baker, M. B. (1996). Fire severity effects on water resources. Effects of Fire on Madrean Province Ecosystems, 77–84.
Google Scholar
Dennison, P. E., Brewer, S. C., Arnold, J. D., & Moritz, M. A. (2014). Large wildfire trends in the western United States, 1984–2011. Geophysical Research Letters, 41(8), 2014GL059576. https://doi.org/10.1002/2014gl059576
10.1002/2014gl059576
Web of Science®Google Scholar
Dore, S., Kolb, T. E., Montes-Helu, M., Eckert, S. E., Sullivan, B. W., Hungate, B. A., et al. (2010). Carbon and water fluxes from ponderosa pine forests disturbed by wildfire and thinning. Ecological Applications, 20(3), 663–683. https://doi.org/10.1890/09-0934.1
10.1890/09-0934.1
CASPubMedWeb of Science®Google Scholar
Driscoll, D. G., & Carter, J. M. (2004). Hydrologic effects of the 1988 Galena Fire, Black Hills area, South Dakota (Water-Resources Investigations Report 03-4323, p. 73). U.S. Department of Interior, U.S. Geological Survey.
Google Scholar
Ellis, A. W., & Sauter, K. (2017). The significance of snow to surface water supply: An empirical case study from the southwestern United States. Physical Geography, 38(3), 211–230. https://doi.org/10.1080/02723646.2017.1281014
10.1080/02723646.2017.1281014
Web of Science®Google Scholar
Fazel, N., Torabi Haghighi, A., & Kløve, B. (2017). Analysis of land use and climate change impacts by comparing river flow records for headwaters and lowland reaches. Global and Planetary Change, 158, 47–56. https://doi.org/10.1016/j.gloplacha.2017.09.014
10.1016/j.gloplacha.2017.09.014
ADSWeb of Science®Google Scholar
Ffolliott, P. F., Stropki, C. L., Chen, H., & Neary, D. G. (2011). The 2002 Rodeo-Chediski wildfire’s impacts on southwestern ponderosa pine ecosytems, hydrology, and fuels (RMRS-RP-85, p. 36). U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station.
Google Scholar
Flerchinger, G. N., & Cooley, K. R. (2000). A ten-year water balance of a mountainous semi-arid watershed. Journal of Hydrology, 237(1–2), 86–99. https://doi.org/10.1016/s0022-1694(00)00299-7
10.1016/S0022-1694(00)00299-7
ADSWeb of Science®Google Scholar
Frank, J. M., Massman, W. J., Ewers, B. E., Huckaby, L. S., & Negrón, J. F. (2014). Ecosystem CO2/H2O fluxes are explained by hydraulically limited gas exchange during tree mortality from spruce bark beetles. Journal of Geophysical Research: Biogeosciences, 119, 1195–1215. https://doi.org/10.1002/2013JG002597
10.1002/2013JG002597
CASADSWeb of Science®Google Scholar
Garner, B. D., Pool, D. R., Tillman, F. D., & Forbes, B. D. (2013). Human effects on the hydrologic system of the Verde Valley, Central Arizona, 1910-2005 and 2005-2110, using a regional groundwater flow model. Retrieved from https://pubs.usgs.gov/sir/2013/5029/
10.3133/sir20135029
Google Scholar
Goeking, S. A., & Tarboton, D. G. (2020). Forests and water yield: A synthesis of disturbance effects on streamflow and snowpack in western coniferous forests. Journal of Forestry, 118(2), 172–192. https://doi.org/10.1093/jofore/fvz069
10.1093/jofore/fvz069
Web of Science®Google Scholar
Golding, D. L., & Swanson, R. H. (1986). Snow distribution patterns in clearings and adjacent forest. Water Resources Research, 22(13), 1931–1940. https://doi.org/10.1029/wr022i013p01931
10.1029/WR022i013p01931
ADSWeb of Science®Google Scholar
Guardiola-Claramonte, M., Troch, P. A., Breshears, D. D., Huxman, T. E., Switanek, M. B., Durcik, M., & Cobb, N. S. (2011). Decreased streamflow in semi-arid basins following drought-induced tree die-off: A counter-intuitive and indirect climate impact on hydrology. Journal of Hydrology, 406(3–4), 225–233. https://doi.org/10.1016/j.jhydrol.2011.06.017
10.1016/j.jhydrol.2011.06.017
