Fire retardants are an overlooked source of phosphorus to western US ecosystems

Abstract

Excessive nutrient loading to surface waters endangers drinking water supplies, recreation, aquatic life, and many other water quality endpoints. Unfortunately, concentrations of phosphorus (P) and nitrogen (N) remain high in many US waterbodies and may be increasing in remote, relatively pristine watersheds. Several hypotheses have been advanced to explain this increase, including warming temperatures, increased dust, and wildfire-associated smoke and ash deposition. Notably, nutrients from fire retardants have been heretofore overlooked. For the first time, we estimate P and N inputs from ammonium phosphate-based fire retardants in the western US. Remarkably, when expressed on a per area basis, retardant P and N are applied at rates 4–44x and 0.3–16x greater, respectively, than agriculture fertilizer rates for corn, wheat, and other row crops. Moreover, aggregated across subbasins, retardant P—but not N—is comparable to estimated atmospheric deposition rates. Fire retardants help protect human lives and property, and measures are taken to avoid application directly to waterbodies and riparian areas. Nevertheless, the potential for runoff exists, and, even if this does not occur, nutrients from retardants may alter terrestrial ecosystem productivity and carbon cycling, particularly given their usage is only likely to increase under future climate change.

Graphical Abstract

Introduction

Excessive nutrient loading, particularly phosphorus (P) and nitrogen (N), to the environment causes a myriad of downstream water quality problems, including the formation of harmful algal blooms (HABs) and hypoxic zones^1–4. These in turn negatively impact aquatic life, recreation, property values, and drinking water, among many other endpoints (e.g., see Smith et al)⁵. Eutrophication caused algal blooms can lead to low oxygen conditions and sublethal and lethal impacts to biotic assemblages and aquatic food webs^6–8. Beyond ecological effects, blooms often increase organic matter in sources of drinking water and the potential for disinfection by-product formation with elevated cancer risks^9,10. In addition, certain types of algae release cyanotoxins, with the potential to harm humans, pets, livestock, and wildlife through ingestion or contact^11,12. Thus, excessive nutrient loading, and the subsequent cascading water quality effects, present challenges to both human and animal health, and ecosystem services.

Despite national efforts to reduce nutrient non-point source inputs, nutrient concentrations have remained high in many lakes, rivers, and streams across the contiguous US and have even increased in some locations¹³. Indeed, Stoddard et al¹⁴ found that total P concentrations increased even in relatively undisturbed waterbodies¹⁴, although subsequent national stream and lakes assessments have shown improvements^15,16. Several hypotheses have been advanced to explain higher nutrient concentrations, even in seemingly undisturbed areas, including soil deacidification, warming temperatures, increases in atmospheric deposition of P associated with dust, and an increase in fire mobilization of P^14,17–19. In the latter case, large wildfire frequency and area burned by high severity fire are increasing in the western US due in part to climate change^20,21, as is fire activity in other locations globally^22,23. Wildfires have well documented impacts on nutrient loads to surface water^24–26. Post-fire erosion and leaching increases nutrient loading to waterbodies in burned watersheds following the mortality of the vegetation and removal of the biotic nutrient sink²⁴. There is also the potential for fires to mobilize N and P in smoke and other materials to the atmosphere, where downwind ecosystems may be affected via depositional processes²⁷.

The use of fire retardants—in a sense, the aerial distribution of fertilizers—has been overlooked as a source of nutrients to ecosystems up until now. The main types of fire-fighting chemicals are long-term and pre-treatment retardants, foam suppressants, and water enhancers (see https://www.fs.usda.gov/rm/fire/wfcs/index.php for additional information). Aerially applied long-term retardants are typically applied in front of the fire and contain ammonium phosphate salts. These salts work by providing a protective coating and releasing water and carbon dioxide which cools and suffocates the advancing fire. Fire retardants are intended to be applied away from waterbodies, yet the potential for misapplication exists or subsequent leaching or runoff can occur. Moreover, even if the nutrients fail to reach nearby waterbodies, nutrient additions to vegetation and soils can impact terrestrial ecosystems, including changes in plant diversity^28–30, plant productivity and carbon cycling^31–36, and soil microbial responses^37,38. Paradoxically—and perhaps uniquely as a nutrient source—fire retardants can either add to the problem of mobilization of nutrients directly or indirectly by fire if the area burns despite their use, or conversely, they could reduce nutrient mobilization by fire if they prevent or minimize burning as intended. To fully assess this tradeoff at the landscape scale, an estimate of the nutrient additions of fire retardants is first required.

