CoChat
Sign Up
Cited by 189Fire effects on ecosystemsLand Use and Ecosystem ServicesFire Detection and Safety Systems

Impact of human population density on fire frequency at the global scale

Open access
Knorr, W., Kaminski, T., Arneth, A., Weber, U.·2014·Biogeosciences
DOI 10.5194/bg-11-1085-2014

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

Abstract. Human impact on wildfires, a major earth system component, remains poorly understood. While local studies have found more fires close to settlements and roads, assimilated charcoal records and analyses of regional fire patterns from remote-sensing observations point to a decline in fire frequency with increasing human population. Here, we present a global analysis using three multi-year satellite-based burned-area products combined with a parameter estimation and uncertainty analysis with a non-linear model. We show that at the global scale, the impact of increasing population density is mainly to reduce fire frequency. Only for areas with up to 0.1 people per km2, we find that fire frequency increases by 10 to 20% relative to its value at no population. The results are robust against choice of burned-area data set, and indicate that at only very few places on earth, fire frequency is limited by human ignitions. Applying the results to historical population estimates results in a moderate but accelerating decline of global burned area by around 14% since 1800, with most of the decline since 1950.

AI snapshot

Written from the paper’s full text.

TL;DR

At the global scale, increasing human population density mainly reduces fire frequency; only in extremely sparsely populated areas (up to 0.1 people km⁻²) does fire frequency rise by 10–20% relative to no population. Applying the fitted population effect to historical population estimates implies a moderate, accelerating decline in global burned area of about 14% since 1800, mostly after 1950.

Key findings

  • All optimisations with maximum population density ≥1 km⁻² against the standard burned-area data yielded an exponential decline of fire frequency with population (negative *e*; relative uncertainty 1–15%, highly significant). Only the two successful cases capped at 0.1 km⁻² had positive *e*, indicating a significant positive impact at very low densities.
  • The inferred decline is moderate: a >50% reduction in fire frequency is only predicted for population densities between 40 and 70 inhabitants km⁻².
  • Historical extrapolation using observed burned area and HYDE population gives about a 14% decline in global burned area since 1800 (1−1/1.16 = 0.14), with most of the decline since 1950.
  • Results are robust across burned-area data sets; GFED-SF (including small fires) also yielded negative *e* in all cases, though *e* was consistently lower than for GFED3, not enough to reverse the negative impact.

Methods

  • SIMFIRE: fire frequency *f*ᵢ,ₜ = *a*(*l*ᵢ) *F*ᵢᵇ *N*ᵢ,ₜᶜ logit(*d* + *eP*ᵢ), with FAPAR as a fuel proxy, Nesterov index as a climate/fire-risk proxy, population density *P*, and land-cover multiplier *a*; optimised against GFED3, MODIS MCD45, L3JRC, and GFED-SF at 0.5° resolution.
  • 23 optimisation experiments (four burned-area data sets × five population caps for eight regions, plus three global 32-region cases), each repeated from 20 random starting points (460 optimisations); Levenberg-Marquardt minimisation, with posterior parameter uncertainties from the inverse Hessian.
  • Historical application used HYDE 3.1 population and Monte Carlo sampling (10,000-member ensemble) to propagate parameter uncertainty; agricultural grid cells (crop fraction ≥50%) were handled separately in a sensitivity test.

Limitations

  • Stated: Eq. (1) cannot represent a humped relationship between fire frequency and population; attempts to optimise a double-logistic peaked form failed consistently. Space-for-time substitution is limited because low population density today may reflect land abandonment, and historical land-clearing phases differ by region. Small fires are undetected by the satellite sensors, and GFED-SF is indirect and uncertain. The eight-region model overestimates fire frequency in some regions (e.g., western US/northwestern Mexico), and the simple approach cannot capture complex socio-economic interactions.
  • Evident from text: one optimisation (GFED3, up to 0.1 km⁻²) was excluded because the Hessian was too close to singular; only two of three optimisations succeeded for the 0.1 km⁻² cap. The model’s implicit dependence on population via FAPAR is minimised but not eliminated by excluding predominantly agricultural areas.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.5194/bg-11-1085-2014

Cites

48 of 48 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 48 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    The optimised land cover scalar, either a(l_i) or a_d(l_i), shows patterns that are consistent with the main global distribution of fire frequency previously discussed.

