Life after a fiery death: Fire and plant biomass loading affect dissolved organic matter in experimental ponds

Drier and hotter conditions linked with anthropogenic climate change can increase wildfire frequency and severity, influencing terrestrial and aquatic carbon cycles at broad spatial and temporal scales. The impacts of wildfire are complex and dependent on several factors that may increase terrestrial deposition and the influx of dissolved organic matter (DOM) from plants into nearby aquatic systems, resulting in the darkening of water color. We tested the effects of plant biomass quantity and its interaction with fire (burned vs. unburned plant biomass) on dissolved organic carbon (DOC) concentration and degradation (biological vs. photochemical) and DOM composition in 400 L freshwater ponds using a gradient experimental design. DOC concentration increased nonlinearly with plant biomass loading in both treatments, with overall higher concentrations (>56 mg/L) in the unburned treatment shortly after plant addition. We also observed nonlinear trends in fluorescence and UV‐visible absorbance spectroscopic indices as a function of fire treatment and plant biomass, such as greater humification and specific UV absorbance at 254 nm (a proxy for aromatic DOM) over time. DOM humification occurred gradually over time with less humification in the burned treatment compared to the unburned treatment. Both burned and unburned biomass released noncolored, low molecular weight carbon compounds that were rapidly consumed by microbes. DOC decomposition exhibited a unimodal relationship with plant biomass, with microbes contributing more to DOC loss than photodegradation at intermediate biomass levels (100–300 g). Our findings demonstrate that the quantity of plant biomass leads to nonlinear responses in the dynamics and composition of DOM in experimental ponds that are altered by fire, indicating how disturbances interactively affect DOM processing and its role in aquatic environments.