Here, for the first time, we estimate nutrient additions—specifically P and N—via aerially applied long-term fire retardants to 8-digit subbasin hydrologic units (HUC8s) on National Forest land in the western US. Forest fires are common in and near US Forest Service (USFS) land (Figure 1a, b), and these areas are generally located away from—and upstream of—agricultural and urban areas that receive significant nutrient inputs already. By contrast, these lands often contain relatively pristine waterways, with low nutrient inputs, and therefore the biological systems are particularly vulnerable to exogenous nutrient additions. In this study, we compare our estimates of fire retardant nutrients to fertilizer rates of agricultural row crops on a per area basis and atmospheric deposition during this time period at the subbasin scale¹⁸. Lastly, we discuss the potential implications for vulnerable ecosystems, such as mountain waterbodies, particularly given the likely greater use of these chemicals as climate-driven fire activity increases.

Figure 1:

(A) National Forest boundaries (green) and all western fires from 1984 to 2019 (red) over 1000 acres (∼404.7 ha) in size (Monitoring Trends in Burn Severity) https://www.mtbs.gov with (B) total hectares burned in all western states (red) and hectares burned in National Forests (teal). (C) Total estimated kilograms of phosphorus (P; light purple) and nitrogen (N; dark purple) based on reported volumes of aerially applied fire retardant used by the USFS from 2012 to 2019.

Methods

To estimate fire retardant P and N, we used annual reports of aerial long-term fire retardant use on all National Forests (NFs) for 2012 to 2019 from the US Forest Service (https://www.fs.usda.gov/managing-land/fire/chemicals). From these, we included only NFs (e.g., eliminated National Scenic Areas) overlapping with the 11 most western contiguous states. There were 451 instances of a NF receiving fire-retardants from 2012 to 2019. During the 2012 to 2019 period, eight long-term fire retardants containing ammonium-phosphate salts were fully qualified for fixed-wing and helicopter application by the USFS. Using the product information and safety data sheets available from the USFS Wildland Fire Chemical Systems (https://www.fs.usda.gov/rm/fire/wfcs/), we calculated the kilograms of P per volume of fire retardant using the reported percent P₂O₅ by weight of all eight approved retardants. To estimate N inputs, we used the median value of the given estimated percent composition range for each ammonium phosphate ingredient listed for the four retardants comprised of diammonium and monoammonium phosphates but excluded the ammonium polyphosphates. For example, the fire retardant PhosChek MVP-Fx contains 80–90% monoammonium phosphate by weight so a median value of 85% was used for calculating the kg N of fire retardants applied. Since only the total volumes of retardants were provided in the database and not the specific retardant used, we then averaged across the kg of P and N estimates for each of the eight (or four) approved long-term retardants, using this average value for estimating kg P(N) per m² NF land burned (see Table S1 for details). Since ammonium polyphosphates can have different chemical formulations—depending on the number of orthophosphate monomers and ammonium cations present—and the exact formulation used in Phos-Chek’s ammonium polyphosphates are proprietary information, we are unable to estimate the percent composition of N for these retardants. Every gallon of fire retardant contains approximately 116 g P (s.d. of 16 g) and 47 g N (s.d. of 9 g).

To identify fires within each NF, we overlaid proclaimed NF boundaries (https://data.fs.usda.gov/geodata/edw/datasets.php; with an additional 10 km buffer) with fire boundaries for each year (2012 through 2019) and extracted all fires that intersected a NF. We used the fire boundary dataset from the Monitoring Trends in Burn Severity (MTBS) interagency program (https://www.mtbs.gov) for single fires over 1000 acres (~404.7 hectares) in size, and the historic GeoMac fire perimeter datasets from the National Interagency Fire Center (NIFC) (https://data-nifc.opendata.arcgis.com/; accessed in early 2020) for complex fires and fires smaller than 1000 acres. This combined approach of buffering and multiple fire datasets allowed us to link ~95% of reported fire retardant use in a NF each year to at least one fire event. While aerial fire retardant application rates vary from 0.4 to over 2.5 liters per m² (1 to over 6 gallons per 100 ft²) depending, at least in part, on multiple fuel models (see USFS Final Environmental Impact Statement ³⁹; Table S2) we assumed an even coverage of retardant across area burned, given the lack of fuel load and retardant information tied to specific fires.

