Incorporating Anthropogenic Influences into Fire Probability Models: Effects of Human Activity and Climate Change on Fire Activity in California

Anthropogenic influences account for up to fifty percent of explanatory power in the model historically.

“Historically, we find that anthropogenic influences account for up to fifty percent of explanatory power in the model.”

Model 1 does a reasonable job estimating fire area over California.

“The climate only specification (Model 1) does a reasonable job in estimating fire area over California (Table 2), but the inclusion of human drivers of wildfire (Model 2) significantly improves model performance for all statistical indicators (AIC, Log-Likelihood) as well as prediction error for the 1951–1975 and 1976–2000 periods (Table 2).”

Inclusion of human drivers (Model 2) significantly improves model performance for all statistical indicators (AIC, Log-Likelihood) and prediction error for 1951-1975 and 1976-2000.

“The climate only specification (Model 1) does a reasonable job in estimating fire area over California (Table 2), but the inclusion of human drivers of wildfire (Model 2) significantly improves model performance for all statistical indicators (AIC, Log-Likelihood) as well as prediction error for the 1951–1975 and 1976–2000 periods (Table 2).”

We can reject the null hypothesis that the data are more likely described by the Poisson distribution (p<0.01), supporting Negative Binomial regression.

“We find that we can reject the null hypothesis that the data are more likely described by the Poisson distribution (p<0.01), strongly supporting the use of Negative Binomial regression.”

We reject equality of the full model and null model (p<0.01) in favor of the full model.

“Finally, we also test model performance against the null model (intercepts only) and reject equality of the models (p<0.01) in favor of the full model.”

Same results were true for suite of possible indicators of variability for AET, precipitation, and maximum temperature.

“These same results were also true for the suite of possible indicators of variability for AET, precipitation, and maximum temperature (results not shown).”

The omission of anthropogenic variables in our model would overstate the importance of climatic ones by at least 24%.

“The omission of anthropogenic variables in our model would overstate the importance of climatic ones by at least 24%.”

• AssumptionRests on an assumption: The relationships between fire and its predictors estimated from 1976–2000 remain stationary and valid for forecasting 2001–2050.

We can reject the null hypothesis that zero-inflated and non-inflated models are equally close in favor of zero-inflation (p<0.01).

“We find that we can reject the null hypothesis that the two models are equally close to the actual model in favor of zero-inflation, which is a superior model (p<0.01) [52].”

• cites [52] Q VH, 1989, Likelihood Ratio Tests for Model Selection and non-nested Hypotheses, Econometrica sourceNot checked yet

  Not checked in the free run.

Omitting anthropogenic influences significantly overstates the role of lightning as an ignition source.

“For example, comparing models 1 and 5 demonstrates that omitting anthropogenic influences significantly overstates the role of lightning as an ignition source, and increases the size coefficients for climate-relevant variables (aetAave and cwdAave) by 24% on average.”

We do not find evidence that inclusion of measures of inter-annual variability in climatic water deficit, precipitation, or temperature statistically improves model performance.

“Despite the paramount role of water and energy balances variation for fire occurrence, we do not find evidence that inclusion of measures of inter-annual variability in climatic water deficit, precipitation, or temperature statistically improves model performance.”

The effect of neighboring housing density (DenMax25km) on fire likelihood is non-linear.

“A similar non-linear fit applies to the effects of neighboring housing density (DenMax25km).”

Total area burned is expected to increase for both climates (PCM, GFDL) and time periods (2001-2025, 2026-2050).

“Our analysis reveals that total area burned is expected to increase for both climates (PCM, GFDL) and time periods (2001–2025, 2026–2050) (Fig 4).”

Between 1951-1975 and 1976-2000, there was a decline in mFRI (increase in fire frequency) across areas of coastal influence in southwestern California.

“Between the two periods we estimate a decline in mFRI (i.e., an increase in fire frequency) across areas of coastal influence in southwestern California by 1976–2000.”

In the second forecast period (2026-2050), GFDL simulations show higher fire activity than the PCM model.

“In the second forecast period (2026–2050), GFDL simulations show higher fire activity than the PCM model (Fig 5B, Fig 6B).”

Our findings suggest, on average, higher increases in the number of fires across public lands.