| INTRODUC TI ON
Wildfires play a major role in global carbon (C) cycles and are on the rise in many regions as climate change makes vegetation more susceptible to burning (Giglio et al., 2013;Halofsky et al., 2020;Harvey & Enright, 2022).Geography, climate conditions, human disturbance, and plant community composition shape regional fire regimes (Bond, 2013).The relative significance of these factors for driving trends in fire activity differs greatly across global ecoregions (Bowman et al., 2009;McKenzie et al., 2004).Nonetheless, it is clear that drier and hotter conditions, linked with anthropogenic climate change, increase wildfire frequency, severity, and intensity, thereby influencing terrestrial C cycles at broad spatial and temporal scales (Abatzoglou & Williams, 2016;Keeley, 2009;Lasslop et al., 2019).
Despite covering a small fraction of the Earth's surface, freshwater ecosystems play a critical role in global C cycles.Global estimates suggest that approximately 0.2 Pg C year −1 is stored in lake sediments, 0.8 Pg C year −1 is emitted back as gas exchange, and the remaining 0.9 Pg C year −1 is transported to the oceans (Cole et al., 2007;Mulholland & Elwood, 1982).The processing of organic and inorganic materials from both internal (autochthonous) production as well as terrestrial (allochthonous) sources regulates lentic and lotic metabolism that directly influences trophic dynamics and nutrient cycling.For instance, dissolved organic matter (DOM) leached from terrestrial sources can lead to net heterotrophy where lake community respiration exceeds internal primary production, indicating the degree to which allochthonous inputs control lake ecosystem function (e.g., Berggren et al., 2022;Fonseca et al., 2022;Hanson et al., 2003).The annual amount of C buried in lake sediments is equivalent to the amount of C stored in the ocean through the C pump (Ward et al., 2017), further illustrating how the fate of DOM in lakes and other inland water bodies impacts global C cycles (Cole et al., 2007;Paul et al., 2022;Wall et al., 2023).
Wildfires liberate C from terrestrial sources and serve as catalysts for ecosystem change.First, wildfires can trigger landscape shifts such as the replacement of woody vegetation or shrublands with more flammable grasslands (D'Antonio & Vitousek, 1992;Zedler et al., 1983).Second, burning releases C through combustion and depletes stored C in plant biomass and soils (Pellegrini et al., 2015).
Wildfires may cause changes in plant and soil organic matter composition, and nutrient cycling processes through heat-induced alterations of plant tissues and modifications in nutrient availability and storage (Cornelissen et al., 2017;Pellegrini et al., 2015).Lastly, soil destabilization and hydrophobicity from severe wildfires accelerate runoff and erosion of soil and ash that can be deposited into streams, rivers, and lakes (Larsen et al., 2009;Lewis et al., 2019;Santos et al., 2019).
The link between wildfires and aquatic ecosystem function depends on the severity, proximity to water, and rate of post-fire ecosystem recovery (Stevens-Rumann & Morgan, 2016).The persistence and bioaccumulation of C particulates within aquatic ecosystems from burned (i.e., combustion-derived carbon [CDC]) and unburned terrestrial sources can have severe consequences on the feeding behavior, growth, and reproduction of aquatic organisms (Bistarelli et al., 2021;Campos & Abrantes, 2021;Cooper et al., 2015;Robidoux et al., 2015) as well as on post-fire nutrient and metal concentrations in water (Paul et al., 2022).These effects have been associated with changes in the quantity and quality of DOM, including the dissolved organic carbon (DOC) fraction of DOM imported from the surrounding landscape (reviewed in McCullough et al., 2019).
For instance, lake water color can darken over time as a result of elevated DOC levels (Kritzberg et al., 2020).Light-absorbing chromophoric DOM compounds correspond with the type and amount of imported organic material from the watershed (Cory et al., 2007;Helms et al., 2013;Laurion & Mladenov, 2013).This creates the potential for cascading effects on the cycling of DOM since burned watersheds are more prone to erosion and thus increased terrestrial deposition into aquatic systems (Bixby et al., 2015;Larsen et al., 2009).
The complex interplay among post-fire erosion and sedimentation, DOM, and their subsequent effects on aquatic systems is unclear.Allen et al. (2003) surveyed water chemistry 2 years after a wildfire on Alberta's Boreal Plain and found that mean DOC concentrations in lake water from burned watersheds was 1.4-fold higher compared to reference watersheds.Studies in the Western United States have shown small or variable DOC increases following fires (Emelko et al., 2011), or no changes in DOC despite large increases in other solutes (Granath et al., 2021).In higher latitude ecosystems, DOC responses to wildfires have been reported as increasing (Marchand et al., 2009), decreasing (Betts & Jones, 2009), or remaining unchanged (Olefeldt et al., 2013).Variation in post-fire DOC responses may also be attributed to specific factors like fire intensity (i.e., the temperatures at which vegetation is burned, see Keeley, 2009), or DOM composition from vegetation in the watershed that results in greater selection pressures on decomposer communities (Chen et al., 2022;Olefeldt et al., 2013).Thus, heterogeneity in the vegetation of burned watersheds, DOM sources (i.e., plants, soils, or ashes), changes in organic matter (i.e., formation of CDC compounds) can all determine the fate and processing of C compounds in aquatic ecosystems.