We calculated loading rates for P and N for individual NFs per year and aggregated at the HUC8 scale. We did this by summing the overlapping areas (m²) of fires and NFs for each NF and for each identified fire for each year. Summing the overlapping area by NF provided the total area (m²) burned in each NF, allowing for calculation of the loading rate (kg P(N)/m²) unique to each NF each year. Summing by individual fire allowed us to calculate the overlap between the area of each fire and the area of one or more NFs. Since NFs contained differing amounts of area burned and thus loading rates of retardants applied, a fire that burned across two NFs had two different kg P(N) per m² burned values for each NF. We multiplied each NF-specific loading rate by the area each individual fire burned within each NF to calculate the total kg P(N) per fire. We then partitioned the kg P(N) per fire based on proportion overlap of each individual fire within subbasin HUC8s (https://apps.nationalmap.gov/downloader/#/) and then summed by subbasin to estimate the total kg P(N) entering each subbasin. For an individual fire overlapping two subbasins, the area and associated loading rate of the fire was partitioned into the overlapping subbasins accordingly (see Figures S1a, S1b and Table S3a–c for a visual example). These steps were done separately for each year.

For context, we compared the magnitude of nutrient additions from fire retardants to reported agricultural fertilizer inputs and estimated rates of atmospheric deposition. We obtained fertilizer usage for select crops reported during the 2012–2019 period from the USDA’s National Agricultural Statistics Service (https://www.nass.usda.gov/). For comparable fire retardant rates, we used the retardant coverage levels reported in the USFS Final Environmental Impact Statement³⁹ ranging from 0.4 to over 2.5 liters per m² (1 to over 6 gallons per 100 ft²), multiplying the average kg P(N) per gallon of fire retardant by reported coverage levels. These coverage levels are built on fuel models that include information on location and vegetation structure. For the atmospheric deposition comparison, we used the modeled P and N depositional estimates from Sabo et al. 2021 and 2019, respectively, averaging the three years of estimated deposition rates (2002, 2007, 2012). Briefly in these studies, P deposition rates are estimated using “sea salt, dust, biogenic, wildland fire and fossil fuel combustion P emission data” and integrated into a combination aerosol chemistry and climate model. N deposition rates come from the National Atmospheric Deposition Program’s Total Deposition maps (TDep) which combine measured and modeled dry and wet deposition measurements (see Wang et al. and Schwede and Lear and for further details on methodology).^40,41 This average atmospheric deposition value was subsequently multiplied by the number of years a subbasin received fire retardants to approximate the total kg of P(N) each subbasin would have received from deposition over the same time period. These deposition estimates were compared to the summed kg of P(N) from fire retardants for each subbasin as a ratio of kg P(N) from fire retardants to deposition.

Finally, we considered the condition of waterbodies in the western US during this time period. To do this, we used the National Aquatic Resource Surveys (NARS) 2012 and 2017 National Lakes Assessments, and 2013/14 and 2018/2019 National Rivers and Streams Assessments (data available at https://www.epa.gov/national-aquatic-resource-surveys), and binned sampled waterbodies into NARS good, fair, poor categories based on measured total P concentrations and aggregated Omernik level III ecoregions^{15,16,42–45}. We combined this information with the spatial information on fire retardants to assess waterbody condition in areas that received fire retardant and are likely to receive future retardant applications to highlight potential areas and ecosystems vulnerable to eutrophication effects.

All mapping overlays and extractions were performed using R version 4.0.3. Packages used included ggplot2, dplyr, ggpattern, usmap, and sf. All shapefiles were converted to Conus Albers projection (EPSG 5070) before overlaying and extraction and then converted to NAD83 (EPSG 4269) for plotting purposes. All data used are publicly available.

Results

Not surprisingly, fire retardant use generally mirrors fire activity. Volume of retardant applied generally increased from 2012 to 2018, alongside increasing area burned across the western US during this time (Figure 1b and 1c). In 2019, when area burned was markedly lower than previous years, so was retardant use. In total, fire retardant use on National Forest lands was greatest in 2016 and lowest in 2019 (Figure 1c). This resulted in an estimated addition of 2.2 and 0.8 million kilograms of P and 0.9 and 0.3 million kilograms of N applied in those years. Notably, when expressed on a per area basis, recommended fire retardant application rates exceed most reported agricultural application rates (Table 1). For example, across the US, fertilizer inputs average 17 to 73 kg P per hectare for wheat and corn, respectively with a crop average of 32 kg P, while input from long-term fire retardants ranged from 125 kg P to nearly 750 kg P per hectare. This is also the case for N, where fire retardant inputs typically exceed those of most agricultural crops (Table 1).

