Refining the cheatgrass-fire cycle in the Great Basin: Precipitation timing and fine fuel composition predict wildfire trends

Abstract

Larger, more frequent wildfires in arid and semi-arid ecosystems have been associated with invasion by non-native annual grasses, yet a complete understanding of fine fuel development and subsequent wildfire trends is lacking. We investigated the complex relationships among weather, fine fuels, and fire in the Great Basin, USA. We first modeled the annual and time-lagged effects of precipitation and temperature on herbaceous vegetation cover and litter accumulation over a 26-year period in the northern Great Basin. We then modeled how these fine fuels and weather patterns influence subsequent wildfires. We found that cheatgrass cover increased in years with higher precipitation and especially when one of the previous 3 years also was particularly wet. Cover of non-native forbs and native herbs also increased in wet years, but only after several dry years. The area burned by wildfire in a given year was mostly associated with native herb and non-native forb cover, whereas cheatgrass mainly influenced area burned in the form of litter derived from previous years' growth. Consequently, multiyear weather patterns, including precipitation in the previous 1-3 years, was a strong predictor of wildfire in a given year because of the time needed to develop these fine fuel loads. The strong relationship between precipitation and wildfire allowed us to expand our inference to 10,162 wildfires across the entire Great Basin over a 35-year period from 1980 to 2014. Our results suggest that the region's precipitation pattern of consecutive wet years followed by consecutive dry years results in a cycle of fuel accumulation followed by weather conditions that increase the probability of wildfire events in the year when the cycle transitions from wet to dry. These patterns varied regionally but were strong enough to allow us to model annual wildfire risk across the Great Basin based on precipitation alone.

Keywords: Bromus tectorum; annual forb; climate; litter; sagebrush shrublands; weather.

Figure 1

Conceptual model of hypothesized relationships among weather, fine fuel, and fire. Fine fuel is characterized as herbaceous non‐native and native vegetation, and litter. Litter is shown as dark brown plant matter lying horizontally, which can persist across years. Numbered circles correspond to a subset of our study objectives (1a–d) examined at a focal study area within the northern Great Basin. Objective 1a refers to the relationship between weather and herbaceous fine fuel, 1b refers to the relationship between herbaceous plant cover in previous years and litter cover in a focal year, 1c refers to the relationship between herbaceous fine fuel and fire, and 1d refers to the relationship between weather and fire. Objectives 2 and 3 (not shown) were examined across the entire Great Basin, with Objective 2 addressing the relationship between weather and fire at the Great Basin‐scale, and Objective 3 assessing our ability to use this relationship to predict and forecast relative fire risk across this area

Figure 2

Focal and Great Basin study area maps. Historic fire polygons are shown in gray. The focal study area is located at the Morley Nelson Snake River Birds of Prey National Conservation Area located within the Snake River Plains Major Land Resource Area (MLRA) of the northern Great Basin. Long‐term vegetation plots are shown. The Great Basin study area is delineated as three Ecoregions (see text) and subdivided by Major Land Resource Areas (MLRA) numbered: (1) Malheur High Plateau, (2) Central Rocky and Blue Mountain Foothills, (3) Owyhee High Plateau, (4) Snake River Plains, (5) Eastern Idaho Plateaus, (6) Klamath and Shasta Valleys and Basins, (7) Fallon‐Lovelock Area, (8) Humboldt Area, (9) Central Nevada Basin and Range, (10) Great Salt Lake Area, (11) Wasatch and Uinta Mountains, (12) Sierra Nevada Mountains, (13) Carson Basin and Mountains, and (14) Southern Nevada Basin and Range. The location of the focal study area is shown for reference (red polygon)

Figure 3

Modeled relationships between precipitation (mm) and (a) cheatgrass, (b) non‐native forb, (c) native herb, and (d) herbaceous litter cover from 57 plots sampled annually 1989–2014 in sagebrush ecological sites at the Morley Nelson Snake River Birds of Prey National Conservation Area in southwestern Idaho. Negative values of delta precipitation variables (e.g., ΔWIN + SPR2yrPave) indicate seasons that were drier than corresponding seasons in previous years. Panels at right show observed (black lines) and model estimated (red lines) cover values (primary y‐axes) and observed values of the most influential precipitation variable from each model (blue lines; secondary y‐axes) through time

Figure 4

(a, b) Modeled relationships between plant cover in previous years and litter cover from 57 plots sampled annually 1992–2014 in sagebrush ecological sites at the Morley Nelson Snake River Birds of Prey National Conservation Area in southwestern Idaho. Data from 1989 to 1991 could not be included because this analysis required plant cover data collected 1–3 years prior to litter cover data. (c) Observed (black line) and model estimated (red line) cover values (primary y‐axis) and observed values of the most influential vegetation variable (brown line; secondary y‐axis) through time

Figure 5

(a, b) Modeled relationships between fine fuels and area burned annually in sagebrush ecological sites at the Morley Nelson Snake River Birds of Prey National Conservation Area in southwestern Idaho. Fine fuels were characterized from vegetation data collected from 57 plots sampled annually 1989–2014. Gray areas are regions of predictor space with too few observations to make reliable estimates of area burned. (c) Observed (black line) and model estimated (red line) cover values (primary y‐axis) and observed values of the most influential vegetation variable (brown line; secondary y‐axis) through time