Changes in DOM composition from fire can significantly affect the two dominant controls of degradation (photochemical and microbial) in aquatic biogeochemical C cycling (Morris & Hargreaves, 1997;Obernosterer & Benner, 2004;Wetzel, 1984).Burned organic material (collectively, CDC and pyrogenic carbon [PyC]) can produce hydrophobic compounds and chars, or contribute to alterations in pH, redox potential, and thermal decomposition, all of which significantly influence photo-and microbial degradation (Tshering et al., 2023;Wang et al., 2015;Wetzel, 1984).For example, Bostick et al. (2021) found that PyC compounds strongly influence the composition and diversity of bacterial communities with some PyC fractions promoting and inhibiting certain taxa, suggesting that different types of PyC have varying degrees of biolability.Additionally, Chen et al. (2022) demonstrated that high concentrations of aromatic PyC compounds can significantly decrease microbial growth, diversity, and metabolism.Aromatic PyC, as well as other CDC compounds, can thus shape the structure of microbial communities in aquatic environments (also see Bertilsson & Tranvik, 2000;Olefeldt et al., 2013).Identifying changes to DOM composition from fire will provide insight into how DOM persistence, fate, and reactivity impact ecosystem processes such as C storage, nutrient cycling, and water quality.
This study addresses the effects of fire and plant biomass loading on DOC concentration and degradation, as well as DOM composition, in experimental freshwater ponds (hereafter, ponds).Using an array of 30 (400 L) ponds with a gradient design of increasing plant biomass from burned or unburned terrestrial sources (hereafter, treatments), we ask the following: (1) How do fire (burned or unburned plant biomass) and the quantity of plant biomass jointly affect DOC concentration?(2) How does burning plant biomass affect DOM chemical composition?and (3) What are the relative contributions of microbial activity and photodegradation on DOC degradation between ponds receiving burned or unburned plant biomass?We predict an increase in DOC concentrations from initial C leaching, but a smaller increase in the ponds receiving burned plant biomass due to C depletion and stoichiometric changes from combustion (Cornelissen et al., 2017;Pellegrini et al., 2015).We also hypothesize that fire and greater plant biomass loading should weaken photo-and microbial DOC degradation from the combined effects of increased light attenuation due to shading, reduced dissolved oxygen concentrations at high plant biomass loading, and greater DOM aromaticity derived from burned biomass.
While past studies have tackled similar questions regarding aquatic ecosystem C dynamics: comparisons of DOC quantity (McEachern et al., 2000;Olivares et al., 2019;Tshering et al., 2023) or quality (Bixby et al., 2015;Burd et al., 2020;Marchand et al., 2009;Olefeldt et al., 2013) in watersheds with/without or before/after burns, or lab incubations of single concentrations (Bostick et al., 2021;Chen et al., 2022), we specifically investigate how fire induces C transformations across a broad concentration range in diverse experimental pond ecosystems that include microbial decomposers, primary producers, and zooplankton consumers.This unique experimental approach allows us to detect critical thresholds where variations in plant biomass alter the chemistry and dynamics of DOM in aquatic ecosystems and to determine how these thresholds are affected by chemical transformations due to burning.

| Plant biomass
We selected two California native plant species Salvia leucophylla (hereafter, sage) and Salix lasiolepis (hereafter, willow) to simulate terrestrial loading effects.We chose sage because it has a wide distribution, growing on arid, sandy, and rocky soils, and is common throughout western North America at elevations from 762 to 3000 m (Keeley & Fotheringham, 2001).We selected willow because it grows in riparian zones along rivers and streams, making it an important component of these habitats' ecosystem dynamics.
Additionally, as native species to California found in fire-prone scrub and woodland ecosystems, including willow and sage allowed us to assess the effects of plant biomass loading and fire on DOM dynamics in a regionally significant context.
We purchased 23 sage plants from a nursery on June 9, 2021, transplanted them into pots with a sand-vermiculite mixture, and grew them for 60 days.We harvested willow from the University of California Dawson Los Monos Reserve in San Diego, CA.We airdried cut stems (~1 cm diameter, 3-5 cm long) and leaves from sage and willow branches inside a greenhouse for 24 h.The plant material was then placed in a drying oven at 45°C for 24 h until dry before being introduced into the ponds.

