Installation Matters: Implications for In Situ Water Quality Monitoring

Get Complete Project Material File(s) Now! »

Chapter 3. Effects of Prescribed Fire on Soil Physical and Chemical Properties, Discharge, and Water Quality in a Headwater Catchment in Southwest Virginia 3.1 Introduction

Headwaters, or the upper reaches of larger river networks, make up approximately 70% of total global stream length (Alexander et al. 2007). Headwater streams provide many ecosystem services including sediment and nutrient retention, organic matter processing, flood control, and valuable wildlife habitat. As such, headwater catchments can strongly influence downstream communities and water chemistry (Creed et al. 2017), and any disturbances to a headwater catchment can potentially have local and downstream impacts (Division of Surface Waters 2015). Soils within headwater catchments provide a primary control mechanism for the storage and delivery of water, solutes, and particles to streams (Cruse et al. 1989; Alley et al. 2002). Therefore, it is of utmost importance to quantify and understand how land management practices within headwater catchments affect soil and water quality.
Water chemistry is an important determinant of stream function (Alexander et al. 2007). Key chemical constituents in headwaters include nitrogen (N), suspended sediment (e.g., turbidity), dissolved organic carbon (DOC), and total dissolved salts (e.g., specific conductance). These constituents can have variable sources, effects, and interactions with flow regimes. Nitrogen sources may be either natural (e.g., plant decay, atmospheric deposition) or anthropogenic (e.g., increased deposition, runoff of agricultural fertilizers) and may be transported to the stream through overland flow or as groundwater (Scott et al. 2007). Nitrogen is beneficial to streams as it supports aquatic primary production (Warren et al. 2017), but elevated N concentrations can result in eutrophication (Yang et al. 2018). Sources of sediments include natural or enhanced erosion from both stream banks and adjacent hillslopes (Collins et al. 2017).
Sediment loads can help to create and maintain benthic substrate, but increased amounts can inhibit light penetration through increased turbidity and also smother benthic habitat (Anthony et al. 2004). Lastly, DOC comes from organic carbon sources introduced into the stream mostly originating from allochthonous, terrestrial organic matter (Palmer et al. 2017; Wetzel 1992; Hope et al. 1997). DOC in streams is a carbon source for downstream heterotrophs, serving as the main energy source for biological activity that is non-photosynthetic (Schiff and Aravena 1990; Wetzel 1992; Hornberger et al. 1994; Boyer et al. 1997; Webster and Meyer 1997). Additionally, DOC functions as a major control on metal mobility and concentrations in streams (Mcknight and Bencala 1990). However, elevated DOC may result in a high demand of oxygen and associated low dissolved oxygen conditions (Filter et al. 2017).
There are many different types of natural and anthropogenic disturbances that can potentially impact both headwater catchment properties and water quality. These might include pesticide overuse, logging, and wildland fires (Webster et al. 1992; Roger et al. 1994; Certini 2005). Such impacts to water quality are often an outcome of disturbed soil properties. Disturbance effects to water quality can vary in headwater catchments, ranging from altered DOC delivery dynamics (Jeffrey et al. 2006) and increased nitrate (NO₃) (Knoepp and Clinton 2009) to elevated sediment loads and associated turbidity (Henley et al. 2000). For example, aggregate stability indicates erosion potential and thus sediment loss to downgradient waters (Kemper and Rosenau 1986). Similarly, soil carbon, N, and their ratio (C:N ratios can be used as a measure of soil quality), where changes in these can affect terrestrial productivity (Wieder et al. 2015) and alter delivery of nutrients and DOC to receiving streams (Cooper et al. 2005). Therefore, effective assessment of catchment disturbance effects requires coupled monitoring of both water chemistry and soil properties.
Forest fires can be a natural (e.g., wildfires) or intended (e.g., prescribed) disturbance to forested systems. Both wildfires and prescribed fires can be characterized by burn intensity and severity. Fire intensity is a measure of combustion or energy released (Binkley and Fisher 2020), and is most often quanitified by the flame length at the flaming front of a fire or by examining char heights remaining on trees (Keeley 2009). Fire severity is a measure of fuel consumption and is often measured by organic matter loss. This characteristic can be quantified in a number of ways; some examples include recording the mass of litter in a known area before and after fire or recording the change in litter and duff layer depths after fire (Sikkink and Keane 2012).