“Finally, our findings also suggest, on average, higher increases in the number of fires across public lands.”

Burned area is expected to increase by 2.2% and 5.0% by 2050 for the PCM and GFDL models, respectively.

“We find that burned area is expected to increase by 2.2% and 5.0% by 2050 for the PCM and GFDL models, respectively.”

The two climate models forecast a significant reduction in mFRI for the coastal influence zones of the northwestern part of the state, the higher elevations of the Sierra Nevada, and the Modoc Plateau.

“The two climate models forecast a significant reduction in mFRI for the coastal influence zones of the northwestern part of the state, the higher elevations of the Sierra Nevada, and the Modoc Plateau.”

On average, there are more fires in public lands than on private lands throughout California.

“Finally, the PublicLand variable indicates that, on average, there are more fires in public lands than on private lands throughout the state.”

Our models indicate that fire activity may decline for many of California's central and north coastal zones by midcentury.

“Our models also indicate that fire activity may decline for many of California’s central and north coastal zones by midcentury.”

Our finding of moderate increases in the number of fires in the two forecast periods evaluated (2001-2025, 2026-2050) is in agreement with Lenihan et al [11], who forecast relatively small increases (on the order of 4-7%) to burned area.

“Our finding of a moderate increases in the number of fires in the two forecast periods evaluated (2001–2025, 2026–2050) are in agreement with Lenihan et al [11], who use dynamic global vegetation models (DGVMs) to forecast relatively small increases (on the order of 4–7%) to burned area but with broad shifts in fire regime and plant community patterns.”

• cites [11] JM Lenihan, 2008, Response of vegetation distribution, ecosystem productivity, and fire to climate change scenarios for California, Clim Change sourceNot checked yet

  Not checked in the free run.

Urban areas of central and southwest coasts have strong suppressive effects close to their centers.

“We can also see that the urban areas of central and southwest coasts have strong suppressive effects close to their centers.”

In many areas such as the WUI of Southern California, anthropogenic variables explain around half of the total fire count.

“We find, however, that in many areas such as the WUI of Southern California, anthropogenic variables explain around half of the total fire count (Fig 3A and 3B).”

The non-linear fit of CWD and AET indicates that the number of fires initially increases with higher levels of AET or CWD but then decreases at very high levels of either.

“As expected, the non-linear fit of CWD and AET suggest that the number of fires initially increases with higher levels of AET or CWD but then decreases at very high levels of either (Fig A in S1 File).”

Model estimates capture the decreasing mFRI (increasing fire frequency) throughout coastal southern California and the Sierra Nevada between 1951 and 2000.

“For instance, model estimates capture the decreasing mFRI (increasing fire frequency) throughout coastal southern California and the Sierra Nevada between 1951 and 2000 [12,53] (Fig 2AB).”

• cites [12] AL Westerling, 2008, Climate change and wildfire in California, Clim Change sourceNot checked yet

  Not checked in the free run.

• cites [53] JE Keeley, 2001, Historic Fire Regime in Southern California Shrublands, Conserv Biol sourceNot checked yet

  Not checked in the free run.

Remaining fire counts (areas burned more than once) are over- or underestimated by less than 0.5% of actual count for both periods.

“The remaining fire counts (i.e., areas that burned more than once) are over- or underestimated by less than 0.5% of the actual count for both the estimation period and the previous period.”

The influences of climate and terrain contribute the most to fire counts on the southern coast and in the southwestern Sierra, roughly tracing the current distribution of chaparral plant communities.

“The influences of climate and terrain (Fig 3A) contribute the most to fire counts on the southern coast and in the southwestern Sierra, roughly tracing the current distribution of chaparral plant communities.”

Model forecasts suggest substantial agreement about spatial and temporal variation in future fire activity in California by mid-21st century, irrespective of specifications of future climate and human activities.

“Model forecasts suggest substantial agreement about spatial and temporal variation in future fire activity in California by mid-21st century, irrespective of specifications of future climate and human activities (Figs 5A–7B).”

Across model specifications, non-linear relationships between fire and each predicting variable were established through both theory and indicators of model fit.

“Across model specifications, non-linear relationships between fire and each predicting variable were established through both theory and indicators of model fit.”