| Simulated wildfire on plant biomass
We burned a portion of sage and willow in a 75-L aluminum container using a butane torch to simulate the wildfire effects (Figure S1a).We used the container's lid to regulate oxygen flow and smother the flames as needed which ensured that the plant biomass was not completely combusted.We used visual assessment to divide the plant biomass into two burn severity types: low severity (intact green leaves showing noticeable black burns on the stems and sticks) and high severity (gray leaves, sticks, and stems including ash; Figure S1a,b).The two severity types generally represent the characteristics of post-fire vegetation as wildfire intensity and severity can lead to variable appearances (Coppoletta et al., 2016;Keeley, 2009).We pooled the biomass of the two severity types to form one burned treatment.We then individually packed the burned and unburned biomass of each species into 25 × 15 cm nylon bags with a mesh size of 250 μm (Figure S1c).For example, we packed all burned sage into its own bags, all unburned sage into separate bags, and the same was done for each type of willow.Leaf litter bags facilitated submersion and containment upon introduction to the ponds and allowed for microbial colonization and grazing by small invertebrates as well as measuring mass loss at the end of the experiment (Figure S1e).
After weighing the initial mass, we dried all litter bags in a drying oven at 45°C for 24 h before being placed in their assigned ponds.
We used equal amounts of sage and willow to achieve the target mass for each pond.For instance, a pond with 10 g of total plant biomass contained 5 g of sage and 5 g of willow.This was done for each treatment.We recovered all litter bags at the end of the experiment (oven-dried at 45°C for 7 days) for final mass measurements.
We used initial and final masses to calculate leaf litter decomposition as percent mass loss over the course of the experiment.

| Experimental pond design
In October of 2021, we assembled and filled 30 ponds (400 L plastic cattle ponds) with fresh water at the University of California, San Diego Biology Field Station (Figure S1d).We used equal aliquots of a plankton mixture collected from Lake Miramar in San Diego, CA, using a vertical tow (64 μm mesh net; 7.6 m depth) to stock the ponds.This controlled approach allowed us to account for potential variation in the initial conditions across the ponds (i.e., Day 0, see Section 3) before plant addition and to ensure that the ponds contained similar resident plankton communities.We refilled the ponds each week with fresh water to replace losses to evaporation.
In the gradient design, each pond received incremental increases of plant biomass ranging from 0 to 400 g for a total of 15 ponds per treatment (burned vs. unburned).This design allowed us to test for thresholds and nonlinearities in the response of pond ecosystems along a broad concentration gradient (Cottingham et al., 2005).We filled the ponds on October 27, 2021; added the plankton mixture on November 1, 2021; added the plant biomass on November 5, 2021; and ended the experiment on February 2, 2021.

| In situ incubations
We reported the C fraction of DOM as DOC which was measured in mg/L, while DOM reflected the broader mixture of aliphatic and aromatic compounds, which we measured using fluorescence and UV-visible absorbance spectroscopy.We conducted three incubation experiments (on Days 10, 31, and 59) using Sterile Whirl-Pak© bags (hereafter, whirlpaks) within each pond to test the contributions of photo-and biodegradation on DOC loss (Figure S1f).We Each pond received four whirlpaks subjected to one of four conditions manipulating the presence of microbes and UV light.We used clear and opaque whirlpaks for UV transmission or complete light exclusion, respectively.We collected 100 mL of water from each pond and filtered (F) it through pre-combusted (2 h at 550°C) glass-fiber filters (0.7 μm GF/F; Whatman, Maidstone, UK) to limit microbial abundance.We filled the other whirlpaks with 100 mL of unfiltered water (UF) that included microbes resident in the ponds.This resulted in four conditions in a 2 × 2 factorial design: (3) UV excluded-unfiltered [NoUV-UF]; and (4) UV excluded-filtered [NoUV-F].We secured all whirlpaks (four whirlpaks per pond × 30 ponds, n = 120) to 8.8 × 25.4 × 7.6 cm plastic trays and suspended them ca.15 cm below the surface in every pond.At the end of the incubation period (7 days; see Dempsey et al., 2020), we removed and filtered all whirlpaks through pre-combusted 0.7 μm GF/Fs into muffled (5 h at 550°C) borosilicate amber vials (60 mL) and transported them on ice to the Water Innovative and Reuse Lab (WIRLab) at San Diego State University (SDSU).
Filtration is essential for measuring microbial decomposition of DOC with studies recommending a 0.45 or 0.2 μm pore size to remove microbiota from samples (Bertilsson & Tranvik, 2000;Magyan & Dempsey, 2021).Many 0.45 and 0.2 μm filters, however, are composed of organic material such as cellulose acetate, which can induce contamination problems, leading some researchers to recommend 0.7 μm GF/F filters (Denis et al., 2017;Khan & Subramania-Pillai, 2007).We were unable to confirm the effectiveness of filtration in removing microbes at the time of sampling or during storage due to methodological constraints.
We used UV transmission and exclusion, and filtration to distinguish between the effects of photodegradation versus microbial decomposition on DOC loss.The percent change in DOC concentration between the four whirlpaks yielded effects for individual degradation pathways: (1) photodegradation, (2) microbial decomposition, and (3) the combined effect of both photo-and microbial degradation.That is, we calculated the percent change in DOC concentration over the 7-day incubation for each whirlpak in every pond: the photodegradation effect (or UV-only) as the difference in DOC between the UV-F and NoUV-F relative to NoUV-F, the microbe effect (or microbes-only) as the difference in DOC percent change between the NoUV-UF and NoUV-F relative to NoUV-F, and the combined microbial and photochemical effect (or UV and microbes) as the difference between the UV-UF and NoUV-F relative to NoUV-F.