Table 1.

Fire retardant and agriculture P and N input rates

Coverage Level	Fire Retardant Input			Agriculture Input (US Average)
(L/m2)	kgP/ha	kgN/ha	CROP	kgP/ha	kgN/ha
0.4	125	51	corn	32	164
0.8	250	102	cotton	21	97
1.2	374	153	soybean	25	19
1.6	499	203	wheat	17	81
2.5	749	305	all (ave)	24	90

When expressed on a subbasin scale, fire retardant P inputs are comparable to those of atmospheric deposition in many subbasins in the western US, while N inputs from retardants are considerably lower. Over the eight years of available data, 413 subbasins (HUC8s) in the western US received at least one year of P and N input from fire retardants (52% of western subbasins, 61% of western land area) (Figure 2a). The median 8-year combined input into a subbasin was approximately 8.3 Mg P (3.4 Mg N) and the mean was 29 Mg P (12 Mg N). The highest estimated P(N) input to a single western subbasin was 498,522 kg (200,986 kg) and overlapped with the northern portion of Los Padres National Forest, near Santa Barbara, CA, and occurred during 2016 (HUC 18060006). When comparing P from fire retardants to deposition, 4% of subbasins (16 of 413 subbasins) received two times or more P from retardants than from atmospheric deposition, while an additional 16% of subbasins (66) received fire retardant P comparable to deposition (ratio between 0.5 and 2) (Figure 2b). The average P and N addition from fire retardants to these 413 subbasins are approximately 29 Mg and 12 Mg, respectively, while the average P and N additions from estimated atmospheric deposition are 88 Mg and 6,350 Mg, respectively, resulting in atmospheric deposition inputs being 3 times higher for P and over 540 times higher for N than retardant inputs on average. Consequently, the additional N from fire retardants is a comparatively small portion of N deposition (~0.2%), while the addition of P from fire retardants represents a much greater proportion relative to estimated P deposition (~30%).

Figure 2:

(A) Eight-year total (2012−2019) estimated kg fire retardant P per subbasin (HUC8) and (B) ratio of total kg P from fire retardants to kg P from deposition per subbasin (HUC8) from 2012 to 2019. Darkest subbasins received more than two times the kg P from fire retardants compared to atmospheric deposition (i.e., greater than a 2:1 ratio), while the lightest subbasins received less than half of P from fire retardants compared to atmospheric deposition (i.e., less than a 1:2 ratio). Intermediate subbasins represent those that received fire retardant P roughly equal to atmospheric deposition (between 1:2 and 2:1).

Finally, our overlay with NARS assessment data shows that subbasins with fire retardant use also contain the majority of good condition waterbodies in the western US. According to the 2012 and 2017 National Lakes Assessments and the 2013/2014 and 2018/2019 National Rivers and Streams Assessments, only 40% of lake, river, and stream samples can be classified as being in good condition based on their P concentrations in the western US, with 24% in fair and 36% in poor condition. The “Western Mountains” ecoregion—one of four western US ecoregions used by the EPA National Aquatic Resource Surveys and aggregated based on shared environmental characteristics^42,43,46—covers only one third of western land yet half (56%) of all good condition samples came from waterbodies within this ecoregion. This mountain ecoregion overlaps over 85% of NF lands—lands that are receiving fire retardant applications—and consequently, the majority of good, fair, and poor condition water samples taken within this ecoregion also fall within a subbasin that we estimate to have received fire retardant at least once over the eight years examined (Figure 3).

Figure 3:

National Forest boundaries (light green) and aggregated Omernik III ecoregions representing Western Mountains (dark green). Blue circles (lakes) and squares (rivers and streams) represent waterbody samples in “good” condition from the 2012 and 2017 (lakes) and 2013/2014 and 2018/2019 (streams and rivers) National Aquatic Resource Surveys based on their P concentrations.