ADSWeb of Science®Google Scholar
Hallema, D. W., Sun, G., Bladon, K. D., Norman, S. P., Caldwell, P. V., Liu, Y., & McNulty, S. G. (2017). Regional patterns of postwildfire streamflow response in the Western United States: The importance of scale-specific connectivity. Hydrological Processes, 31(14), 2582–2598. https://doi.org/10.1002/hyp.11208
10.1002/hyp.11208
ADSWeb of Science®Google Scholar
Hallema, D. W., Sun, G., Caldwell, P. V., Norman, S. P., Cohen, E. C., Liu, Y., et al. (2017). Assessment of wildland fire impacts on watershed annual water yield: Analytical framework and case studies in the United States. Ecohydrology, 10(2). https://doi.org/10.1002/eco.1794
10.1002/eco.1794
PubMedWeb of Science®Google Scholar
Hallema, D. W., Sun, G., Caldwell, P. V., Norman, S. P., Cohen, E. C., Liu, Y., et al. (2018). Burned forests impact water supplies. Nature Communications, 9(1), 1307. https://doi.org/10.1038/s41467-018-03735-6
10.1038/s41467-018-03735-6
ADSPubMedWeb of Science®Google Scholar
Harpold, A. A., Biederman, J. A., Condon, K., Merino, M., Korgaonkar, Y., Nan, T., et al. (2014). Changes in snow accumulation and ablation following the Las Conchas Forest Fire, New Mexico, USA. Ecohydrology, 7(2), 440–452. https://doi.org/10.1002/eco.1363
10.1002/eco.1363
Web of Science®Google Scholar
Helvey, J. D. (1980). Effects of a North Central Washington wildfire on runoff and sediment production. Journal of the American Water Resources Association, 16(4), 627–634. https://doi.org/10.1111/j.1752-1688.1980.tb02441.x
10.1111/j.1752-1688.1980.tb02441.x
ADSGoogle Scholar
Hibbert, A. R. (1979). Managing vegetation to increase flow in the Colorado River Basin. Rocky Mountain Forest and Range Experiment Station, Forest Service, U.S. Department of Agriculture.
Google Scholar
Kinoshita, A. M., & Hogue, T. S. (2015). Increased dry season water yield in burned watersheds in Southern California. Environmental Research Letters, 10(1), 014003. https://doi.org/10.1088/1748-9326/10/1/014003
10.1088/1748-9326/10/1/014003
Web of Science®Google Scholar
Knighton, J., Vijay, V., & Palmer, M. (2020). Alignment of tree phenology and climate seasonality influences the runoff response to forest cover loss. Environmental Research Letters, 15(10), 104051. https://doi.org/10.1088/1748-9326/abaad9
10.1088/1748-9326/abaad9
ADSWeb of Science®Google Scholar
Knowles, J. F., Scott, R. L., Biederman, J. A., Blanken, P. D., Burns, S. P., Dore, S., et al. (2020). Montane forest productivity across a semiarid climatic gradient. Global Change Biology, 26(12), 6945–6958. https://doi.org/10.1111/gcb.15335
10.1111/gcb.15335
ADSPubMedWeb of Science®Google Scholar
Krogh, S. A., Broxton, P. D., Manley, P. N., & Harpold, A. A. (2020). Using process based snow modeling and Lidar to predict the effects of forest thinning on the Northern Sierra Nevada snowpack. Frontiers in Forests and Global Change, 3. https://doi.org/10.3389/ffgc.2020.00021
10.3389/ffgc.2020.00021
PubMedGoogle Scholar
Leaf, C. F. (1975). Watershed management in the Rocky Mountain subalpine zone: The status of our knowledge. US Dep. Agric. For. Serv. Res. Pap., RM-137, 31. Retrieved from http://digitalcommons.usu.edu/aspen_bib/5098
Google Scholar
Lilliefors, H. W. (1967). On the Kolmogorov-Smirnov test for normality with mean and variance unknown. Journal of the American Statistical Association, 62(318), 399–402. https://doi.org/10.1080/01621459.1967.10482916
10.1080/01621459.1967.10482916
Web of Science®Google Scholar
Livneh, B., Deems, J. S., Buma, B., Barsugli, J. J., Schneider, D., Molotch, N. P., et al. (2015). Catchment response to bark beetle outbreak and dust-on-snow in the Colorado Rocky Mountains. Journal of Hydrology, 523, 196–210. https://doi.org/10.1016/j.jhydrol.2015.01.039