| DOC concentration analysis
We measured DOC concentration in all 30 ponds at four distinct time intervals: 10, 31, 59, and 89 days following the introduction of plant material.Day 0 shows consistent baseline DOC concentrations across the ponds before plant addition (see Section 3).We accomplished this using an integrated water column sampler, collecting water at different locations to create a DOC sample across the water column (approximately 45 cm from the surface) for each pond.We filtered the water samples through pre-combusted 0.7 μm GF/Fs into muffled borosilicate amber vials, and then acidified them with 37% HCl to a pH of 3. We analyzed DOC concentration using a high-temperature combustion method (Shimadzu TOC-L Total Organic Carbon Analyzer) in the WIRLab at SDSU.
We calibrated samples with potassium hydrogen phthalate standards (ranging from 1 to 50 mg C/L) and analyzed them according to WIRLab standard protocols with ~10% of samples in duplicate.
For all duplicated samples, standard deviations were within 10% of mean values; we either re-analyzed or excluded sample results falling outside of this range.

| Excitation emission matrix spectral analysis
We used fluorescence excitation emission matrix (EEM) data to characterize DOM chemical composition.We collected and individually filtered water from every pond (n = 30) through a 64-μm mesh into 100 mL acid-washed (10% HCl) polypropylene containers for EEM spectral analysis.The filter mesh was rinsed with ddH 2 O between collections to prevent cross-pond contamination.We filtered samples into muffled borosilicate amber vials using 0.7 μm GF/Fs, refrigerated them (<24 h), and then transported the vials on ice to the WIRLab for analysis.EEMs measure the excitation and emission wavelengths at which fluorescence occurs to characterize specific molecular structures (Coble, 1996).A fraction of chromophores in DOM are fluorescent compounds (Stedmon & Nelson, 2015), and certain regions present in the EEMs were used to characterize aspects of the chemical composition in the ponds.
We acquired optical properties, UV-visible absorbance and fluorescence simultaneously (at room temperature of 21-23°C), using a Horiba Scientific Aqualog Fluorometer on filtered water samples in a quartz cuvette with a path length of 1 cm.We inner filter corrected EEMs fluorescence, blank-subtracted and Raman-normalized, and excised first-and second-order Rayleigh scatter bands (see Laurion & Mladenov, 2013).

| Indices generated by UV-visible absorbance and fluorescence spectroscopy
We calculated three fluorescence indices to characterize DOM chemistry (Table 1): the humification index (HIX = ratio of areas under the emission curve at 435-480 nm and 300-345 nm plus 435-480 nm at an excitation wavelength 245 nm) that described the degree of humification (high HIX = more terrestrial humic-like material; Zsolnay et al., 1999); the freshness index (BIX), representing the degree to which organic matter has undergone processing and degradation, as the ratio of emission intensity at 380 nm to the maximum emission intensity between 420 and 435 nm at excitation wavelength 310 nm (Parlanti et al., 2000); and the fluorescence index (FI = emission intensity at 470 nm divided with that of 520 at 370 nm excitation) which indicated source material (microbial-precursor FI value ≈ 1.8 or terrestrial-precursor FI ≈ 1.3;McKnight et al., 2001).
Additionally, we further classified the DOM chemical characteristics by calculating two UV-visible absorbance indices (Table 1): the ratio of the slope (S R ) parameters (S 275-295 :S 350-400 ) as a proxy for molecular weight (MW; Helms et al., 2008), and the specific UV absorbance at 254 nm (SUVA 254 ) to indicate aromaticity (Weishaar et al., 2003).

| Statistical analyses
We examined the effects of treatment and plant biomass quantity on several response variables using general additive models (GAMs) implemented in the mgcv package (Wood, 2017) using R version 4.3.1 (R Core Team, 2022).All data were plotted across the plant biomass gradient with points representing individual ponds.Flexible smoothers in a GAM framework were well suited to model the nonlinear responses in our gradient experimental design.We analyzed the relationship between plant biomass loading and response metrics to detect potential nonlinearities and thresholds.Regression designs without replication outperform replicated designs for quantifying nonlinear responses (Kreyling et al., 2018).Many ecological experiments rely on replicated designs with limited sampling along environmental gradients.Unreplicated gradient designs may be more powerful and appropriate for addressing questions aimed at understanding nonlinear responses and complex interactions among multiple environmental factors (Cottingham et al., 2005;Kreyling et al., 2018).We used a plant biomass gradient to detect thresholds and nonlinear relationships in response to loading of plant detritus, and how these are affected by burning.
We estimated the effects of treatment and plant biomass quantity using separate GAMs with a Gaussian error distribution for all response variables.Models for each parameter were individually analyzed at the four discrete sampling days since results for a single full model introduced extreme concurvity (>0.9).We selected final models with the lowest generalized cross-validation (GCV) and Akaike information criterion scores (Tables S1-S4), and greatest deviance explained (DE) using backward stepwise selection.Lower GCV scores typically correspond with higher DE indicating that a model minimizes the smoothed predictors while maximizing explanatory power (Wood, 2017).In our fitted GAMs, the significance of

| Pond DOC concentrations
DOC concentrations increased nonlinearly 10 days after plant addition with a significant difference between treatments (Figure 1).
DOC values ranged from 5.0 to 56.5 mg/L along the plant biomass gradient.Intermediate plant biomass amounts for the burned treatment at 10 days showed about 20% less DOC compared to the unburned treatment as evidenced by the differences between the fitted smooth functions (difference in trends; solid lines) shown in red with approximate 95% confidence intervals (p = .005,Figure S2).
Averaged across all plant biomass amounts, the burned treatment had 8 mg/L less DOC than the unburned treatment at Day 10.After 31 days, we observed a subsequent decline in DOC concentrations that continued until the end of the experiment at 89 days.By Day 89, DOC concentrations decreased to an average of 15 mg/L across the biomass gradient with no significant difference between treatments (Table S5).

