3.3. Predicting Fire Hotspot Density (FHD) from AcFDI by Vegetation Type and Region
Non-linear models based on fuel dryness showed higher goodness of fit statistics compared to linear models. The R2
values for non-linear models using AcFDI and FDI (Equations (8) and (9), respectively) are shown in Table S1
. For all vegetation types and regions, higher or similar predictive capacity for estimation of FHD was obtained using AcFDI compared to FDI (Table S1
). The highest improvements for FHD using AcFDI instead of FDI were observed for temperate and tropical forests of Centre and temperate forests of the Northwest (NW), where R2
values increased from 0.19, 0.21, and 0.35 to 0.51, 0.50, and 0.74, respectively (Table S1
). Smaller gains or marginal increases were obtained for tropical forests of the South (S) and Northeast (NE) regions.
The best models for predicting fire FHD by vegetation type and region and the corresponding goodness of fit coefficients are summarized in Table 2
. The best fitted models were obtained with AcFDI, calculated with FDIp
threshold values of p
= 95% (FDI95
) from Table 1
, using a non-linear expression, including autoregressive terms (Equation (10)), for all vegetation types.
The chosen percentile for the best fit percentile regression models ranged from 65 to 90, with most vegetation types having a best fit for percentile values of 70 to 80. As defined in the criteria for model selection, the bias of the selected models in Table 1
ranged from −5 to −15, which was chosen to represent a conservative overestimation for cautions risk decision making. The R2
values of the best fit models ranged from approximately 0.5 to 0.7 (Table 2
), with the highest R2
values for temperate forests of the Northwest and Centre regions and for dry tropical forest of the Centre region. The lowest values for several tropical forests of the South and Northeast regions, and for the temperate forest of the arid North-Centre region. RMSE
values ranged from approximately 20 to 80, with most vegetation types having RMSE
values in the range 30–60, as shown in Table 2
The highest values of the b coefficient were obtained for temperate forests of the South, North-Centre, Northwest, Centre, and Northeast regions; with b values ranging from approximately 2 to 4.7. This suggests that for these vegetation types, increases in accumulated FDI index resulted in strong increases in fire occurrence. In contrast, most of the tropical forests had lower b coefficients, ranging between 1 and 2, suggesting a less pronounced effect of AcFDI on FHD for these more humid regions.
The value of the c coefficient for autoregression of FHD was highest in the temperate and tropical forests of the Centre region, suggesting a larger effect of previously existing fires on fire occurrence in these regions.
Examples of the predicted FHD values against observed MODIS FHD with the selected models are shown in Figure 3
for several vegetation types.
The observed temporal evolution of FHD shown in Figure 3
varied by region. The Centre and South regions had earlier starts of the fire season, in the months of January and February, compared to a later start of the fire season in the Northwest region, where fire activity generally started in the months of March and April. (Figure 3
Interestingly, for most vegetation types and years, predicted FHD showed a gradual increase from 0 (no risk) to levels of 50 (medium fire risk) in the first months of the year, prior to the start of the fire season (Figure 3
). This might be useful as an early warning of the levels of fuel dryness accumulation before the start of the fire season, as will be briefly discussed below in Section 4.3
For the temperate forests of the Northwest, years 2011, 2012, and 2013 had high-to-medium fire occurrence as shown in the observed FHD plots (Figure 3
a), whereas years 2014 and 2015 had low and very low fire occurrence, respectively. This matches with the observed FDI values (Figure 2
), where lower FDI values were reached in these latter years, and FDI remained above the FDI threshold for a shorter period compared to the first above-mentioned three years. In this region, the AcFDI index explained a substantial part of the variability in FHD, with R2
values of more than 0.7 for the models with only FDI (Table S1
For the temperate forests of the Northeast, the highest FDI values (up to 80) were achieved in the years 2011 and 2013, together with large periods of FDI well above the threshold (Figure 2
), corresponding to years of high and medium FHD, respectively (Figure 3
b). The fires occurring in this region in the very dry year 2011 were the largest fires recorded in the country, according to CONAFOR records, cited in [54
]. Years 2012 and 2014—when lower FDI values of up to 70 were observed (Figure 2
)—corresponded to low fire activity (Figure 3
b), and year 2015—where the period above the threshold was extremely reduced—had very low fire activity.
In the case of temperate forests from the Centre, both high FDI values and high fire occurrence were reached in the very dry year 2011. In contrast, 2014 was both a year with wetter fuel conditions as noted by lower FDI (Figure 2
) and low fire activity as measured by FHD (Figure 3
c). For the year 2015, in spite of wet fuel conditions as noted in the FDI plot, the observed fire FHD was higher than expected based on only fuel dryness. This seems to suggest a strong anthropogenic influence resulting in a large number of ignitions in spite of wet fuel conditions. In this region, the c
coefficient for autoregression of FHD was higher than in the Northwest, and the increase in R2
was higher with the addition of the autoregressive term compared to modeling with only AcFDI (R2
of 0.7 for the model including autoregression of FHD (Equation (10), Table 2
) compared to a value of 0.5 for the model with only AcFDI (Equation (9), Table S1
For the wetlands of the South (Figure 3
d), very high fire occurrence was recorded for the very dry year 2011 and high FHD was observed for the dry year 2013. Although the maximum FDI values reached (Figure 2
) did not vary strongly between years, the period where the FDI remained above the threshold for fire occurrence was longer in those dry years with higher fire occurrence, resulting in higher AcFDI values. Those values were larger than the AcFDI values in 2014, when fuel dryness accumulated above the threshold for a short period, resulting in a year of low fire occurrence. For the year 2015, the FDI index and the associated FHD predicted from the AcFDI seems to capture both the decrease in month 06 (June) and the rapid increase in the following weeks.