10.1016/j.jhydrol.2015.01.039
ADSWeb of Science®Google Scholar
Love, L. D. (1955). The effect on stream flow of the killing of spruce and pine by the Engelmann spruce beetle. Eos, Transactions American Geophysical Union, 36(1), 113–118. https://doi.org/10.1029/TR036i001p00113
10.1029/TR036i001p00113
ADSGoogle Scholar
Mann, H. B., & Whitney, D. R. (1947). On a test of whether one of two random variables is stochastically larger than the other. The Annals of Mathematical Statistics, 18(1), 50–60. https://doi.org/10.1214/aoms/1177730491
10.1214/aoms/1177730491
Web of Science®Google Scholar
Marlon, J. R., Bartlein, P. J., Gavin, D. G., Long, C. J., Anderson, R. S., Briles, C. E., et al. (2012). Long-term perspective on wildfires in the western USA. Proceedings of the National Academy of Sciences of the United States of America, 109(9), E535–E543. https://doi.org/10.1073/pnas.1112839109
10.1073/pnas.1112839109
CASADSPubMedWeb of Science®Google Scholar
McCabe, G. J., Wolock, D. M., Pederson, G. T., Woodhouse, C. A., & McAfee, S. (2017). Evidence that recent warming is reducing upper Colorado river flows. Earth Interactions, 21(10), 1–14. https://doi.org/10.1175/EI-D-17-0007.1
10.1175/EI-D-17-0007.1
Web of Science®Google Scholar
Milly, P. C. D., & Dunne, K. A. (2020). Colorado River flow dwindles as warming-driven loss of reflective snow energizes evaporation. Science, 367(6483), 1252–1255. https://doi.org/10.1126/science.aay9187
10.1126/science.aay9187
CASADSPubMedWeb of Science®Google Scholar
Moeser, C. D., Broxton, P. D., Harpold, A., & Robertson, A. (2020). Estimating the effects of forest structure changes from wildfire on snow water resources under varying meteorological conditions. Water Resources Research, 56(11), e2020WR027071. https://doi.org/10.1029/2020WR027071
10.1029/2020WR027071
ADSWeb of Science®Google Scholar
Molotch, N. P., Fassnacht, S. R., Bales, R. C., & Helfrich, S. R. (2004). Estimating the distribution of snow water equivalent and snow extent beneath cloud cover in the Salt–Verde River basin, Arizona. Hydrological Processes, 18(9), 1595–1611. https://doi.org/10.1002/hyp.1408
10.1002/hyp.1408
ADSWeb of Science®Google Scholar
Musselman, K. N., Molotch, N. P., & Brooks, P. D. (2008). Effects of vegetation on snow accumulation and ablation in a mid-latitude sub-alpine forest. Hydrological Processes, 22(15), 2767–2776. https://doi.org/10.1002/hyp.7050
10.1002/hyp.7050
ADSWeb of Science®Google Scholar
Neary, D. G., Koestner, K. A., & Youberg, A. (2011). Hydrologic impacts of high severity wildfire: Learning from the past and preparing for the future.
Google Scholar
O’Donnell, F. C., Flatley, W. T., Springer, A. E., & Fulé, P. Z. (2018). Forest restoration as a strategy to mitigate climate impacts on wildfire, vegetation, and water in semiarid forests. Ecological Applications, 28(6), 1459–1472. https://doi.org/10.1002/eap.1746
10.1002/eap.1746
PubMedWeb of Science®Google Scholar
Potts, D. F. (1984). Hydrologic impacts of a large-scale mountain pine beetle (Dendroctonus Ponderosae Hopkins) epidemic. Water Resources Bulletin, 20(3), 373–377. https://doi.org/10.1111/j.1752-1688.1984.tb04719.x
10.1111/j.1752-1688.1984.tb04719.x
ADSWeb of Science®Google Scholar
Ren, J., Adam, J. C., Hicke, J. A., Hanan, E. J., Tague, C. L., Liu, M., et al. (2021). How does water yield respond to mountain pine beetle infestation in a semiarid forest? Hydrology and Earth System Sciences, 25(9), 4681–4699. https://doi.org/10.5194/hess-25-4681-2021
10.5194/hess-25-4681-2021
ADSWeb of Science®Google Scholar