| Fluorescence and UV-visible absorbance indices
We detected significant effects of plant biomass quantity and fire-induced alterations on the fluorescence (HIX, BIX, and FI) and UV-visible absorbance (SUVA 254 and S R ) indices (Table S6), which The BIX revealed a significant effect of treatment and plant biomass quantity on Day 10, with more recently produced DOM in the unburned treatment at high biomass levels (>250 g; p = .01,Figure 2d).The differences between treatment persisted to Day 31 but stabilized and remained relatively constant over time.
The FI index displayed a positive, nonlinear trend along the plant biomass gradient.We observed greater microbial sources (average FI value ~2.0) in both treatments at intermediate plant biomass levels (75-175 g) on Day 10 (Figure 2e).The burned treatment showed a stronger unimodal trend, though greater FI values in the unburned treatment also increased with biomass that declined rapidly after a certain threshold (>250 g).The trends in FI became more linear by Day 59, with values around 1.8, suggesting a mixture of both terrestrial and microbial sources of DOM in the ponds, with no significant difference between treatments.

| In situ DOC degradation
Proportional DOC loss was more strongly linked to the amount of plant biomass than treatment, with the greatest loss occurring at intermediate biomass levels (Figure 3).DOC loss due to photodegradation was consistent across the biomass gradient for the unburned treatment.Despite the nonlinear trend where DOC loss increased in the burned treatment at the higher plant biomass amounts (> 300 g), the UV-only effect on average DOC losses across the gradient did not significantly differ between burn treatments (Table S7).
Microbial decomposition of DOC occurred more at intermediate plant biomass amounts (100-300 g) with an average >9% reduction in DOC and no significant difference between treatments.The quantity of plant biomass more strongly influenced the rate of decomposition with greater DOC loss that increased with biomass but declined rapidly after a certain threshold (>300 g).In contrast, there was only about an average 1% reduction in DOC at low and high plant biomass levels.The microbial effect on DOC loss, therefore, showed a unimodal relationship along the biomass gradient.
The combined UV and microbes effect showed the greatest change in DOC with an average >13% reduction at intermediate biomass amounts (100-300 g), similar to that of the microbes-only condition.Plant biomass quantity significantly influenced the rate of DOC degradation with no significant difference between the treatments.We found that the R 2 value for the UV and microbes model explained a larger proportion of the variance in DOC loss (R 2 = .463),indicating that this combined effect had a stronger overall impact on DOC degradation compared to the microbes-only model (R 2 = .322).

| Sage and willow decomposition
We calculated percent mass loss for sage and willow as a function of plant biomass (per species) and treatment.Sage decomposed more at the lowest levels of plant addition, and the relationship differed significantly between treatments (p = .03,Figure 4a).Unburned sage exhibited a linear decrease in mass loss with increasing plant biomass.
In contrast, burned sage showed a significant nonlinear trend with the greatest mass loss at the lowest plant biomass amounts.Willow decomposition also differed significantly between treatments (p = .02) with the greatest mass loss occurring in the unburned treatment that remained constant across all plant biomass levels (Table S8).Sage decomposed more than willow, with the greatest difference between species in the burned treatment (p < .0001, Figure 4b).

| DISCUSS ION
We demonstrate that burned and unburned plant biomass loading elicits nonlinear responses on the dynamics of DOM in experimental ponds.First, we found strong nonlinear effects of plant biomass on DOC concentration, with the greatest pulse of DOC soon after plant addition.Smaller loading amounts exhibited lower DOC concentrations, but once a certain biomass threshold was exceeded (>100 g), the release of DOC increased exponentially (Figure 1).
Although the upper threshold of DOC values measured in the ponds exceeds the global average concentrations in natural systems (mean = 7.58 ± 0.19 mg/L), they fall within the range of variability (0.1-332 mg/L, median of 5.71 mg/L; see Sobek et al., 2007).
Experiments with many drivers are essential to produce generalizable and predictive models and to detect nonlinear responses (Collins et al., 2022;Kreyling et al., 2018).Lakes in burned watersheds have around 13.2 mg/L, ranging to maximum concentrations between 30 and 32 mg/L (Allen et al., 2003;McEachern et al., 2000).All ponds fell within these concentration ranges which demonstrates that the DOC levels in our study are comparable to that of natural systems.Moreover, these threshold effects modeled along a gradient of increasing plant biomass further suggest that the dynamics of plant-biomass-DOC interactions are sensitive to subtle changes in ecosystem conditions, and that simple linear relationships may not adequately capture these nuances.Holgerson et al. (2022) found nonlinear relationships between lake metabolism and concentrations of total dissolved phosphorus (TDP) and DOC, suggesting that critical thresholds and internal feedbacks govern ecosystem responses to the input of abiotic resources.Our gradient design allowed us to evaluate how fire and terrestrial loading extremes impact experimental pond ecosystems with effects that may extend to larger inland water bodies like lakes.
Second, we observed differences in DOC concentration between treatments with lower DOC levels in the burned treatment, especially at intermediate biomass amounts (Figure 1).This may be attributed to chemical transformations that occur during combustion which can alter the C and nutrient concentrations in plant tissues (Cornelissen et al., 2017).A companion study (Wall et al., 2023) found that burning sage decreased leaf % nitrogen (N) but increased stem % N and % potassium (K), while burning willow consistently increased % N, % P, and % K (as well as other micronutrients like sulfur and zinc) in both leaves and stems.Variations in burn severity can also lead to the production of volatile compounds, non-dissolvable residues like ash, or solubility changes of certain C compounds (Paul et al., 2022;Wagner et al., 2018).High-severity wildfires tend to result in lower DOC concentrations, while low-severity burns show higher DOC concentrations (Santos et al., 2019), suggesting that burn severity could potentially transform C compounds initially present in plant tissues into less soluble forms, thereby diminishing their contribution to the DOC pool (Burd et al., 2020).These changes parallel similar responses in other plant species to fire (e.g., Betts & Jones, 2009;Bowring et al., 2022;McCullough et al., 2019), highlighting the importance of wildfire severity as a crucial factor affecting stored C in plant tissues with impacts on aquatic DOC concentration.
The fluorescence and UV-visible absorbance indices further elucidate the transformations in DOM chemical composition.We found that the DOM leached 10 days after plant addition largely consisted of LMW, labile compounds as indicated by the higher S R and lower SUVA 254 values, respectively.Fluorescent DOM in water is composed of both stable and labile compounds, with the latter being rapidly consumed by microbes (Cory & Kaplan, 2012).Coupled with the findings from the incubation experiment that showed greater microbial contributions to DOC loss, we observed higher BIX values, and indicators of bacterial metabolism (greater intensities of B and M peaks; Figure S4), indicating that microbes preferentially consumed these labile DOC compounds soon after leaching.After Day 10, recalcitrant DOM began to slowly accumulate in the ponds as evidenced by increases in SUVA 254 and HIX values over time (Figure 2).These findings align with other studies on fire-transformed C where LMW and less aromatic fractions are preferentially mineralized leading to increases in SUVA 254 over time (Bostick et al., 2021;Chen et al., 2022).