Many fire prone areas in and around Los Angeles and San Bernardino National Forests experienced an estimated reduction in mFRI of 50%.

“For instance, many fire prone areas in and around Los Angeles and San Bernardino National Forests experienced an estimated reduction in mFRI of 50%.”

• Also stated in: § Anthropogenic influences on Fire · hedged

  “For example, the model estimates indicate that many fire prone areas in and around San Bernardino National Forests in southern California experienced a reduction in mFRI of 50% (more frequent fire) from 1951–1975 to 1976–2000.”

With greater evapotranspiration and declining levels of available moisture under both future climates examined, our model suggests dramatic decreases in the number of years between fires for many high elevation zones.

“With greater evapotranspiration and declining levels of available moisture under both future climates examined, our model suggests dramatic decreases in the number of years between fires for many of these high elevation zones.”

Under PCM climate projections, the model forecasts a higher increase in the number of expected fires for the 2001-2025 period than GFDL over most of California, especially through much of the south and the Sierra Nevada.

“Under PCM climate projections the model forecasts a higher increase in the number of expected fires for the 2001–2025 period than GFDL over most of California (Fig 6A). This is especially true through much of the south and also the Sierra Nevada.”

Our model is able to simulate that suppression is less effective in drier regions subject to Santa Ana winds than in wetter climes such as in the San Francisco Bay Area.

“Our model is able to simulate that suppression is less effective in drier regions subject to Santa Ana winds than in wetter climes such as in the San Francisco Bay Area, which are also subject to dry but less frequent easterly winds.”

The GFDL model estimates a reduction in the time between fires for many urban areas including Los Angeles, San Diego, and other coastal cities of the southwestern part and the Bay Area.

“The GDFL model also estimates a reduction in the time between fires for many urban areas including Los Angeles, San Diego, and other coastal cities of the southwestern part and the Bay Area (Fig 5B).”

Although the two climate models represent very different climate realizations, their outcomes agree in many regions.

“Although the two climate models considered represent very different climate realizations their outcomes agree in many regions (Fig 7A).”

In these locations, diverging future climates show dramatically different fire outcomes with mFRI change between -150 to +100 years.

“In these locations diverging future climates show dramatically different fire outcomes with mFRI change between -150 to +100 years (Figs 5B and 6B).”

The model suggests declining fire counts in coastal influence areas between approximately 33.5-39° latitude.

“For instance, the model suggests declining fire counts in coastal influence areas between approximately 33.5–39° latitude (Fig 6A and 6B).”

Spatial heterogeneity in fire's response to anthropogenic influences between ecoregions and between urban and rural communities is complex.

“Overall, our results indicate that spatial heterogeneity in fire’s response to anthropogenic influences between ecoregions and between urban and rural communities is complex [1] (Fig 3B).”

• cites [1] AD Syphard, 2007, Human influence on California fire regimes, Ecol Appl sourceNot checked yet

  Not checked in the free run.

Areas with the shortest mFRI follow patterns of chaparral and oak woodlands due to their favorable wildfire conditions.

“Areas currently with the shortest mFRI follow patterns of chaparral and oak woodlands due to their favorable wildfire conditions (Fig 2A).”

Across all landscapes, the number of fires increases with proximity to public and private campsites.

“Across all landscapes the number of fires increases with proximity to public and private campsites.”

Higher lightning density increases the likelihood of false zeroes as many small fires may be ignited but extinguish unobserved during storm events.

“Higher lightning density also increases the likelihood of false zeroes as many small fires may be ignited but extinguish unobserved during storm events.”

Our forecasts suggest the Sonoran or Mojave deserts could see reductions in mFRI by 150+ or more years by 2050 under the GFDL climate.

“Our forecasts suggest this could be the case of the Sonoran or Mojave deserts which could see reductions in mFRI by 150+ or more years by 2050 under the GFDL climate (Fig 5B).”

Omitting anthropogenic influences increases the size coefficients for climate-relevant variables (aetAave and cwdAave) by 24% on average.

“For example, comparing models 1 and 5 demonstrates that omitting anthropogenic influences significantly overstates the role of lightning as an ignition source, and increases the size coefficients for climate-relevant variables (aetAave and cwdAave) by 24% on average.”