Robinne, F.-N., Hallema, D. W., Bladon, K. D., Flannigan, M. D., Boisramé, G., Bréthaut, C. M., et al. (2021). Scientists’ warning on extreme wildfire risks to water supply. Hydrological Processes, 35(5), e14086. https://doi.org/10.1002/hyp.14086
10.1002/hyp.14086
PubMedWeb of Science®Google Scholar
Robles, M. D., Hammond, J. C., Kampf, S. K., Biederman, J. A., & Demaria, E. M. C. (2021). Winter inputs buffer streamflow sensitivity to snowpack losses in the Salt River watershed in the Lower Colorado River Basin. Water, 13(1), 3. https://doi.org/10.3390/w13010003
10.3390/w13010003
Web of Science®Google Scholar
Robles, M. D., Marshall, R. M., O’Donnell, F., Smith, E. B., Haney, J. A., & Gori, D. F. (2014). Effects of climate variability and accelerated forest thinning on watershed-scale runoff in southwestern USA Ponderosa Pine Forests. PLOS One, 9(10), e111092. https://doi.org/10.1371/journal.pone.0111092
10.1371/journal.pone.0111092
ADSPubMedWeb of Science®Google Scholar
Robles, M. D., Turner, D. S., & Haney, J. A. (2017). A century of changing flows: Forest management changed flow magnitudes and warming advanced the timing of flow in a southwestern US river. PLOS One, 12(11), e0187875. https://doi.org/10.1371/journal.pone.0187875
10.1371/journal.pone.0187875
PubMedWeb of Science®Google Scholar
Searcy, J. K., & Hardison, C. H. (1960). Double-mass curves (No. 1541-B). Geological Survey. Retrieved from http://udspace.udel.edu/handle/19716/1592
Google Scholar
Sexstone, G. A., Clow, D. W., Fassnacht, S. R., Liston, G. E., Hiemstra, C. A., Knowles, J. F., & Penn, C. A. (2018). Snow sublimation in mountain environments and its sensitivity to forest disturbance and climate warming. Water Resources Research, 54(2), 1191–1211. https://doi.org/10.1002/2017WR021172
10.1002/2017WR021172
ADSWeb of Science®Google Scholar
Shuttleworth, W. J. (2012). Terrestrial hydrometeorology. John Wiley & Sons.
10.1002/9781119951933
Google Scholar
Singleton, M. P., Thode, A. E., Sánchez Meador, A. J., & Iniguez, J. M. (2019). Increasing trends in high-severity fire in the southwestern USA from 1984 to 2015. Forest Ecology and Management, 433, 709–719. https://doi.org/10.1016/j.foreco.2018.11.039
10.1016/j.foreco.2018.11.039
Web of Science®Google Scholar
Stednick, J. D. (1996). Monitoring the effects of timber harvest on annual water yield. Journal of Hydrology, 176(1–4), 79–95. https://doi.org/10.1016/0022-1694(95)02780-7
10.1016/0022-1694(95)02780-7
ADSWeb of Science®Google Scholar
Svoma, B. M. (2011). Trends in snow level elevation in the mountains of central Arizona. International Journal of Climatology, 31(1), 87–94. https://doi.org/10.1002/joc.2062
10.1002/joc.2062
ADSWeb of Science®Google Scholar
Svoma, B. M. (2017). Canopy effects on snow sublimation from a central Arizona Basin. Journal of Geophysical Research: Atmospheres, 122(1), 20–46. https://doi.org/10.1002/2016JD025184
10.1002/2016JD025184
ADSWeb of Science®Google Scholar
Teuling, A. J., de Badts, E. A. G., Jansen, F. A., Fuchs, R., Buitink, J., Hoek van Dijke, A. J., & Sterling, S. M. (2019). Climate change, reforestation/afforestation, and urbanization impacts on evapotranspiration and streamflow in Europe. Hydrology and Earth System Sciences, 23(9), 3631–3652. https://doi.org/10.5194/hess-23-3631-2019
10.5194/hess-23-3631-2019
ADSWeb of Science®Google Scholar
Udall, B., & Overpeck, J. (2017). The twenty-first century Colorado River hot drought and implications for the future. Water Resources Research, 53(3), 2404–2418. https://doi.org/10.1002/2016WR019638
10.1002/2016WR019638
ADSWeb of Science®Google Scholar
Veatch, W. C., Brooks, P. D., Gustafson, J. R., & Molotch, N. P. (2009). Quantifying the effects of forest canopy cover on net snow accumulation at a continental, mid-latitude site. Ecohydrology, 2(2), 115–128. https://doi.org/10.1002/eco.45