    “The optimised land cover scalar, either a(l i ) or a d (l i ), shows some patterns that are consistent with the main global distribution of fire frequency previously discussed (Table 2 for global, 1 and 0.1 km -2 maximum population density).”
    Quote in the paperNot a substantial claim
  • Findingreasoning

    Observations are more 'spotty' (show more spatial variability) than simulations, in particular in the humid tropics, for shrublands, and in the boreal zone.

    “Another general difference is that observations are more “spotty” (i.e.show more spatial variability) than simulations, Fig. 2.”
    • Also stated in: § 3.2 Comparison with observed fire frequency · plain
      “in particular in the humid tropics, for shrublands, and in the boreal zone.”
    Quote in the paperNot a substantial claim
  • Findingreasoning

    The three 32-region optimisations achieve a slightly lower residual model-data misfit than the three other global cases.

    “With more degrees of freedom, the three 32-region optimisations achieve a slightly lower residual model-data misfit than the three other global cases.”
    Quote in the paperNot a substantial claim
  • Findingown data

    At the global scale, increasing population density mainly reduces fire frequency.

    “We show that at the global scale, the impact of increasing population density is mainly to reduce fire frequency.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Wildfires require a combination of long and hot dry seasons and a continuous cover of sufficient amounts of fuel.
    • The authors state this as an assumption: Fire frequency is proportional to the product of a fuel-continuity/load factor and a fire-risk factor.
    • The authors state this as an assumption: Maximum annual FAPAR equal to zero leads to zero fire frequency.
    • The authors state this as an assumption: If maximum annual FAPAR approaches 1, vegetation cover is continuous and there is also a continuous ground cover of various fuel types.
    • The authors state this as an assumption: The mean fractional vegetation cover is close to the mean fractional cover of fuel.
    • The authors state this as an assumption: Fire frequency reaches zero when the maximum Nesterov index during a given fire year is zero, but increases non-linearly with increasing annual maximum Nesterov index.
  • Findingown data

    All optimisations with a maximum population density of 1 km^-2 or higher against the standard burned-area data sets yield an exponential decline of fire frequency with population, using Eq. (2) with negative values of e.

    “All optimisations with a maximum population density of 1 km -2 or higher against the standard burned-area data sets yield an exponential decline of fire frequency with population, using Eq. ( 2) with negative values of e (Table 2, see Table A1 of Appendix A for further cases).”
    Quote in the paperSubstantial claim
  • May combine two claims. 41
  • Classified as probably not a substantial claim. 27

  • May combine two claims.

    Findingnone given

    L3JRC shows much higher values for the boreal zone than the other two data sets, but lower values for the African savannas and grasslands.

    “For example, L3JRC shows much higher values for the boreal zone than the other two, but lower val-ues for the African savannas and grasslands.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Those patterns are also consistent between data sets used for optimisation.

    “Those patterns are also consistent between data sets used for optimisation.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Highest values of the optimised land cover scalar are found for savanna/grassland and shrubland, and low values for broadleaf forest and tundra.

    “Highest values are found for savanna/grassland and shrubland, and low values for broadleaf forest and tundra.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    SIMFIRE with optimised parameters reproduces the main features of global fire distributions: very high values in African and Australian savannas and grasslands, intermediate values in grasslands of South America and Central Asia, moderate fire frequency for Mediterranean ecosystems and the boreal zone, and low values for deserts and humid-temperate areas of North America, Europe and Asia.

    “SIMFIRE with optimised parameters reproduces the main features of global fire distributions (Figs.2345): very high values in African and Australian savannas and grasslands, intermediate values in grasslands of South America and Central Asia, moderate fire frequency for Mediterranean ecosystems and the boreal zone, low values for deserts and humidtemperate areas of North America, Europe and Asia.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The results for eight regions are similar to those with 32 regions.

    “The results for eight regions are similar to those with 32 regions (see Table A2).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning

    Applying the results to historical population estimates results in a moderate but accelerating decline of global burned area by around 14% since 1800, with most of the decline since 1950.

    “Applying the results to historical population estimates results in a moderate but accelerating decline of global burned area by around 14 % since 1800, with most of the decline since 1950.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The spatial dependence of fire frequency on population density can be substituted by a temporal dependence.
    • The authors state this as an assumption: The impact of population density alone can be considered while keeping all other factors constant, without the possible influence of a changing climate.
    • The authors state this as an assumption: X(P_i,t) describes the dependence of fire frequency on population density.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Model-data agreement is comparable between the burned-area data sets used.

