Characterization and evaluation of controls on post-fire streamflow response across western US watersheds

Abstract

Abstract. This research investigates the impact of wildfires on watershed flow regimes, specifically focusing on evaluation of fire events within specified hydroclimatic regions in the western United States, and evaluating the impact of climate and geophysical variables on response. Eighty-two watersheds were identified with at least 10 years of continuous pre-fire daily streamflow records and 5 years of continuous post-fire daily flow records. Percent change in annual runoff ratio, low flows, high flows, peak flows, number of zero flow days, baseflow index, and Richards-Baker flashiness index were calculated for each watershed using pre-and post-fire periods. Independent variables were identified for each watershed and fire event, including topographic, vegetation, climate, burn severity, percent area burned, and soils data.Results show that low flows, high flows, and peak flows increase in the first 2 years following a wildfire and decrease over time. Relative response was used to scale response variables with the respective percent area of watershed burned in order to compare regional differences in watershed response. To account for variability in precipitation events, runoff ratio was used to compare runoff directly to PRISM precipitation estimates. To account for regional differences in climate patterns, watersheds were divided into nine regions, or clusters, through k-means clustering using climate data, and regression models were produced for watersheds grouped by total area burned. Watersheds in Cluster 9 (eastern California, western Nevada, Oregon) demonstrate a small negative response to observed flow regimes after fire. Cluster 8 watersheds (coastal California) display the greatest flow responses, typically within the first year following wildfire. Most other watersheds show a positive mean relative response. In addition, simple regression models show low correlation between percent watershed burned and streamflow response, implying that other watershed factors strongly influence response.Spearman correlation identified NDVI, aridity index, percent of a watershed's precipitation that falls as rain, and slope as being positively correlated with post-fire streamflow response. This metric also suggested a negative correlation between response and the soil erodibility factor, watershed area, and percent low burn severity. Regression models identified only moderate burn severity and watershed area as being consistently positively/negatively correlated, respectively, with response. The random forest model identified only slope and percent area burned as significant watershed parameters controlling response.Results will help inform post-fire runoff management decisions by helping to identify expected changes to flow regimes, as well as facilitate parameterization for model application in burned watersheds.

Long‐term hydrologic recovery after wildfire and post‐fire forest management in the interior Pacific Northwest

“…Due to increased wildfire activity, there has been substantial research in recent years on the effects of wildfire on discharge; however, the vast majority of studies have focussed on the initial (<5 years) post‐fire effects (Hallema et al, ; Kinoshita & Hogue, ; Smith, Sheridan, Lane, & Bren, ). Comparatively fewer studies have investigated the longer‐term legacy of wildfire effects on catchment hydrology (Hallema et al, ; Saxe, Hogue, & Hay, ). Certainly, the majority of studies have illustrated the greatest effects during the first several years, followed by a decline at various rates before returning to a near pre‐fire condition (Ebel & Mirus, ; Vieira, Malvar, Fernandez, Serpa, & Keizer, ).…”

Hillslope sediment production after wildfire and post‐fire forest management in northern California

“…As such, concerns have grown regarding the immediate and longer‐term effects on forest resilience and the water supply originating in forests (Hallema, Robinne, & Bladon, 2018; Stevens‐Rumann et al, 2018). Increasingly severe wildfires have produced substantial and long‐lasting (>10 years) effects on annual streamflow and peak flows (Hallema, Sun, et al, 2018; Niemeyer, Bladon, & Woodsmith, 2020; Saxe, Hogue, & Hay, 2018), debris flows (Langhans et al, 2017; Nyman et al, 2015), physical and chemical water quality (Rhoades et al, 2019; Rust, Hogue, Saxe, & McCray, 2018), aquatic ecosystem health (Bixby et al, 2015; Emelko et al, 2016), and downstream drinking water supply (Emelko, Silins, Bladon, & Stone, 2011; Hohner, Terry, Townsend, Summers, & Rosario‐Ortiz, 2017).…”

Year-ahead predictability of South Asian Summer Monsoon precipitation

“…The predicted value is then the mean across the regression tree ensemble [43]. RF is robust to the presence of correlated or unhelpful predictors [44] and has been successfully used for many earth science applications, generally comparing favorably with linear and other nonlinear methods [45][46][47][48][49][50][51][52]. The implementation in the R randomForest package [53] was used.…”