Discussion

To our knowledge, this study is the first to assess nutrient inputs from fire retardants over a large geographic region, such as the western US. Our results show that fire retardant P and N are added at very high rates onto forest lands, even exceeding those of agricultural fertilizers when considered on a one-time, individual area basis (e.g., kg N or P ha⁻¹). Unlike agricultural inputs, however, fire retardants are not added to the same location year after year, but rather are patchily distributed over time and space. For this reason, it is necessary to consider the cumulative inputs from fire retardants over larger spatial scales and multiple years. At the subbasin (HUC8) scale, we found fire retardant P is comparable in magnitude to the inputs from atmospheric deposition in many areas across the western US, while N additions from fire retardant are comparably small (on average ~0.2% of N deposition). Inputs of P to these predominantly forested systems raise water quality concerns given an estimated half of surface drinking water supplies in the western US come from forested lands⁴⁷. While fire retardant derived nutrients may not enter waterways directly—in fact the USFS follows established protocols in a concerted effort to ideally eliminate direct additions (see https://www.fs.usda.gov/managing-land/fire/chemicals for relevant policy and guidance documents)—there is the potential for subsequent runoff and effects on downstream water quality. This is especially likely if, despite the fire retardant use, the area burns, resulting in higher runoff and soil erosion to waterbodies.

Among aquatic ecosystems with the potential to be most impacted by fire retardants are higher elevation lakes and headwaters, often oligotrophic systems, where a small change in nutrient inputs can profoundly shift the water quality and ecology of these waterbodies^48–52. Our findings regarding fire retardant P inputs are particularly salient given concerns about reported increases in P concentrations in otherwise relatively undisturbed waterways in the region¹⁴. Catchment characteristics largely determine the sources and total inputs of P, and P inputs to high elevation lakes and streams in forested catchments are typically dominated by atmospheric deposition in contrast to waterbodies within downstream, lower elevation catchments surrounded by agriculture or urban areas.^17,53 Notably, there is substantial overlap with areas containing “good condition” waterbodies in the NARS survey and fire retardant usage. Within the western US, 40% of lake, river and stream samples are in good condition and over half (56%) fall within the Western Mountains aggregated ecoregion classification—an area that covers only one third of western land area but encompasses over 85% of National Forest lands and therefore are within the same areas receiving fire retardants. It may be argued these water bodies are still in good condition despite being located within the same areas receiving fire retardant inputs. However, it is often easier to protect good conditions rather than restore poor or degraded ecosystems, and thus there is a need to better understand the increasing risks to these systems. Recent work by Sabo and others found climate variability (monthly and annual temperature), erosional losses of legacy sources, and increased contemporary inputs determine total P concentrations in lakes and streams¹⁷. Fire retardant P is comparable to the largest anthropogenic P source (deposition) accounted for in these mountainous aquatic systems and is likely contributing to soil P and N pools with the potential to impact water quality both currently and in the future with subsequent runoff, leaching, and mobilization. At the same time, temperature and fire risk are only predicted to rise, making it likely that fire retardant usage will also increase in future years as well. Given these stressors and others, the protection of these relatively pristine systems is likely to become increasingly more challenging in coming decades.

Eutrophication of these aquatic ecosystems can lead to a myriad of undesirable outcomes for both aquatic life and human health^5–10. The increase in P could be problematic for algal bloom formation, particularly in combination with higher water temperatures accompanying climate change^17,54. Cyanobacteria can access organic P more readily than other phytoplankton and, consequently, can bloom at lower nutrient levels, such as those found within oligotrophic waters⁵⁵. A recent study examining large lakes across the contiguous US found only 4% of lakes showed an increase in cyanobacterial bloom magnitude between sampling years, yet most of those lakes are in the western US due to the hotter and drier growing season conditions⁵⁶. Wildfires typically occur at times of elevated air temperatures, often coinciding with higher water temperatures as well. A recent study suggested that the addition of nutrients via fire mobilized aerosols could lead to cyanobacteria blooms in downwind lakes²⁷. Adding nutrients from fire retardants could be an additional factor contributing to blooms due to nutrient runoff and leaching, air movement of retardant-enriched dust, or both. Warming temperatures, and changes in precipitation intensity under future climate change are likely to exacerbate this dynamic¹⁷. Other fire-retardant components besides P could also contribute to eutrophication if they reach a receiving waterbody. While the subbasin total of N from fire retardants is far less than N deposition, fire retardants contain N in the form of ammonium, which can in turn increase P release from lake sediments^57,58, resulting in greater total P concentrations in the water column in comparison to P release due to nitrate loading⁵⁹. This indicates the N from fire retardants may have a potentially larger impact on aquatic conditions than just the total fire retardant N amount may otherwise suggest. Furthermore, some fire retardants contain iron (Fe) oxides and recent HAB research has found evidence of P and Fe co-limitation by N-fixing cyanobacteria in stream habitats⁶⁰.