    “In general, model-data agreement is comparable between the burned-area data sets used.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Only for the two cases with a maximum of 0.1 km^-2 is e positive, but with a much larger relative uncertainty.

    “Only for the two cases with a maximum of 0.1 km -2 , e is positive, but with a much larger relative uncertainty.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The optimisations against GFED-SF all yield an exponential decline of fire frequency as a function of population density, even for the case up to 0.1 km^-2, and in each case the value of e is less than zero by many standard deviations.

    “The optimisations against GFED-SF all yield an exponential decline of fire frequency as a function population density, even for the case up to 0.1 km -2 , and in each case, the value of e is less than zero by many standard deviations (see Table A3).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    For the other two optimisations, the uncertainty of one optimisation is much smaller than the difference between the two, so the results of the two optimisations are shown separately.

    “For the other two optimisations, the uncertainty of one optimisation is much smaller than the difference between the two.We therefore show the results of the two optimisations separately (blue and green lines in Fig. 6).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The temporal extrapolation shows a decline of fire frequency by about 14% since 1800 as a result of changes in population density, with most of the decline happening since 1950.

    “The results show a decline of fire frequency by about 14 % since 1800 (1-1/1.16= 0.14) as a result of changes in population density, with most of the decline happening since 1950 (Fig. 7).”
    • Also stated in: § 4 Discussion · hedged
      “Extrapolation of our results back in time yields an estimate of 14 % for the decline in burned area since 1800, or about the same since the late 19th century.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The way population density is related to human impact on fire frequency remains the same across time.

  • May combine two claims.

    Findingown data

    In a test without replacing observed by modelled fire frequency for predominantly cropland areas, the median value of the confidence interval for MCD45 at 1800 changed from 1.137 to 1.124, and for L3JRC at 1800 from 1.169 to 1.150.

    “In this case, the median value of the confidence interval for MCD45 at 1800 changed from 1.137 to 1.124, and for L3JRC also at 1800 from 1.169 to 1.150.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    For the cases with up to 0.1 inhabitants km^-2, only two optimisations were successful.

    “For the cases with up to 0.1 inhabitants km -2 , only two optimisations were successful.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    For maximum values between 1 and 100 km^-2, the inferred dependence declines earlier than for the global case, but only the result with up to 1 km^-2 shows a significant difference (marked by non-overlapping confidence ranges).

    “For maximum values between 1 and 100 km -2 , the inferred dependence declines earlier than for the global case, but only the result with up to 1 km -2 shows a significant difference (marked by non-overlapping confidence ranges).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning

    Even though explicit models of ignitions and human fire suppression can successfully reproduce global patterns of fire frequency, this study does not find empirical evidence that representing the processes separately is necessary in a global model.

    “Even though explicit models of ignitions and human fire suppression can successfully reproduce global patterns of fire frequency (Li et al., 2012), we do not find empirical evidence that representing the processes separately is necessary in a global model.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Li et al., 2012

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    There is a small uncertainty range, but a larger difference between the results with different burned-area data sets.

    “There is a small uncertainty range, but a larger difference between the results with different burned-area data sets.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning

    In the global analysis based on spatial instead of temporal patterns, there is only an increase of about 20% between the low end of population density and a threshold value of 0.1 inhabitants km^-2, where the threshold is of a similar magnitude as that found by Guyette et al. (2002).

    “In our global analysis based on spatial instead of temporal patterns, we only find an increase of about 20 % between the low end of population density and a threshold value of 0.1 inhabitants km -2 , where the threshold is of a similar magnitude as that found by Guyette et al. (2002).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Guyette et al. (2002)

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    The difference in e decreases when the optimisation was constrained to lower and lower population densities and reverses for a limit of 1 inhabitant km^-2.

    “The difference decreases when the optimisation was constrained to lower and lower population densities and reverses for a limit of 1 inhabitant km -2 .”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    For the case with the sparsest population density and GFED-SF, the inferred relationship is still negative, but there is no result for GFED3.

    “It is also interesting to note that for the case with the sparsest population density and GFED-SF, the inferred relationship is still negative (but there is no result for GFED3).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The decline starts earlier when the population density was capped at lower values, but only reaches moderate reductions in fire frequency at the maximum allowed value.