Wildfires are typically severe and intense, igniting either naturally as a result of lightning strikes or by humans via arson or carelessness. Within the United States, wildfires frequently occur in the western region during the drier summer months. The relatively arid climate and high combustible fuel loads during that season often create ideal conditions for wildfires to become large, hot, and destructive (Scasta et al. 2016). The average annual acreage burned from wildfires between 2010 to 2017 was 6.6 million (Hoover 2018). Studies performed in the western United States have demonstrated that wildfires may affect catchment soils, vegetation, and stream water chemistry. These stream water impacts may include increased erosion potential and turbidity, altered base flow and stormflows via changes in plant water uptake, and elevated stream DOC and NO₃ concentrations (Rhoades et al. 2011).
Specific wildfire-related water quality effects vary across site conditions, fire behavior characteristics, and specific water quality parameters. There are three main factors related to site and fire conditions that affect water quality in response to wildfires. These factors are 1) burn frequency, duration, and area, including proximity to the stream; 2) climate and weather patterns, specifically antecedent wetness and post-fire precipitation events; and 3) watershed characteristics (e.g., soil types, slope, vegetation type and regeneration after burn) (Hahn et al. 2019). Additionally, fire effects may vary across parameters of interest. For example, sediment loads and turbidity have been shown to increase after wildfires, with subsequent effects on biological processes (e.g., primary production) and downstream systems (Smith et al. 2012; Tecle and Neary, 2015). It is also common for stream nutrient concentrations to increase during and shortly after wildfires (Rhodes et al., 2011) due to increased nutrient-rich ash being transported into streams via overland flow (Sherson et al., 2015).
In contrast to the western United States, large portions of the southeastern U.S. are managed with frequent, low intensity and severity prescribed fires (Melvin, 2018). These fires are used to reduce fuel loads, enhance stand composition (e.g., to restore shortleaf pine and mixed oak forests; Elliott et al. 2005; McArthur 1966), and improve understory habitat. In general, prescribed fires are typically less intense and less severe than wildfires (Fernandes and Botelho 2003). In 2011, over 7.5 million acres of land in the southeastern U.S. were burned with managerial intent to meet a specific prescription (Melvin 2018), representing 69% of total acreage of prescribed fire in the U.S. Most studies concerning fire as a disturbance have focused on wildfires, yet, in many years, prescribed fires account for nearly equal acreage across the United States (Waldrop et al. 2012; Hoover 2018). While some studies have documented negligible effects of prescribed fire on long-term soil nutrients and site productivity (Vose 2000; Elliott, James, and Service 2005), less is known regarding potential prescribed fire effects on watershed attributes (e.g., soil characteristics), water storage and flows, and water quality (Elliott et al. 2005; Harr 2011; Hahn et al. 2019).
Several studies have examined water quality after prescribed fires but have largely focused on changes in sediment loads. For example, Elliot et al. (2005) found no statistically significant differences in total suspended solids (TSS) before and after prescribed fire.
Several mechanisms have been proposed to explain why sediment loads may remain unchanged after prescribed burns, including low fire intensities and severities, non-combustible litter, and riparian areas acting as buffers. Typical prescribed fires burn less than 20% of forest floor biomass, which suggests that much of the forest floor may remain undisturbed (Elliott et al., 2005). The southeastern United States has a humid climate and hardwood or mixed hardwood-conifer forests comprise a large portion of the representative land cover (McNulty et al. 2013). Litter from hardwood species such as maples and hickories can have lower flammability than coniferous species, limiting their combustion (Kane et al. 2008). Further, ash particles, solutes, and sediments may be filtered by unburned forest floor cover and soil layers, particularly in riparian areas, preventing them from being transported to streams (Elliott et al., 2005). While these properties of prescribed fire and southeastern catchments may reduce other water quality impacts, little research has been conducted on the effects of prescribed fire on DOC or specific nutrients. Additional work is needed across different site conditions and water quality parameters to further assess prescribed fire effects.
Fire effects are confounded by temporal drivers, including weather and streamflow, making it difficult to assess effects with infrequent water quality monitoring. For instance, the timing and magnitude of precipitation events before and after fires may play a critical role in wildland fire water quality impacts (Sherson et al., 2015). Post-fire, high intensity rain events have been found to have the greatest effects on post-fire water quality, due to their ability to deliver nutrients and sediment to streams (Sherson et al., 2015). Still, specific water chemistry effects of prescribed, low intensity, low severity fires on headwater streams remain unclear due to the difficulties in measuring multiple water quality parameters at requisite frequencies (e.g., to capture a large rain event immediately after a prescribed fire).