Burn severity and terrestrial loading impact DOM composition
with effects that become more pronounced over time.Before plant addition (i.e., Day 0), initial HIX and SUVA 254 values did not significantly differ between treatments or along the biomass gradient (Figure S8).Over time, and especially after Day 59, the differences in DOM humification became more apparent with increasing biomass as indicated by the peak intensities for terrestrial humic-like DOM (peaks A and C; refer to Figures S4-S7), as well as the higher and lower HIX values in the unburned and burned treatments, respectively.Since DOM humification refers to the degree to which DOM is processed (Zsolnay et al., 1999), the unburned treatment may have contained more easily biodegradable compounds leading to greater HIX values, while the burned treatment, with lower HIX values, likely contained higher proportions of recalcitrant C that were not as readily processed (see Chen et al., 2022).Furthermore, higher burning temperatures can form more aromatic and HMW fractions (Chen et al., 2022;Wang et al., 2015), and since we combined the plant biomass of different burn severities into a single burned treatment, this mixture likely produced DOM with a range of solubility characteristics (including both hydrophobic and hydrophilic fractions) that subsequently led to reduced DOM processing in the burned treatment compared to the unburned treatment over time.In order to fully understand how CDC compounds change over time, conducting longer experiments with similar variables is crucial.While our study, along with a related one by Bostick et al. (2021), lasted for approximately 90 days, other studies (e.g., Chen et al., 2022;Wang et al., 2015), were considerably shorter, often lasting less than 60 days.Our study highlights the profound influence of wildfires and their severity on DOM composition in experimental ponds, with time emerging as a critical factor in shaping the bioavailability of CDC compounds.
Third, we found that plant biomass loading regulates DOC degradation more than whether or not it is burned, and that DOC loss across a gradient of increasing biomass is more strongly attributed to microbial activity than photodegradation.The distinct unimodal pattern in DOC loss shows the most significant reduction occurred at intermediate biomass levels (Figure 3).This suggests that critical thresholds, as it relates to the amount of terrestrial loading, regulates DOC processing in aquatic systems.Studies using first-order decay models of microbial DOC decomposition found proportional DOC loss relative to initial concentrations, and that DOC loss increased linearly (Chen et al., 2022;Wilske et al., 2020), a likely to the greatest DOC loss in our incubation experiment.Higher bacterial productivity has also been linked to browner waters, indicating how elevated DOC levels, or the subsequent availability of organic compounds, may impact aquatic microorganismal responses (Robidoux et al., 2015).
Here, we found that a successive photochemical-microbial degradation pathway (i.e., UV and microbes) also contributed to the greatest DOC loss at intermediate biomass amounts (Figure 3).
While we did not detect treatment differences in DOC loss among the microbes-only or UV and microbes conditions, aromatic, humic compounds can moderate bacterial activity through photochemical transformations that yield labile compounds (Cory & Kling, 2018;Dempsey et al., 2020;Moran & Hodson, 1990).For instance, photolysis of aromatic DOM can depolymerize these complex chemical structures into smaller, labile compounds such as LMW organic acids that fuel microbial respiration (Chen et al., 2022;Morris & Hargreaves, 1997;Obernosterer & Benner, 2004;Wetzel, 1984).
This is a likely explanation since the additional UV component in In some cases, linear regressions have been favored for their explanatory power (Sobek et al., 2007) with responses like bacterial productivity and respiration to increase linearly with increasing supply of terrestrial DOM (Lennon & Pfaff, 2005).This complexity underscores the need for comprehensive analyses of the various relationships at play.Understanding post-fire DOM dynamics presents complex challenges, especially when considering how natural variations such as the responses of aquatic microbial communities to terrestrial loading are tightly coupled with nutrient or DOC gradients.Nutrients such as P and N can become liberated during fire and subsequent in situ decomposition (McEachern et al., 2000;Paul et al., 2022;Santos et al., 2019).The companion study (Wall et al., 2023) found that fire altered the nutrient content of both sage and willow, which increased TDP and TDN along the biomass gradient with overall higher TDP and TDN concentrations in the burned treatment.Increased nutrient access (especially TDP) likely enhanced microbial decomposition of DOC at these intermediate biomass levels (see Hanson et al., 2003;Lennon & Pfaff, 2005).
It is also crucial to recognize the significant interactions among nutrients (e.g., N and P), DOC degradation, and other environmental factors like greenhouse gasses and dissolved oxygen when assessing the effects of disturbances such as fire and plant biomass loading on aquatic ecosystems.Nutrient and DOM availability likely enhanced degradation processes, leading to CO 2 supersaturation in the ponds with as little as 50 g of biomass that resulted in larger amounts of greenhouse gas emissions over a shorter time period (approx. 60 days; Wall et al., 2023).At the highest biomass levels (>250 g), increased microbial activity and subsequent oxygen demand led to hypoxia in the ponds, inhibiting microbial respiration (Wall et al., 2023).
Other studies (e.g., Hanson et al., 2003;Mallin et al., 2006)  future studies employing a similar approach should extend the measurement period to gain a better understanding of the potential physio-chemical effects of fire on DOM residence time.Additionally, it is essential to consider other contributing factors such as total area or temperature that can influence DOC concentration variability and subsequent degradation (e.g., Hanson et al., 2011;Sobek et al., 2007).This broader approach will help assess the role of residence time in relation to DOM derived from burned or unburned plant biomass, further elucidating its significance in controlling C cycling in aquatic systems, especially when plant biomass loading exceeds the rate of DOM turnover (Stedmon & Nelson, 2015).
Our study differs from previous work on microbial decomposition of DOC (e.g., Bostick et al., 2021;Hanson et al., 2003;Lennon & Pfaff, 2005;Zhang et al., 2013) since filtering in the incubation experiment likely reduced microbial abundance but not complete exclusion.The GF/F pore size (0.7 μm) may have allowed the passage of some microorganisms.Although the disinfecting effect of UV radiation exposure likely reduced abundance in the photodegradation experiment (see Tranvik & Bertilsson, 2001;Uzun et al., 2020;Yang et al., 2020), our results indicate that filtration consistently decreased DOC decomposition across most plant biomass amounts (Figure 3).
This suggests that filtering microorganisms with a 0.7 μm pore size limited microbial abundance enough to impede decomposition.
Lastly, our findings reveal distinct variations between treatments in the decomposition of willow and sage along the biomass gradient (Figure 4).Willow decomposed less than sage in our study, suggesting an effect of its specific physiology such as its lignin-rich tissues or woody structure which decompose more slowly than leaves (Cornelissen et al., 2017).Wildfire severity may affect plant species differently (Coppoletta et al., 2016;Keeley, 2009;Pellegrini et al., 2015), leading to char production in some species that further reduces decomposition (Bostick et al., 2021).The effect of wildfires on terrestrial plants, however, is a subject of ongoing debate.While wildfires can change how nutrients are distributed within soils, plant community composition and the stoichiometry of individual plant species have remained surprisingly resilient to the effects of burning (Pellegrini et al., 2015).Since we added both the woody and leafy parts of sage and willow plants to the experimental ponds, and observed clear differences in decomposition between the treatments, changes to the traits of these two different plant species as it relates to fire could have significantly affected how quickly they decomposed.The breakdown of burned or unburned particulate organic detritus may then determine the quality of the compounds released from the decomposing material, influencing how they are ultimately degraded (i.e., via photo-or microbial degradation pathways) within aquatic ecosystems.Additionally, we studied the effects of burning and terrestrial loading using only two riparian species, future work should consider the heterogeneity of the surrounding vegetation near a watershed with a framework that includes a greater variety of plant species since plant community composition plays a critical role in the quantity and quality of DOM leached into water (Cornelissen et al., 2017;McCullough et al., 2019;Sobek et al., 2007).
Understanding the relationship between fire, terrestrial loading, and their impacts on aquatic DOM chemistry and C budgets is crucial for developing effective strategies to mitigate climate change and promote ecosystem health.