    “The decline starts earlier when the population density was capped at lower values, but only reaches moderate reductions in fire frequency at the maximum allowed value.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The effect of missing small fires is not pronounced enough to reverse the generally negative impact of population density on fire frequency because the inferred value of e is still highly significantly below zero.

    “However, the effect is not pronounced enough to reverse the generally negative impact of population density on fire frequency because the inferred value of e is still highly significantly below zero.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Even for the 32-region case, parameter e remains highly significantly below zero, leading to the same conclusion that the dominant influence of humans on the global scale is to suppress fires.

    “Nonetheless, we find that even for the 32-region case, parameter e remains highly significantly below zero, leading to the same conclusion that the dominant influence of humans on the global scale is to suppress fires.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The predominant effect of increasing population is to reduce fire frequency, except for extremely sparsely populated areas, where the effect is only slightly positive.

    “We present the first global analysis showing that the predominant effect of increasing population is to reduce fire frequency, except for extremely sparsely populated areas, where the effect is only slightly positive.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    The population density formulation is able to capture many different functional forms with only two parameters, without any prior assumption about the direction of the impact.

    “we choose a formulation that is able to capture many different functional forms with only two parameters, without any prior assumption about the direction of the impact.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    In one case (GFED3, up to 0.1 km^-2), the Hessian matrix of J was too close to singular to be inverted numerically, and the specific case was excluded from further analysis.

    “In one case (GFED3, up to 0.1 km -2 ), the Hessian matrix of J (Eq. 3) was too close to singular to be inverted numerically, and the specific case was excluded from further analysis.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    In the two cases with a maximum of 0.1 km^-2, e is still significantly larger than 0, so there is a significant positive impact of population density on fire frequency.

    “However, e is still significantly larger than 0, so that in this case we infer a significant positive impact of population density on fire frequency.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Archibald et al. (2009) found that fractional area burnt decreased for values above about 20 people per km2 for Southern Hemisphere Africa, but showed no clear trend for values below.

    “Archibald et al. (2009) found that fractional area burnt decreased for values above ca.20 people per km 2 for Southern Hemisphere Africa, but showed no clear trend for values below.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Archibald et al., 2009

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    The spotty feature occurs when observed fire return time (the inverse of fire frequency) is larger than the length of the observational period of no more than 10 yr.

    “This feature occurs when observed fire return time (the inverse of fire frequency) is larger than the length of the observational period of no more than 10 yr (essentially all areas shown as blue, green or yellow).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    A frequently observed pattern for fire density is few fires at low population density, a peak at intermediate levels around 20-40 people per km2, and a rapid drop above 100 people per km2.

    “A frequently observed pattern for fire density -as opposed to fire frequency -is one with few fires at low population density, a peak at intermediate levels (ca.20-40 people per km 2 ), and a rapid drop above 100 people per km 2”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Syphard et al., 2007

      Not checked in the free run.

    • cites Archibald et al., 2009

      Not checked in the free run.

    • cites Archibald et al., 2010

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Guyette et al. (2002) identified distinct historical phases in the midwestern United States: fire frequency was limited by low numbers of human ignitions up to a population density of 0.64 km-2, then reached an ignition-saturated plateau where fuel was limiting, and then declined again above 3.4 km-2 due to human-caused landscape fragmentation.

    “Fire frequency appeared limited by low numbers of human ignitions up to a population density of 0.64 km -2 , then reached an ignitionsaturated plateau where fuel was limiting, and then declined again for values higher than 3.4 km -2 due to human-caused landscape fragmentation.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Guyette et al., 2002

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    GFED3 and MCD45 are more similar, but are based largely on the same original satellite data from the MODIS sensor.

    “GFED3 and MCD45 are more similar, but are based largely on the same original satellite data from the MODIS sensor (Giglio et al., 2010).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Giglio et al., 2010

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    The authors deliberately did not introduce any formulation that would have pre-supposed a humped or peaked shape of the population function, in order to avoid introducing prior assumptions.

    “We deliberately did not introduce any formulation that would have pre-supposed any humped or peaked shape of the population function in order to avoid introducing prior assumptions.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    The higher values for urban/crop/natural vegetation mosaic are possibly due to some remaining presence of agricultural fires.