Areas with the longest mFRI expected to receive few if any fires include most of the cool and wet western edge of the Klamath Mountains, the high deserts east of the Sierra Nevada, the warm deserts of the southeast, and the isolated sagebrush steppe of the Modoc Plateau.

“Areas with the longest mFRI that are expected to receive few if any fires (Fig 2A) include most of the cool and wet western edge of the Klamath Mountains, the high deserts east of the Sierra Nevada, the warm deserts of the southeast, and the isolated sagebrush steppe of the Modoc Plateau.”

The two models forecast much of the coastal influence zones of the central Coast to see moderate reductions in the number of fire events, with the exception of the Bay Area.

“In contrast, the two models forecast much of the coastal influence zones of the central Coast to see moderate reductions in the number of fire events, with the exception of the Bay Area.”

In areas with low adjacent housing density, the likelihood of fires increases rapidly with additional density up to a turning point, beyond which effects tend toward increased suppression or lack of combustible materials due to the built urban environment.

“In areas with a low adjacent housing density, the likelihood of fires increases rapidly with additional density (i.e., increased ignitions) up to some turning point, beyond which, the effects err towards increased levels of suppression or lack of combustible materials due to the built urban environment.”

Our model captures changes in observed fire frequency that is likely due to changes in anthropogenic forces.

“Our model also captures changes in observed fire frequency that is likely due to changes in anthropogenic forces.”

AET and CWD are good indicators of potential changes in the distribution of fuel amount and its flammability under changing climates.

“As such, AET and CWD are good indicators of potential changes in the distribution of fuel amount and its flammability (i.e., two of the three biophysical categories shown in Fig 1), under changing climates.”

Previous findings of substantially increased numbers of fires and burned area for California may be tied to omitted variable bias from the exclusion of human influences.

“Previous findings of substantially increased numbers of fires and burned area for California may be tied to omitted variable bias from the exclusion of human influences.”

• AssumptionRests on an assumption: The relationships between fire and its predictors estimated from 1976–2000 remain stationary and valid for forecasting 2001–2050.

Californians have built in or near areas of natural vegetation (WUI), much of which is fire-prone.

“Californians have built in or near areas of natural vegetation, known as the wildland-urban interface (WUI), much of which is fire-prone [2,3].”

• cites [2] VC Radeloff, 2005, The wildland-urban interface in the United States, Ecol Appl sourceNot checked yet

  Not checked in the free run.

• cites [3] RB Hammer, 2007, Wildland–urban interface housing growth during the 1990s in California, Oregon, and Washington, Int J Wildl Fire sourceNot checked yet

  Not checked in the free run.

The distribution of vegetation types is largely controlled by the biophysical effects of climatic variables.

“The distribution of vegetation types is largely controlled by the biophysical effects of climatic variables [42–44].”

• cites [42] FA Douglas, 1994, Temporal variation in actual evapotranspiration of terrestrial ecosystems: Patterns and ecological implications, J Biogeogr sourceNot checked yet

  Not checked in the free run.

• cites [43] NL Stephenson, 1990, Climatic control of vegetation distribution: the role of the water balance, Am Nat sourceNot checked yet

  Not checked in the free run.

• cites [44] JR Leathwick, 2001, Competative interactions between tree species in New Zealand’s old-growth indigenous forests, Ecology sourceNot checked yet

  Not checked in the free run.

Projections for the number of wildfire events by end of century vary dramatically, e.g., +9 to 13%, +40 to 49%, and +34 to 53%.

“projections for the number of wildfire events by the end the century vary dramatically (e.g., +9 to 13% [11], +40 to 49% [14], and +34 to 53% [12]).”

• cites [11] JM Lenihan, 2008, Response of vegetation distribution, ecosystem productivity, and fire to climate change scenarios for California, Clim Change sourceNot checked yet

  Not checked in the free run.

• cites [14] C Price, 1994, The impact of a 2xCO2 climate on lightning caused fires, J Clim sourceNot checked yet

  Not checked in the free run.

• cites [12] AL Westerling, 2008, Climate change and wildfire in California, Clim Change sourceNot checked yet

  Not checked in the free run.

Empirical work in California supports the proposed relationship between fire and human activity and the 'pyric transition'.