10.1002/eco.45
Web of Science®Google Scholar
Viviroli, D., Dürr, H. H., Messerli, B., Meybeck, M., & Weingartner, R. (2007). Mountains of the world, water towers for humanity: Typology, mapping, and global significance. Water Resources Research, 43(7). https://doi.org/10.1029/2006WR005653
10.1029/2006WR005653
ADSPubMedWeb of Science®Google Scholar
Westerling, A. L., Hidalgo, H. G., Cayan, D. R., & Swetnam, T. W. (2006). Warming and earlier spring increase western U.S. forest wildfire activity. Science, 313(5789), 940–943. https://doi.org/10.1126/science.1128834
10.1126/science.1128834
CASADSPubMedWeb of Science®Google Scholar
Wine, M. L., & Cadol, D. (2016). Hydrologic effects of large southwestern USA wildfires significantly increase regional water supply: Fact or fiction? Environmental Research Letters, 11(8), 085006. https://doi.org/10.1088/1748-9326/11/8/085006
10.1088/1748-9326/11/8/085006
Web of Science®Google Scholar
Wine, M. L., Cadol, D., & Makhnin, O. (2018). In ecoregions across western USA streamflow increases during post-wildfire recovery. Environmental Research Letters, 13(1), 014010. https://doi.org/10.1088/1748-9326/aa9c5a
10.1088/1748-9326/aa9c5a
Web of Science®Google Scholar
Wine, M. L., Makhnin, O., & Cadol, D. (2018). Nonlinear long-term large watershed hydrologic response to wildfire and climatic dynamics locally increases water yields. Earth's Future, 6(7), 997–1006. https://doi.org/10.1029/2018EF000930
10.1029/2018EF000930
ADSWeb of Science®Google Scholar
Woodhouse, C. A., Pederson, G. T., Morino, K., McAfee, S. A., & McCabe, G. J. (2016). Increasing influence of air temperature on upper Colorado River streamflow. Geophysical Research Letters, 43(5), 2174–2181. https://doi.org/10.1002/2015GL067613
10.1002/2015GL067613
ADSWeb of Science®Google Scholar
Wyatt, C. J. W., O’Donnell, F. C., & Springer, A. E. (2015). Semi-arid aquifer responses to forest restoration treatments and climate change. Groundwater, 53(2), 207–216. https://doi.org/10.1111/gwat.12184
10.1111/gwat.12184
CASPubMedWeb of Science®Google Scholar
Zhang, F., Biederman, J. A., Dannenberg, M. P., Yan, D., Reed, S. C., & Smith, W. K. (2021). Five decades of observed daily precipitation reveal longer and more variable drought events across much of the western United States. Geophysical Research Letters, 48(7), e2020GL092293. https://doi.org/10.1029/2020GL092293
10.1029/2020GL092293
ADSWeb of Science®Google Scholar
Zhang, M., & Wei, X. (2012). The cumulative effects of forest disturbance on streamflow in a large watershed in the central interior of British Columbia, Canada. Hydrology and Earth System Sciences Discussions, 9(3), 2855–2895.
10.5194/hessd-9-2855-2012
ADSGoogle Scholar
Zhao, F., Zhang, L., Xu, Z., & Scott, D. F. (2010). Evaluation of methods for estimating the effects of vegetation change and climate variability on streamflow. Water Resources Research, 46(3), W03505. https://doi.org/10.1029/2009WR007702
10.1029/2009WR007702
ADSWeb of Science®Google Scholar
Zou, C. B., Breshears, D. D., Newman, B. D., Wilcox, B. P., Gard, M. O., & Rich, P. M. (2008). Soil water dynamics under low-versus high-ponderosa pine tree density: Ecohydrological functioning and restoration implications. Ecohydrology, 1, 309–315. https://doi.org/10.1002/eco.17
10.1002/eco.17
CASWeb of Science®Google Scholar
Zou, C. B., Ffolliott, P. F., & Wine, M. (2010). Streamflow responses to vegetation manipulations along a gradient of precipitation in the Colorado River Basin. Forest Ecology and Management, 259(7), 1268–1276. https://doi.org/10.1016/j.foreco.2009.08.005
10.1016/j.foreco.2009.08.005
Web of Science®Google Scholar