    “The values for urban/crop/natural vegetation mosaic are somewhat higher than for broadleaf forest, possibly due to some remaining presence of agricultural fires.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    All 460 optimisations converged to a minimum indicated by a norm of the gradient of the simple model-data misfit (Eq. 4) below 10^-3.

    “All 460 optimisations converged to a minimum indicated by a norm of the gradient of the simple model-data misfit (Eq.4) below 10 -3 .”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    Areas with greater than 50% agricultural land were excluded in order to minimise the impact of agricultural waste burning on the results.

    “We also exclude areas with greater than 50 % agricultural land using data by Ramankutty and Foley (1999) for the latest year (2005) in order to minimise the impact of agricultural waste burning on the results.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Ramankutty and Foley, 1999

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    Marlon et al. (2008) suggest that the Earth is currently in a state of low fire frequency as a result of global population growth.

    “Based on the global sedimentary charcoal record, Marlon et al. (2008) suggest that the earth is currently in a state of low fire frequency as a result of global population growth.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Marlon et al., 2008

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Lehsten et al. (2010) found that for the entire African continent, fire frequency generally decreased towards higher population density if other factors were taken into account.

    “Lehsten et al. (2010) in an analysis for the entire African continent found that fire frequency generally decreased towards higher population density if other factors were taken into account.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Lehsten et al., 2010

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    The algorithms used differ, which has a visible impact in some regions, in particular Australia.

    “The algorithms used differ, however, which has a visible impact in some regions, in particular Australia.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Most fuel is either woody plant litter or dried grass, and it is co-located with the photosynthesizing vegetation it is derived from.

    “Because most fuel is either woody plant litter or dried grass (Stocks and Kauffman, 1997), it is co-located with the photosynthesizing vegetation it is derived from.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Stocks and Kauffman, 1997

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Global-scale studies based on satellite observations have found a strong human impact on the seasonality of fire activity.

    “Such global-scale studies based on satellite observations have found a strong human impact on the seasonality of fire activity (Le Page et al., 2010).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Le Page et al., 2010

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Either the very few humans present in almost uninhabited areas can cause as much burned area as humans in moderately populated areas, or lightning strikes will always fill the void in ignitions if no humans are present.

    “Either the very few humans present in almost uninhabited areas can cause as much burned area as humans in moderately populated areas, or lightning strikes will always fill the void in ignitions if no humans are present.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Phases of intensive land clearing tend to occur in tropical forests today, but in temperate forests during the previous centuries.

    “Also, phases of intensive land clearing tend to occur in tropical forests today, but in temperate forests during the previous centuries (Mouillot and Field, 2005).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Mouillot and Field, 2005

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Long-term observations for Canada show that lightning-caused fires had on average 2.4 times higher burned area than human-caused fires, based on lightning causing 85% of burned area and 70% of fires.

    “Long-term observations for Canada show that lightning-caused fires had on average 2.4 times higher burned area than human-caused fires (Stocks et al., 2002) (based on lightning causing 85 % of burned area and 70 % of fires).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Stocks et al., 2002

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Guyette et al. (2002) note their upper limit of 0.67 km^-2 for the ignition-limited regime might be unusually high because their study area was located in a region of Missouri where lightning fires were rare and which was dissected by numerous steep ridges and streams, thus containing many natural barriers to fire spread.

    “As Guyette et al. (2002) note, their upper limit of 0.67 km -2 for the ignition-limited regime might be unusually high because their study area was located in a region of Missouri where lightning fires were rare and which was also dissected by numerous steep ridges and streams, thus containing many natural barriers to fire spread.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Guyette et al. (2002)

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    If humans have the tendency to decrease the size of fires, then global data sets might miss burned area in particular areas with higher population density.

    “If humans have the tendency to decrease the size of fires, then these global data sets might miss burned area in particular areas with higher population density.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationown data

    The results do not show an increase prior to this period, which is also consistent with the view that this increase was caused by climate change.

    “They do not show, however, an increase prior to this period, which is also consistent with the view that this increase was caused by climate change.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    Between 0.1 and 10 km^-2, the ignition-saturated regime dominates, while above 10 inhabitants km^-2 the dominant regime appears to be one of fuel-fragmentation.

    “Between 0.1 and 10 km -2 , the ignition saturated regime dominates, while above 10 inhabitants km -2 the dominant regime appears to be one of fuel-fragmentation.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisereasoning

    There is no consideration allowing human population density to influence the average burned area per fire, and thus no possibility of a fire-saturated regime where the number of ignitions can increase with no impact on total area burned.