“Empirical work in California [1] supports the proposed relationship between fire and human activity and the “pyric transition”–that is, the observation that fire activity initially increases in parallel with increased human influence, but then decreases beyond some point as characteristics of the built urban environment and increased suppression effort reduce it [21,22].”

• cites [1] AD Syphard, 2007, Human influence on California fire regimes, Ecol Appl sourceNot checked yet

  Not checked in the free run.

• cites [21] D Bowman, 2011, The human dimension of fire regimes on Earth, J Biogeogr sourceNot checked yet

  Not checked in the free run.

• cites [22] V Butsic, 2015, Land Use and Wildfire: A Review of Local Interactions and Teleconnections, Land sourceNot checked yet

  Not checked in the free run.

The spatial and temporal resolutions (1080m, 25-year) can be considered representative of landscape scale fire regimes for California.

“These spatial and temporal resolutions can be considered representative of landscape scale fire regimes for California [8,9].”

• cites [8] MA Parisien, 2009, Environmental controls on the distribution of wildfire at multiple spatial scales, Ecol Monogr sourceNot checked yet

  Not checked in the free run.

• cites [9] M Krawchuk, 2012, Fire and climate change in California: Changes in the distribution and frequency of fire in climates of the future and recent past (1911–2099)Not checked yet

  Not checked in the free run.

Because ignition data are limited, distance to population centers and housing density are used as proxies for ignitions.

“because ignition data are limited, we use two variables correlated with ignitions (distance to population centers, housing density) as proxies [1].”

• cites [1] AD Syphard, 2007, Human influence on California fire regimes, Ecol Appl sourceNot checked yet

  Not checked in the free run.

The failure to include anthropogenic effects in many models likely overstates the response of wildfire to climatic change.

“As such, the failure to include anthropogenic effects in many models likely overstates the response of wildfire to climatic change.”

• AssumptionRests on an assumption: The relationships between fire and its predictors estimated from 1976–2000 remain stationary and valid for forecasting 2001–2050.

Humans are currently responsible for igniting approximately 95% of wildfires in California.

“humans are currently responsible for igniting approximately 95% of wildfires in the state [1].”

• cites [1] AD Syphard, 2007, Human influence on California fire regimes, Ecol Appl sourceNot checked yet

  Not checked in the free run.

The wildfire perimeter data is considered the most complete record of fire perimeters for the state despite shortcomings.

“Despite some shortcomings this data is considered the most complete record of fire perimeters for the state, compiling information from CALFIRE/FRAP, United States Department of Agriculture, Bureau of Land Management, the National Park Service, and Bureau of Indian Affairs and Department of Defense [27].”

• cites [27] FRAP. Fire Perimeter Project Introduction [Internet]. 2015 [cited 10 Aug 2015]. Available: <ext-link xmlns:xlink="http://www.w3.org/1999/xlink" ext-link-type="uri" xlink:href="http://frap.fire.ca.gov/projects/fire_data/fire_perimeters_index.php" xlink:type="simple">http://frap.fire.ca.gov/projects/fire_data/fire_perimeters_index.php</ext-link>Not checked yet

  Not checked in the free run.

Sparse sampling fraction ensures each observation contributes relatively independent information, minimizing spatial autocorrelation.

“We use a sparse sampling fraction to ensure that each observation contributes relatively independent information, thereby minimizing the effects of spatial autocorrelation.”

Relatively little is known about influences of fire suppression, land use patterns, and other human activities at finer scales of analysis.

“Relatively little is known, however, about influences of fire suppression, land use patterns, and other human activities at finer scales of analysis [1,9,20].”

• cites [1] AD Syphard, 2007, Human influence on California fire regimes, Ecol Appl sourceNot checked yet

  Not checked in the free run.

• cites [9] M Krawchuk, 2012, Fire and climate change in California: Changes in the distribution and frequency of fire in climates of the future and recent past (1911–2099)Not checked yet

  Not checked in the free run.

• cites [20] P Laris, 2013, Integrating Land Change Science and Savanna Fire Models in West Africa, Land sourceNot checked yet

  Not checked in the free run.

Fire data over broader temporal scales can be thought of as 'count' data.