Stream water chemical concentrations can demonstrate great variation and nonlinear responses over a broad range of temporal scales (Krause et al. 2013), but our ability to capture such variation across parameters has been historically limited. Water chemistry sampling and analysis has been extremely limited due to laboratory and personal time costs, and infrequent sampling resolutions (days to seasonal) (Bowes et al., 2009). However, newly developed high resolution sensors have enabled monitoring of fine scale variations of various water quality parameters, including dissolved organic carbon (DOC), turbidity, and NO₃. To adequately analyze water chemistry dynamics and identify possible drivers, high frequency in situ monitoring is necessary (Blaen et al., 2017). High resolution measurements can help identify flow-driven constituent sources and potential catchment disturbance impacts (Kirchner et al. 2004).
It has been well documented that relationships between parameter concentrations (C) and discharge (Q) (i.e., C-Q relationships) can be used as indicators of constituent sources and subsequent stream processing (Moatar et al. 2017), offering a useful tool to explore source and stream dynamics. Several studies indicate that the shape of a C-Q curve can provide valuable insight into particulate and/or solute abundance and source location (Moatar et al., 2017), as well as information concerning biogeochemical uptake and primary production (Montheith et al., 2015; Moatar et al., 2017). C-Q curves can generally be divided into two different types: 1) chemostatic, in which concentration does not vary with discharge, and 2) chemodynamic, in which concentration varies as a function of discharge(Moatar et al. 2017). Chemostatic patterns are due to consistent or homogeneous diffusion from geologic and agricultural sources (Moatar et al., 2017), where concentrations are not affected by changes in hydrological connectivity or flow paths (Musolff et al., 2015; Moatar et al., 2017). Chemostatic relationships have been identified in solutes tied to geological weathering (Moatar et al., 2017), DOC (Creed et al., 2015; Moatar et al., 2017) and for abundant solutes from legacy inputs in agricultural catchments (Basu et al., 2010; Moatar et al., 2017). Chemodynamic C-Q curves can be either positive (i.e., solute enrichment) or negative (i.e., solute dilution) (Godsey et al., 2009; Meybeck and Moatar, 2012; Musolff et al. 2015; Moatar et al., 2017). Positive C-Q curves have been identified with turbidity and some nutrients that likely result from increased erosion during high precipitation events. Further, positive C-Q trends are associated with increased solute supply under conditions of higher hydrologic connectivity, such as DOC from connected wetlands (Holmes et al., 2002; Thomas et al., 2016; Moatar et al., 2017). Typically, this positive slope behavior indicates transport limitations as opposed to production or availability of the solute (Basu et al., 2011; Moatar et al., 2017). Negative patterns have been observed for solutes that become diluted during high flow events (Shanley et al., 2011; Moatar et al., 2017). These solutes are typically source limited because mass flux is based on availability and/or rate of production instead of hydrologic transport (Basu et al. 2010; Moatar et al., 2017).
C-Q relationships are useful at many different temporal scales to provide insight for important watershed characteristics and processes (Hooper et al. 1990; Genereux et al. 1993; Hornberger et al. 2001; Moatar et al. 2017). For example, they have been used at individual precipitation event scales to analyze storm event source dynamics (Evans & Davies, 1998; Chanat et al. 2002; Dupas et al. 2015; Moatar et al., 2017). Relationships have also been used at seasonal and multi-year scales to gather information about how export dynamics vary with different catchment qualities, such as topography, vegetation, and even human activity (Butturini et al. 2008; Godsey et al. 2009; Moatar et al., 2017). Additionally, C-Q relationships may reveal long term changes in hydrological and biogeochemical functioning related to changing environmental conditions, i.e., land-use, climate, or disturbance (Petrone et al. 2006; Guzman et al. 2013; Törnqvist et al. 2015). Thus, analyzing and understanding C-Q relationships is a valuable tool for understanding catchment processes and to observe catchment responses to changing drivers.
Recent advances in water quality monitoring technologies and methodologies (Rhoades et al., 2011) have created an opportunity to better assess prescribed fire effects by using high temporal resolution, in situ water quality sensors (Sherson et al., 2015). In this project, we assessed soil attributes and collected high frequency measurements for flow and multiple water quality parameters (NO₃, DOC, and turbidity) before and after a large, prescribed fire in a headwater catchment (No Business Creek) in southwest Virginia. We also monitored flow at an unburned neighboring headwater catchment (Dismal Creek) to evaluate if prescribed fire influenced water flows at No Business. Using soil property characterization and high resolution, in situ monitoring, we addressed the following primary objectives:
Assess how soil particle size, aggregate stability, and carbon and N contents change following prescribed fire.
Evaluate effects of prescribed fire on stream discharge and water chemistry dynamics.