    “There is no consideration allowing human population density to influence the average burned area per fire, and thus -by designno possibility of a fire-saturated regime, where the number of ignitions can increase with no impact on the total area burned (but an increase in number of fires).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The optimisation against GFED3 did not yield an uncertainty estimate for numerical reasons because the condition number of the Hessian at the minimum was larger than 10^15.

    “The optimisation against GFED3 did not yield an uncertainty estimate for numerical reasons because the condition number -i.e. the ratio of the highest and lowest eigenvalues -of the Hessian at the minimum was larger than 10 15 .”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    The presence of humans in sparsely to moderately populated regions (up to ca. 10 inhabitants km^-2) was found to cause an increase in fire number and a corresponding decrease in fire size.

    “The presence of humans in sparsely to moderately populated regions (up to ca. 10 inhabitants km -2 ) was found to cause an increase in fire number and a corresponding decrease in fire size (Syphard et al., 2007(Syphard et al., , 2009;;Archibald et al., 2010).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Syphard et al., 2007

      Not checked in the free run.

    • cites Syphard et al., 2009

      Not checked in the free run.

    • cites Archibald et al., 2010

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning

    The simple approach chosen here for the global scale necessarily cannot capture many of the complex interactions between socio-economic factors and fire activity.

    “However, the simple approach chosen here for the global scale necessarily cannot capture many of the complex interactions between socio-economic factors and fire activity.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Methane isotope data indicate an approximately twofold increase since 1800, but carbon monoxide data indicate a ~70% decline since the late 1800s.

    “Of these, methane isotope data indicate an approximately twofold increase since 1800 (Ferretti et al., 2005), but carbon monoxide (CO) data a ∼ 70 % decline since the late 1800s (Wang et al., 2010).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Ferretti et al., 2005

      Not checked in the free run.

    • cites Wang et al., 2010

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Humans also influence vegetation cover (represented by FAPAR), creating an implicit dependence.

    “However, humans also influence vegetation cover (represented by FAPAR), creating an implicit dependence.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    van der Werf et al. (2013) concluded that the strong decline in emissions reported by Wang et al. (2010) is difficult to reconcile with what is known about emission sources, and that emissions were likely not as high during historical periods.

    “A further constraint is given by van der Werf et al. (2013), who used bottom-up calculations and atmospheric transport modelling to conclude that the strong decline in emissions reported by Wang et al. (2010) is difficult to reconcile with what we know about emission sources, and that emissions were likely not as high during historical periods.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites van der Werf et al. (2013)

      Not checked in the free run.

    • cites Wang et al. (2010)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    A possible explanation for the much lower increase from the lowest end of population density to the threshold is that only few fire-affected areas below the inferred average threshold of 0.1 km^-2 experience an ignition-limited fire regime.

    “A possible explanation for the much lower increase from the lowest end of population density to the threshold is that only few fire-affected areas below the inferred average threshold of 0.1 km -2 experience an ignition-limited fire regime.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationreasoning

    A possibility is that land management in those land cover types differs substantially between major regions.

    “A possibility is that land management in those land cover types differs substantially between major regions.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationreasoning

    This may be because lightning ignitions are common, or because many areas have few natural barriers and very few human or lightning ignitions are sufficient to create an ignition-saturated regime.

    “This may be because lightning ignitions are common, or because many areas have few natural barriers and very few hu-man or lightning ignitions are sufficient to create an ignitionsaturated regime.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Guyette et al. (2002) developed a theoretical framework of fire regimes at varying stages of human settlement with three main regimes: an ignition-limited one at low population density, an ignition-saturated one where fire frequency peaks, and a fuel-fragmentation regime where human activity limits the spread of fire.

    “A useful theoretical framework of fire regimes at varying stages of human settlement has been developed by Guyette et al. (2002), where three main regimes were observed: an ignition-limited one at low population density, an ignition-saturated one where fire frequency peaks, and Fig. 7.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Guyette et al. (2002)

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Space-for-time substitution has limitations: low population density in some areas today may be the result of land abandonment and the fire regime in such cases may not be comparable to the pre-industrial situation preceding population expansion.