“Further complicating things, fire data over broader temporal scales can be thought of as ‘count’ data, because any particular location is capable of burning in more than one discrete event.”

Much of the inter-model variation can be explained by differences between scenarios, climate models, or assumptions used in simulations.

“Much of this inter-model variation can be explained by differences between the scenarios of change, the climate models, or the assumptions used in such simulations (e.g., effects of climate changes on plant communities, and the omission or inclusion of anthropogenic effects).”

At high distances from populated places, more small fires go unobserved or ideal conditions unignited, creating more false zeroes.

“In contrast, at high distances from populated places more small fires go unobserved or ideal conditions unignited, creating more false zeroes.”

The measure of wildfire counts used in this study should be considered incomplete particularly before 1980.

“The measure of wildfire counts used in this study, like most historical data, should be considered incomplete particularly before 1980; it is however the most comprehensive spatial data set for California.”

The significant influence of distance to public campsites in our model highlights that even low levels of human activity can have a measurable effect on the likelihood of fires in wilderness areas.

“Additionally, the significant influence of distance to public campsites in our model highlights that even low levels of human activity (as may be the case in many public lands) can have a measurable effect on the likelihood of fires in wilderness areas [64].”

• cites [64] AD Syphard, 2008, Predicting spatial patterns of fire on a southern California landscape, Int J Wildl Fire sourceNot checked yet

  Not checked in the free run.

The use of a zero-inflated model partially controls for additional 'false zeros' due to omission from the data, ameliorating the issue of incomplete wildfire counts.

“The use of a zero-inflated model, whereby we partially control for the additional ‘false zeros’ due to omission from the data, ameliorates this issue.”

The failure to include anthropogenic effects in future fire estimates may be significantly overstating the response of wildfire to climatic change alone.

“Therefore, our findings evidence that the failure to include anthropogenic effects in future fire estimates may be significantly overstating the response of wildfire to climatic change alone.”

The ability of our model to simulate the effects of variability on wildfire may be limited.

“However, the ability of our model to simulate the effects of variability on wildfire may be limited.”

The reversed trends in relation to the 2001-2025 period are due to greater moisture and fuel availability in the Sonoran and Mojave deserts under the GFDL scenario, drastically increasing the amount of fire-prone land.

“Such reversed trends in relation to the 2001–2025 period are due to greater moisture and fuel availability in the Sonoran and Mojave deserts under the GDFL scenario, drastically increasing the amount of fire-prone land.”

Such heterogeneity depends on the interacting effects of climatic and human drivers on fire occurrence.

“Such heterogeneity depends on the interacting effects of climatic and human drivers on fire occurrence.”

The results should be relatively conservative, fitting the mean fire response to an ensemble of successional stages and vegetation types, rather than over-fitting to a deterministic equilibrium ecosystem type.

“The results therefore should be relatively conservative, fitting the mean fire response to an ensemble of successional stages and vegetation types, rather than over-fitting to a deterministic equilibrium ecosystem type.”

The failure to include critical anthropogenic effects in models will overstate the importance of climatic variables on wildfire.

“The failure to include these critical anthropogenic effects in models will overstate the importance of climatic variables on wildfire.”

Ignoring human contributions to ignitions would strongly overstate the role of lightning as an ignition source.

“This implies that ignoring human contributions to ignitions would strongly overstate the role of lightning as an ignition source.”

• Also stated in: § Anthropogenic influences on Fire · boosted

  “For example, comparing models 1 and 5 demonstrates that omitting anthropogenic influences significantly overstates the role of lightning as an ignition source, and increases the size coefficients for climate-relevant variables (aetAave and cwdAave) by 24% on average.”

At very high levels of AET, the number of fires declines because highly productive areas become too wet to burn frequently.

“Finally, the number of fires declines at very high levels of AET as highly productive areas become too wet to burn frequently.”

Transitioning from an area unsuitable for continuous vegetation growth (e.g., deserts) to more productive environments increases the number of expected fires.

“For AET, this indicates that transitioning from an area unsuitable for continuous vegetation growth (e.g., deserts) to more productive environments increases the number of expected fires.”

As communities develop over time they will influence the balance between increasing and decreasing the likelihood of wildfire, reinforcing or mitigating the effects of climate change.