Citing Literature

Volume58, Issue3

March 2022

e2021WR030687

Streamflow Response to Wildfire Differs With Season and Elevation in Adjacent Headwaters of the Lower Colorado River Basin

Abstract

Key Points

Plain Language Summary

1 Introduction

2 Study Watersheds

3 Data Sets

3.1 Streamflow Data

3.2 Climate Data

3.3 Wildfire Data

4 Methods

4.1 Double-Mass Paired Watershed Comparison

4.2 Classifying Wet and Dry Years

4.3 Comparisons of Streamflow, Runoff Ratio, and Climate

4.4 Linear Regression Modeling With Annually Variable Fire Impacts

4.5 Time-Trend Analysis

4.6 Dry Years and Seasons Analysis

5 Results

5.1 Double-Mass Paired Watershed Comparison

5.2 Climate and Streamflow Comparisons

5.3 Regression Modeling

5.4 Time-Trend Analysis

6 Discussion

6.1 Evaluation of Hypotheses

6.2 Biophysical Processes Affecting Hydrologic Response to Large Wildfire

6.3 Increased Summer Streamflow Following Wildfire

6.4 Methodological Comparison, Alignment, and Limitations

6.5 Challenges to Detecting Hydrologic Change in the LCRB

7 Conclusions

Acknowledgments

Open Research

Data Availability Statement

Supporting Information

References

Citing Literature

Figures

References

Related

Information

RESOURCES

PUBLICATION INFO