| CON CLUSIONS AND IMPLIC ATIONS
Plant biomass quantity and transformations from fire affect both DOM chemistry and DOC concentration in surface waters, which in turn can impact its rate of degradation.The novel approach using a gradient of increasing plant biomass provides evidence for critical thresholds on the processing of organic matter as seen in the compositional shifts in the EEMs, trends in the fluorescence and UV-visible absorbance indices, and the incubation experiment.
Together, these results show that plant biomass loading in experi- excluded the last two incubations on Days 31 and 59 due to positive net changes in DOC which indicated potential contamination or methodological issues.The first incubation (initiated on Day 10) is presented and reflects the period showing the greatest changes in DOC.

TA B L E 1
Descriptions of fluorescence and UV-visible absorbance indices for dissolved organic matter (DOM) used in this study.Inversely related to the molecular weight Helms et al. (2008) Humification index HIX Increases with DOM humification Zsolnay et al. (1999) Freshness index BIX Increases with freshly produced DOM Parlanti et al. (2000) Fluorescence index FI Higher values indicate the relative contribution of autochthonous versus allochthonous DOM McKnight et al. (2001) the smooth plant biomass term is a test of deviation from a flat or null function that is constant at 0 over all observed values of the predictor variable (Wood, 2017); deviation from this indicates the presence of a nonlinear pattern.We performed Wald tests to determine the significance of each parametric and smooth term for all fitted models.Here, the figures presented show either two smooths that indicate a significant nonlinear smooth for each treatment, or one global smooth (in black) representing an estimated mean across the covariate plant biomass for both treatments (see Section 3).We used this model selection procedure for all analyses to identify whether plant biomass quantity affected the different response variables, and whether these relationships differed between treatments.All data and code associated with this study are publicly accessible and available for reference at Zenodo (https://zenodo.org/doi/10.5281/zenodo.10093520).
displayed notable temporal variations.The observed variations at earlier time points (Days 10 and 31) can be attributed to initial DOC leaching while changes through time reflect biological and chemical transformations in the ponds.SUVA 254 , an indicator of DOM aromaticity, exhibited a nonlinear relationship along the plant biomass gradient on Day 10.At this early stage, we observed greater aromaticity at intermediate plant biomass levels for both treatments (Figure2a).The decrease or stability of SUVA 254 at very low biomass levels, followed by an increase at intermediate and high biomass levels (125-400 g), suggested an accumulation of aromatic DOM.This increase in SUVA 254 became more pronounced by Day 59, resulting in a significant difference between treatments (p = .001).Notably, we observed lower SUVA 254 values in the burned treatment compared to the unburned treatment at Day 59, as evidenced by the differences in fitted smooth functions (FigureS3a).On average, SUVA 254 increased by approximately 12.4% from Days 10 to 89 across all plant biomass levels and treatments.S R , a proxy for DOC MW, displayed a negative relationship across the plant biomass gradient on Day 10, indicating low MW (LMW) compounds at low to intermediate biomass levels (5-150 g) and higher MW (HMW) compounds with increasing biomass (Figure 2b).Patterns in the S R become more nonlinear such that LMW compounds were now present at low and high biomass levels while intermediate biomass amounts showed high MW compounds over time with no significant differences between treatments.The HIX exhibited a negative relationship with plant biomass at Days 10 and 31 for both treatments, indicating less humified DOM at greater biomass levels (>250 g, Figure 2c).By Day 59, the HIX displayed F I G U R E 1 Dissolved organic carbon (DOC) concentrations across plant biomass (g) at five time points for the burned (red) and unburned (green) treatments.Day 0 shows baseline DOC concentrations before plant addition.Each dot represents a different pond.Plots with a single smooth (global) indicate that a model with one fit line for the burned and unburned treatments best describes the data, while separate smooths indicate that a model with different fits for the two treatments showed best agreement.F I G U R E 2 Changes in (a) SUVA 254 , (b) spectral slope ratio, (c) humification index, (d) freshness index, and (e) fluorescence index across plant biomass (g) for the burned (red) and unburned (green) treatments.Each dot represents a different pond.Plots with a single smooth (global) indicate that a model with one fit line for the burned and unburned treatments best describes the data, while separate smooths indicate that a model with different fits for the two treatments showed best agreement.a nonlinear increase with a significant difference between treatments (p = .0006),signifying greater humification in the unburned treatment, peaking at intermediate biomass levels (200 g).The significant difference between treatments and the nonlinear trend along the biomass gradient persisted through Day 89, as indicated by the differences in the fitted smooth functions (p < .0001, Figure S3c).

F
I G U R E 3 Dissolved organic carbon (DOC) loss as percent change for burned (red) and unburned (green) treatments across plant biomass (g).Each dot represents a different pond.Plots with a single smooth (global) indicate that a model with one fit line for the burned and unburned treatments best describes the data, while separate smooths indicate that a model with different fits for the two treatments showed best agreement.
shown 1.4 times greater DOC than lakes in unaffected(unburned)    watersheds(Allen et al., 2003) with median DOC concentrations F I G U R E 4 Sage and willow percent mass loss across plant biomass (g) for each species at the end of 89 days.(a) Mass loss for sage and willow between burned (red) and unburned (green) treatments.(b) Mass loss between sage (purple) and willow (brown) within treatments.Each dot represents a different pond.Plots with a single global smooth indicate that a model with one fit line for the burned and unburned treatments best describes the data, while separate smooths indicate that a model with different fits for the two treatments showed best agreement.
explanation for why we observed less DOC change at the lowest plant biomass amounts.A second companion study (C.B. Wall & M. Demmel, unpublished data) found nonlinear increases in total microbial cell counts, and microbial alpha diversity with DOC concentration at intermediate plant biomass levels (150-250 g), corresponding the UV and microbes condition may have transformed HMW fractions into LMW compounds, enhancing DOC loss more compared to the microbes-only condition.A limitation of our study is that we measured DOM compositional changes at the pond level, which restricted our ability to precisely assess transformations during the incubations.Future research should aim for finer temporal and spatial scales to better understand the interactive effects of photoand microbial degradation processes.Additionally, depth plays a crucial role in degradation, as lakes with high DOC concentrations limit light penetration, affecting C mineralization at depth(Hanson et al., 2003).Our study did not account for how DOC degradation varies by depth since all of the incubations occurred at 45 cm from the surface, though in natural ecosystems, depth, DOC degradation, and DOM compositional changes are likely to covary, adding complexity to these processes.The patterns in the concentrations of biologically and photochemically degradable DOC reported here show that degradation is dependent on DOC concentration and DOM composition, highlighting the importance of identifying DOM compounds and other environmental factors linked to fire-affected aquatic systems.While our results clearly exhibit nonlinear patterns, it is important to acknowledge the broader context in biogeochemical studies.
mental ponds is positively associated with DOC concentration that stimulates microbial mineralization at intermediate levels of loading but impedes decomposition at higher DOC levels when biological demand depletes dissolved oxygen concentrations.These findings shed light on the factors controlling DOC processing and highlight the need for future studies to consider the importance of DOM chemical composition.Using a large-scale experimental pond design to deliberately manipulate environmental conditions demonstrates its relevance in reflecting ecological changes that can take place in natural systems, such as lakes.The nonlinear changes to the structure and function of aquatic ecosystems due to plant biomass loading and fire affects their capacity to store and process DOM, and their subsequent role in the global C cycle.