“As these communities develop over time they will influence the balance between increasing and decreasing the likelihood of wildfire, reinforcing or mitigating the effects of climate change.”

Our ability to forecast the effects of variation is limited by the GCMs' reliance on the mean of many simulations, muting the effects of extreme events and intra/interannual variability.

“Our ability to forecast the effects of variation is also limited by the GCMs’ reliance on the mean of many simulations, muting the effects of extreme events and intra/interannual variability [61].”

• cites [61] Z Liang X-, 2008, Do CGCMs Simulate the North American Monsoon Precipitation Seasonal–Interannual Variability?, J Clim sourceNot checked yet

  Not checked in the free run.

The similarities of our statistical approach with process-based approaches suggest that AET and CWD are useful proxies for the determinants of both a shifting plant community as well as wildfire distribution.

“The similarities of our statistical approach with such process-based approaches suggest that AET and CWD are useful proxies for the determinants of both a shifting plant community as well as wildfire distribution [9,62].”

• cites [9] M Krawchuk, 2012, Fire and climate change in California: Changes in the distribution and frequency of fire in climates of the future and recent past (1911–2099)Not checked yet

  Not checked in the free run.

• cites [62] NL Stephenson, 1998, Actual evapotranspiration and deficit: biologically meaningful correlates of vegetation distribution across spatial scales, J Biogeogr sourceNot checked yet

  Not checked in the free run.

Given recent history and population density in these areas, even small shifts in mFRI in and around Los Angeles, San Diego, Del Mar, Encinitas and the Eastern side of the San Francisco Bay may be the most significant for human systems.

“Given recent history and population density in these areas, even small shifts in mFRI in and around Los Angeles, San Diego, Del Mar, Encinitas and the Eastern side of the San Francisco Bay may be the most significant for human systems.”

In locations that experience ideal weather conditions for wildfire but a lack of continuous fuel (e.g., semi-arid deserts), small increases in rainfall would increase fuels available to burn.

“At the other end of the spectrum, in locations that experience ideal weather conditions for wildfire but a lack of continuous fuel (e.g., semi-arid deserts), small increases in rainfall would increase fuels available to burn [59].”

• cites [59] GF Rogers, 1987, Winter precipitation and fire in the Sonoran Desert, J Arid Environ sourceNot checked yet

  Not checked in the free run.

As a result, much (or all) of the variability attributable to direct and indirect anthropogenic influences on fire is instead attributed to climatic ones.

“As a result, much (or all) of the variability attributable to direct and indirect anthropogenic influences on fire is instead attributed to climatic ones.”

Many wildfire forecast models do not, or are unable to, include anthropogenic influences in a meaningful way.

“Many wildfire forecast models do not, or are unable to, include anthropogenic influences in a meaningful way [9,10,12,14].”

• cites [9] M Krawchuk, 2012, Fire and climate change in California: Changes in the distribution and frequency of fire in climates of the future and recent past (1911–2099)Not checked yet

  Not checked in the free run.

• cites [10] JM Lenihan, 2003, Climate change effects on vegetation distribution, carbon, and fire in California, Ecol Appl sourceNot checked yet

  Not checked in the free run.

• cites [12] AL Westerling, 2008, Climate change and wildfire in California, Clim Change sourceNot checked yet

  Not checked in the free run.

• cites [14] C Price, 1994, The impact of a 2xCO2 climate on lightning caused fires, J Clim sourceNot checked yet

  Not checked in the free run.

Historically, many high elevation sites in California rarely burned because of long-lasting snowpack.

“Historically, many high elevation sites in California rarely burned because of long-lasting snowpack [55].”

• cites [55] AE Hessl, 2011, Pathways for climate change effects on fire: models, data, and uncertainties, Prog Phys Geogr sourceNot checked yet

  Not checked in the free run.

This is a classic case of omitted variable bias, whereby statistical models would attribute some part, or all, of the anthropogenic influences on variability to other variables retained in the model.

“This is a classic case of omitted variable bias, whereby statistical models would attribute some part, or all, of the anthropogenic influences on variability to other variables retained in the model [63].”

• cites [63] D Gujarati, 2003, Basic EconometricsNot checked yet

  Not checked in the free run.