Beyond effects on waterbodies, there also remains a potential for impacts to terrestrial ecosystems directly. The consensus has often been that aquatic ecosystems are primarily P-limited while terrestrial systems are primarily N-limited. Recent research, however, shows terrestrial systems across the globe—not just in the tropics—can be P-limited^61,62 while some aquatic systems are becoming increasingly P-limited due, in part, to decades of N deposition^63,64. While there are few studies examining nutrient effects of fire retardants on above and belowground plant and microbial communities, the results are similar to those of conventional fertilizers. Increases in plant growth as well as changes to community structure have been documented^65,66 including loss of or reductions in cover of some native species with a concurrent increase in non-natives^67,68. Belowground, increases in bacterial growth and decreases in fungal growth have been observed even after ten years post-treatment⁶⁹. Persistent changes in soil chemistry have also been documented⁷⁰. Since above and belowground communities and their associated functioning are linked, a better understanding of how and for how long retardant additions, with co-occurring fire or no fire, may alter or interrupt these feedbacks could improve post-recovery efforts.

Human alteration of ecosystem nutrient budgets is a recognized and ongoing issue since the mid-20^th century. Although the total area receiving fire retardants is considerably less than the area under crop production, fire retardants are applied at higher concentrations and the area receiving retardants can be expected to increase in step with projected changes in fire frequency and severity. While a recent study estimated fire suppressants, including long-term retardants, added some 38 Mg of toxic metals to the US landscape over a period of twelve years⁷¹, fire retardant use is essential to protect human life and property. Publicly available and accessible detailed reporting of where, when, which type, and how much fire retardant is used is key to improving large-scale assessments of potential retardant impacts. Furthermore, empirical lab and field studies are needed to determine the bioavailability and fate and transport of these compounds and any environmental or water quality effects to which they may contribute. Such studies would include field experiments to measure mobility, retention, and transformation of retardant-derived nutrients within soils, and monitoring of water quality metrics post-retardant use if specific location data on retardant drops were recorded and available.

As in all studies, there are limitations that should be noted. The fire retardant data available provide volume of retardant used in a particular National Forest but do not specify dates within a year or another way to link usage to specific fires. Instead, our estimate simply assigns retardant use based on recorded area burned per year. However, fires in remote regions of National Forests farther away from human habitation may reach a larger final burn size because resources (personnel and retardant use) were focused in areas critical to protecting life and property. Moreover, retardants are not dropped on the landscape in an even coating but at varying rates depending on fuel types, as well as type of aircraft used, drop height and wind speed.

Conclusions

This is the first study to estimate nutrient additions from fire retardants across a large region, specifically across the western US. Overall, we found that P and N from fire retardants are added at rates higher than those of agricultural inputs when expressed on a per area basis, and fire retardant P input is comparable to or greater than atmospheric deposition in a fifth of western US subbasins. During recent decades, the frequency of large fires has increased in the region ^21,72 and this in turn raises the potential for more fire, including in the wildland-urban interface. The use of fire retardants is essential to protect public safety and property, and thus nutrient additions are likely a necessary side effect of fire retardant usage now and into the future. Nevertheless, resource managers should be aware of the potential for nutrient effects after the use of retardant, beyond the effects of the nutrients mobilized by the fire itself. Researchers could also focus on better understanding how retardant-derived nutrients are retained or lost after application, and the potential consequences of these nutrients to eutrophication of systems, such as a possible connection to algal blooms. By doing so, resource managers could better anticipate the potential impacts of the use of fire retardants as they adapt current resource protection strategies to meet the challenges of climate change.

Supplementary Material

Additional information and visual examples for fire retardant calculations (docx).

Synopsis.

Fire retardant nutrient additions exceed those of agriculture on a per area basis, and retardant phosphorus is comparable to atmospheric deposition at the subbasin scale in the western US.