Figure 6

(a) Modeled relationship between fine fuels in sagebrush ecological sites at the Morley Nelson Snake River Birds of Prey National Conservation Area in southwestern Idaho and uncharacteristically large fire years (BIGYR1SD, see text for definition). Fine fuels were characterized from vegetation data collected from 57 plots sampled annually 1989–2014. (b) Observed large fire years (black circles) and model estimated (red line) probability of large fire year (primary y‐axis), along with observed values of the most influential vegetation variable (brown line; secondary y‐axis) through time

Figure 7

(a) Modeled relationships between precipitation (mm) and the number of fires occurring annually at the Morley Nelson Snake River Birds of Prey National Conservation Area in southwestern Idaho, 1989–2014. (b) Observed (black line) and model estimated (red line) number of fires (primary y‐axis) and observed values of the most influential precipitation variable (blue line; secondary y‐axis) through time

Figure 8

Within the Great Basin, average winter (October–March), spring (April–June), and summer (July–September) precipitation (primary y‐axis) and total area burned by wildfires (ha * 1,000; secondary y‐axis) from 1980 to 2014. Upper panel shows the number of reported fires in the Great Basin annually

Figure 9

(a) Modeled relationship between precipitation anomaly (percent of average) and the number of fires occurring annually within MLRAs of the Great Basin from 1980 to 2014. Negative values of ΔSUMMER3yrPave indicate summers that were drier than the average of the previous three summers. Negative y‐axis values occur due to a high rate of change in the response variable in that region of predictor space. (b) Observed (black line) and model estimated (red line) number of fires across the Great Basin (primary y‐axis) and observed values of the most influential precipitation variable (blue line; secondary y‐axis) through time

Figure 10

(a, b) Modeled relationship between precipitation anomaly (percent of average) and area burned annually within MLRAs of the Great Basin from 1980 to 2014. Negative values of ΔSPRING1yrP indicate springs that were drier than the previous spring. Negative y‐axis values occur due to a high rate of change in the response variable in that region of predictor space. (c) Observed (black line) and model estimated (red line) area burned across the Great Basin (primary y‐axis) and observed values of the most influential precipitation variable (blue line; secondary y‐axis) through time

Figure 11

(a) Modeled relationship between precipitation anomaly (percent of average) and uncharacteristically large fire years (BIGYR1SD, see text for definition) in the Great Basin. Negative values of ΔSPRING1yrP indicate springs that were drier than the previous spring. (b) Observed large fire years (black circles) and model estimated (red line) probability of large fire year (primary y‐axis), along with observed values of the most influential precipitation variable (blue line; secondary y‐axis) through time

Figure 12

Model predicted elevated fire risk (warmer colors) relative to background levels (cooler colors) for three recent years. Observed fire boundaries (black polygons) are shown for each year. White regions of the Great Basin maps had combinations of predictor variable values that were too rare for reliable predictions of fire risk to be made. Panels at right show, for each year, the probability density of fire risk values for all burned (black lines) and unburned (gray lines) pixels in the Great Basin (n = 1.2 million 800‐m pixels total). All three years are plotted on the same x‐axis scale for comparison of distributions across time. Blue lines are reference bands indicating the region that the black and gray lines would both occupy if their distributions were not different according to a randomization test produced through 100 bootstrap runs with replacement

Figure A1

Great Basin study area showing (a) 1980–2014 average annual precipitation, (b) potential vegetation types reclassified from LANDFIRE (see Table A1), (c) areas dominated by introduced annual herbaceous vegetation (IAH) according to LANDFIRE. Major Land Resource Area (MLRA) codes in the lower right panel are: (1) Malheur High Plateau, (2) Central Rocky and Blue Mountain Foothills, (3) Owyhee High Plateau, (4) Snake River Plains, (5) Eastern Idaho Plateaus, (6) Klamath and Shasta Valleys and Basins, (7) Fallon‐Lovelock Area, (8) Humboldt Area, (9) Central Nevada Basin and Range, (10) Great Salt Lake Area, (11) Wasatch and Uinta Mountains, (12) Sierra Nevada Mountains, (13) Carson Basin and Mountains, (14) Southern Nevada Basin and Range, (d) fires burned from 1980 to 2014

Figure A2A

Observed (black lines) and model estimated (red lines) number of fires for each year by MLRA. Model estimates are derived from precipitation variables in the Great Basin‐level model. Note different y‐axes scales

Figure A2B

Observed (black lines) and model estimated (red lines) number of fires for each year by MLRA. Model estimates are derived from precipitation variables in the Great Basin‐level model. Note different y‐axes scales

Figure A3A

Observed (black lines) and model estimated (red lines) area burned for each year by MLRA. Model estimates are derived from precipitation variables in the Great Basin‐level model. Note different y‐axes scales

Figure A3B

Observed (black lines) and model estimated (red lines) area burned for each year by MLRA. Model estimates are derived from precipitation variables in the Great Basin‐level model. Note different y‐axes scales

Figure A4

Model estimated area burned versus one‐year change in spring precipitation for four representative MLRAs. Each point is one year (1980–2014) in a given MLRA. Model estimates are derived from precipitation variables in the Great Basin‐level model

Figure A5

Modeled relationships between precipitation and annual relative fire risk across the Great Basin from 1980 to 2014. The best fitting model contained three precipitation predictor variables: ΔSUMMER3yrPave, SPRING1yrP, and ΔWIN + SPR3yrPave. Panel (a) shows that fire risk is highest the year following wet springs and particularly when summers are dry (negative values) compared to the three years prior. Panel (b) shows that fire risk is highest the year following wet springs and particularly when winter and spring precipitation in the given year is lower than the three years prior