    “We must note, however, that this kind of space-for-time substitution has its limitations.Low population density in some areas today may be the result of land abandonment and the fire regime in such cases not comparable to the pre-industrial situation preceding population expansion.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    Based on the assumption that all main fire regimes that have occurred historically still exist on earth, the sensitivity analysis infers that current observed patterns of fire frequency and human population are consistent with the view that the dominant cause of the observed decline in global fire activity since industrialisation was the further expansion of human populations, as opposed to an ear

    “Based on the assumption that all main fire regimes that have occurred historically still exist on earth (Bowman et al., 2011), we have used a sensitivity analysis to infer that the current observed patterns of fire frequency and human population are consistent with the view that the dominant cause of the observed decline in global fire activity since industrialisation was the further expansion of”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Bowman et al., 2011

      Not checked in the free run.

    • cites Marlon et al., 2008

      Not checked in the free run.

    • The authors state this as an assumption: All main fire regimes that have occurred historically still exist on earth.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationnone given

    These thresholds are not fixed values but carry considerable uncertainty and are merely indicative of the order of magnitude.

    “These thresholds are naturally not fixed values but carry considerable uncertainty and are merely indicative of the order of magnitude.”
    Quote in the paperNot a substantial claim
61 of 61

  • The authors state this as an assumption.

    Wildfires require a combination of long and hot dry seasons and a continuous cover of sufficient amounts of fuel.

    “We formulate a model based on the assumption that wildfires require a combination of long and hot dry seasons and a continuous cover of sufficient amounts of fuel.”

  • The authors state this as an assumption.

    Fire frequency is proportional to the product of a fuel-continuity/load factor and a fire-risk factor.

    “Fire frequency is therefore assumed proportional to the product of two factors, where the first is a function of some quantity approximating fuel continuity and load, the other a function of some indicator of fire risk.”

  • The authors state this as an assumption.

    Maximum annual FAPAR equal to zero leads to zero fire frequency.

    “Because fires cannot spread on bare ground without fuel, maximum annual FAPAR = 0 can be assumed to lead to zero fire frequency.”

  • The authors state this as an assumption.

    If maximum annual FAPAR approaches 1, vegetation cover is continuous and there is also a continuous ground cover of various fuel types.

    “If maximum annual FAPAR approaches 1, vegetation cover is continuous and we assume that there is also a continuous ground cover of various fuel types, such as dead plant litter or dried grass.”

  • The authors state this as an assumption.

    The mean fractional vegetation cover is close to the mean fractional cover of fuel.

    “We assume that this mean fractional cover is close to the mean fractional cover of fuel, which is a reasonable assumption since in the absence of fire, the turnover time of grassy or woody fuel is several years to a few decades (Weedon et al., 2009).”

  • The authors state this as an assumption.

    Fire frequency reaches zero when the maximum Nesterov index during a given fire year is zero, but increases non-linearly with increasing annual maximum Nesterov index.

    “We assume that fire frequency reaches zero when the maximum Nesterov index during a given fire year is zero, but increases non-linearly with increasing annual maximum Nesterov index.”

  • The paper relies on this without support.

    No spatially explicit data of fuel load is available at the global scale, so indirect measures of fuel continuity must be used.

    “Unfortunately, no spatially explicit data of fuel load is available at the global scale and we have to rely on indirect measures of fuel continuity.”

  • The authors state this as an assumption.

    The observational uncertainty is of similar magnitude for the other burned-area products as for GFED3.

    “Only the GFED3 data set contains information about observational uncertainty, but we assume that the uncertainty is of similar magnitude for the other burned-area products.”

  • The paper relies on this without support.

    The spatial dependence of fire frequency on population density can be substituted by a temporal dependence.

    “With this model application, we substitute results derived from the spatial dependence of fire frequency on population density by a temporal dependence.”

  • The authors state this as an assumption.

    The impact of population density alone can be considered while keeping all other factors constant, without the possible influence of a changing climate.

    “Also, we specifically consider the impact of population density alone, without the possible influence of a changing climate, keeping all other factors constant.”

  • The authors state this as an assumption.

    X(P_i,t) describes the dependence of fire frequency on population density.

    “Assuming that X(P i,t ) describes the dependence of fire frequency on population density, we compute fire frequency f i,t for the year t at grid cell i as:”

  • The authors state this as an assumption.

    The uncertainty is initially assumed constant so that minimisation of J becomes independent of its value and specification of model and data error can be postponed.