Methods

Site Descriptions

No Business Creek and Dismal Creek are two neighboring headwater streams located in the Appalachian Mountains of southwest Virginia (Figure 14). In this project, No Business Creek served as the treatment watershed with a prescribed fire, and Dismal Creek acted as the control for stream discharge comparison. Both watersheds are completely forested and are maintained by the US Forest Service. These neighboring watersheds have varying strike and dip in the underlying bedrock, creating planar exposed sheets of bedrock in Dismal, and crumbling blocks of bedrock in No Business (McDowell and Schultz 1990); see Section 2.2 for more details on site characteristics).
The U.S. Forest Service performed a prescribed burn that covered the entirety of the upper No Business Creek watershed, approximately 1560 hectares in total. This burn was conducted on May 2^nd-3^rd, 2018.

Soil Sampling and Analysis

Sampling and Fire Severity

To assess soil attributes before and after fire (Objective 1), soil samples were collected at two different sites (midstream, referred to as the Middle site, and downstream, referred as the Lower site) within the No Business watershed. At both sites, soil samples were collected along four different transects that paralleled the stream, with transect 1 located nearest the stream and transect 4 located farthest upslope. These transects were established approximately 15 m apart at each site. At the Middle site, transect 1 was located ~100 m from the stream bank and transect 4 was ~150 m away from the stream bank (Figure 15a); at the Lower site, transect 1 was located ~5 m from the stream edge and transect 4 was ~ 50 m from the stream bank (Figure 15b). Samples were collected at three locations along each transect ~ 10 m apart, resulting in 12 sampling points at each site for a total of 24 sampling locations.
Fire severity was assessed at each sampling point by quantifying pre- and post-fire forest floor depth and mass (Keeley 2009). To evaluate pre-and post-fire litter depth, nails were driven into the ground. The heads of these nails touched the top of the litter layer after installation.
Litter depth was measured following nail installation. Post-fire, these nails were revisited to measure the distance between the remaining surface litter layer and the top of the nail.
Post-fire char heights were measured on trees within the burned area as an approximation of fire intensity (Keeley 2009). Every tree within 10 m of each sampling location was inspected and measured for char heights. All measurements were taken from the tree center in order to reduce error in burn heights due to slope. In total, 38 trees were analyzed at the Middle site and 33 trees were analyzed at the Lower site.
At each sampling location, separate samples for litter, duff, and mineral soil were collected at three times: April 11^th, 2018 (1 month before the fire); May 8^th, 2018 (4 days after the fire), and September 5^th, 2018 (3 months after the fire). A 930 cm² quadrant was set on the upper surface at each sampling point, and all material in the litter (i.e. the Oi Horizon, consisting of fallen leaves that are fresh or have become slightly decomposed), and duff layers (i.e. Oe+Oa Horizons, consisting of decomposed litter over time where individual tree species can no longer be identified) were collected. Mineral soil samples were also collected from underneath the duff, at an average depth of 10-15 cm. Samples were collected in the field in plastic bags, transported to the lab, and transferred into brown paper bags. Half of each mineral soil sample was air dried, while the remaining mineral soil, duff, and litter samples were placed in a drying oven at 60°C for no less than one week. All mineral samples were then hand-sieved (2 mm), and the coarse fraction was discarded. Additionally, for air dried samples, all large organic pieces (e.g. roots > 2 mm in length) were removed.