    “We start the optimisations by assuming a constant uncertainty, so that the minimisation of J becomes independent of this value, and we can postpone specification of the model and data error.”

  • The authors state this as an assumption.

    The way population density is related to human impact on fire frequency remains the same across time.

    “For the purpose of consistency testing, we also assume that the way population density is related to human impact on fire frequency remains the same across time.”

  • The authors state this as an assumption.

    All main fire regimes that have occurred historically still exist on earth.

    “Based on the assumption that all main fire regimes that have occurred historically still exist on earth (Bowman et al., 2011)”

  • The authors state this as an assumption.

    The difference in e between GFED-SF and GFED3 is caused by higher fire density (smaller fires) caused by more humans.

    “This is in line with the assumption that the difference in e is caused by higher fire density (smaller fires) caused by more humans.”

  • The authors state this as an assumption.

    All other factors besides human population density remain equal in the optimisation.

    “The optimisation presented here was designed in a way to capture only the explicit effect of human population density on fire frequency (Eq. 1 or 2), all other factors remaining equal.”
16 of 16

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Eq. (6) uses an undefined parameter

    Minor

    The historical-population function uses e_a in its exponential branch, but the accompanying definition identifies e_g as the parameter for the global case. No e_a is defined there.

  • Global model parameter count names d instead of e

    Minor

    The stated count of 11 or 35 parameters is consistent with eight or 32 land-cover parameters plus three global parameters. But Eq. (2), which the passage says was used for every global case, has b, c, and e-not b, c, and d.

  • GFED-SF parameter difference is described as both consistent and reversing

    Minor

    The discussion says GFED-SF's inferred e is consistently lower than GFED3's, then says the difference reverses at a population-density limit of 1 inhabitant km−2. A difference that reverses is not consistently in the stated direction.

3 of 3

Cites

Flagged first, then by citations

Research works printed but not linked: 2

  1. Moreira, F., Rego, F. C., and Ferreira, P. G.: Temporal (1958–1995) pattern of change in a cultural landscape of northwestern Portugal: implications for fire occurrence, Landscape Ecol., 16, 557–567, 2001. · DOI
  2. Press, W. H., Teukolsky, S. A., Vetterling, W. T., and Flannery, B. P.: Numerical Recipes: the Art of Scientific Computing, 3rd edn., Cambridge University Press, Cambridge, 2007.

Cited by

Sorted by citations

Authors
Knorr, W., Kaminski, T., Arneth, A., & Weber, U.
Affiliations
  • Lund University (SE)
  • FastOpt (Germany) (DE)
  • Karlsruhe Institute of Technology (DE)
  • Max Planck Institute for Biogeochemistry (DE)
Corresponding author
Arneth, A.
ORCID
Venue
Biogeosciences
Publisher
Copernicus GmbH
Published
2014-02-26
Volume
11
Issue
4
Pages
1085-1102
ISSN
  • 1726-4170 (ISSN-L)
  • 1726-4189
DOI
10.5194/bg-11-1085-2014
OpenAlex
W2150165627
Semantic Scholar
f4aada61b00d877d931780071e71c6b34b791f03
Cited by
189
References
48
Citations by year
2026: 6 · 2025: 20 · 2024: 14 · 2023: 20 · 2022: 16 · 2021: 25 · 2020: 14 · 2019: 16 · 2018: 13 · 2017: 15 · 2016: 18 · 2015: 9 · 2014: 3
FWCI
12.40
Type
journal article
Language
en
Open access
Licence
Locations
  • Biogeosciences (publishedVersion, cc-by)
  • Repository KITopen (Karlsruhe Institute of Technology) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Lund University Publications (Lund University) (submittedVersion)
  • MPG.PuRe (Max Planck Society) (submittedVersion, other-oa)
  • KITopen (cc-by)
Funders
European Commission (265148, 243888)
Topics
  • Physical Sciences › Environmental Science › Global and Planetary Change › Fire effects on ecosystems
  • Physical Sciences › Environmental Science › Global and Planetary Change › Land Use and Ecosystem Services
  • Physical Sciences › Engineering › Safety, Risk, Reliability and Quality › Fire Detection and Safety Systems
Keywords
human population density, fire frequency, forest fires, burned area, human ignition, parameter estimation, uncertainty analysis, historical population estimates

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-06.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-06.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review