Laboratory Analysis

To assess fire severity, litter loss (%) was calculated for each sampling point as:
Litter loss = 100*(m_pre – m_post)/m_pre (3.1) where m_pre is the dry mass of litter before the fire and m_post is the mass of litter after the fire (on May 8^th, 2018).
For aggregate stability, the air-dried, sieved mineral soil samples were used to compare aggregates and particle size distributions. A laser diffraction machine (CILAS 1190, CILAS Inc., Ariane Group, and Orleans, France) was used to determine aggregate size distributions. To determine particle size, 1 g of sample was mixed with 1 ml of 50 g L^-1 sodium hexametaphosphate (Na₆O₁₈P₆) and 4 ml of deionized water. This slurry was shaken on a mechanical shaker for at least 4 hrs and then immediately added to the laser diffraction machine, subjected to 130 s of sonication using the built-in 25 W ultrasound unit, and then analyzed for particle size. To estimate aggregate stability, we compared cumulative distribution functions for aggregate size distributions to those for particle size distributions via the integrated aggregate stability (IAS) metric as developed by Gyawali and Stewart (2019):
where λ is the ratio of specific volumes of the particles versus aggregates, F_p is the cumulative size distribution for particles, and F_a is the cumulative size distribution for aggregates, both estimated at x₁, which represents the particle/aggregate size at which the two samples (particles and aggregates) have the same volume-corrected probability density. Larger values of IAS indicate greater water-stable aggregation and better soil stability.
Oven-dried mineral soil, litter, and duff samples were analyzed for total N and carbon. Litter samples were weighed and then combined with their corresponding duff sample before being ground using a Wiley Mill (2 mm). A vario MAX N/CN (elementar, Analysensysteme GmbH, Hanau, Germany) was used to analyze soil and litter/duff samples for total carbon (%) and total N (%) contents, which were then used to estimate the C:N ratio for each sample. For mineral soils, 0.5-1.0 g soil was used, while forest floor analyses used between 0.2-0.5 g.

Data Analyses

An Analysis of Variance (ANOVA) with Tukey’s HSD test was used to assess significant differences in the means of IAS, total carbon (%), total nitrogen (%), and C:N ratios for the three sampling times (1 month before fire, 4 days after fire, 3 months after fire). An alpha value of 0.05 was used to identify significance.

Discharge and Water Chemistry

Instrumentation

To assess changes in stream discharge and water chemistry between pre- and post-fire conditions (Objective 2), we instrumented the Lower site at No Business Creek with a suite of in situ sensors. First, a S::CAN spectro::lyser UV-Vis (s::can Messtechnik GmbH, Vienna, Austria) was used to record high resolution data for the following parameters: turbidity, dissolved organic carbon (DOC), and nitrate (NO₃). These parameters were collected at 15 min intervals. The S:CAN lamp was kept clean by an automatic brush attachment that cleaned the lamp four times a day and via bi-weekly cleaning during field visits using de-ionized (DI) water and Kimtech wipes. Lamp consistency was tested during field visits by taking readings in DI water and ensuring that these blank readings remained similar throughout the study period.
We also installed screen water level monitoring wells (80 mesh screen, 1 m long, with a 5.08 cm diameter) at both the Lower No Business Creek and Dismal Creek sites. Each well had a sealed pressure transducer (HOBO U20L-04 Onset Corporation, Bourne, Massachusetts, United States). Barometric pressure variations were recorded using an additional pressure transducer installed in a dry well, limiting temperature variation and known induced error (McLaughlin et al., 2011). The pressure data in each well were corrected for barometric pressure before being converted into stage (m) by accounting for water density and gravitational acceleration. Further, we installed specific conductance data loggers (HOBO U20L-04, Onset Corporation, Bourne, Massachusetts, United States) at both Dismal and No Business Creeks in an open housing (see Chapter 2) attached to the water level monitoring well; these data were used during salt tracer tests to estimate discharge (see section 3.2.6)
Each watershed was also equipped with one logging tipping bucket rain gauge (Onset HOBO RG3, Onset Corporation, Bourne, Massachusetts, United States). All sensors were programmed to record at 15 min intervals. Data collection started on July 12^th, 2017 (10 months pre-fire), and continued until March 13th, 2019 (10 months post-fire).

TABLE OF CONTENTS
ABSTRACT
GENERAL AUDIENCE ABSTRACT
ACKNOWLEDGEMENTS
TABLE OF CONTENTS
TABLE OF FIGURES
Chapter 1. Introduction
1.1 Functions and Chemistry of Headwater Streams
1.2 Monitoring Needs
1.3 Fire Effects on Headwater Stream Chemistry
1.4 Objectives
Chapter 2. Installation Matters: Implications for In Situ Water Quality Monitoring
2.1 Introduction
2.2 Methods
2.3 Results
2.4 Discussion
2.5 Conclusions
2.6 Tables
2.7 Figures
Chapter 3. Effects of Prescribed Fire on Soil Physical and Chemical Properties, Discharge, and Water Quality in a Headwater Catchment in Southwest Virginia
3.1 Introduction
3.2 Methods
3.3 Results
3.4 Discussion
3.5 Conclusion
3.6 Tables
3.7 Figures
Chapter 4: Conclusions 
REFERENCES
GET THE COMPLETE PROJECT
Measuring and Understanding Effects of Prescribed Fire in a Headwater Catchment