Background

Wildfire represents the main type of perturbation threatening forests in Europe (San-Miguel-Ayanz et al. 2018). Quantification of the effects of wildfire is important for clarifying ecological impacts and assessing the need to plan and implement measures to restore ecosystems and to reduce flooding and erosion risk (Moody et al. 2016). The main impacts of forest fires include increased runoff and soil erosion, both of which are highly dependent on the level of alteration caused by fire in soil or soil burn severity (Benavides-Solorio and MacDonald 2005; Vega et al. 2005; Cawson et al. 2013; Fernández and Vega 2016; Schmeer et al. 2018; Fernández et al. 2020). Given the potentially large increases in overland flow events after wildfire and the associated flooding and erosion risk, accurate prediction and a good understanding of the relationships between soil properties that control infiltration and burn severity are essential, but remain challenging (Moody et al. 2013).

The normalized burn ratio (NBR) is generally accepted as the standard spectral index for assessing fire severity (e.g., van Wagtendonk et al. 2004; Key and Benson 2006; Lentile et al. 2006; Hudak et al. 2007; De Santis and Chuvieco 2009; Holden et al. 2009; Miller et al. 2009; Harris et al. 2011; Veraverbeke et al. 2011; Veraverbeke et al. 2012). However, there is no agreement about the preferred conditions for determining the spectral indices (Soverel et al. 2010; Cansler and McKenzie 2012). Moreover, most of the above-mentioned studies use the Composite Burned Index (CBI) as a field measure for validating fire severity. The CBI is visually estimated and computes the combined effects of fire on soil and vegetation strata (Key and Benson 2006). However, its use is controversial (e.g., Lentile et al. 2009; Morgan et al. 2014; Smith et al. 2016), and some authors have highlighted the need for independent validation of the approach for specific regions and vegetation types (Picotte and Robertson 2011; McCarley et al. 2018; Tran et al. 2018). In addition, some authors suggest the need for separate vegetation and soil burn severity assessment because these ecosystem components can be affected differently by wildfires depending on stand structure and fuel and fire behavior (Jain et al. 2004; Jain and Graham 2007; Fernández et al. 2020). The risk of soil erosion is higher in forest stands affected by crown fire than in stands in which the tree crowns are only scorched because the presence of needle cast enables water infiltration and thus favors a reduction in post-fire erosion loads (Cerdà and Doerr 2008; Neris et al. 2017; Fernández et al. 2020). Therefore, areas where soil burn severity should be assessed with a view to implementing urgent measures to mitigate post-fire soil erosion are those in which the vegetation is almost completely consumed during fire. However, evaluation of post-fire soil burn severity still relies almost completely on field surveys because the relationships between spectral indices and soil burn severity have scarcely been explored until now (Marcos et al. 2018; Sobrino et al. 2019), and few attempts have been made to quantify alteration in soil properties using spectral indices (Moody et al. 2016).

Post-fire changes in physical and hydraulic properties of soil are important for controlling the magnitude of the hydrological response to rainfall (Wieting et al. 2017). Fire severity alters the strength and persistence of soil water repellency (e.g., Doerr et al. 2000; Huffman et al. 2001; Lewis et al. 2006), and the soil infiltration capacity typically decreases in magnitude when fire severity increases (e.g., Ebel et al. 2012; Moody and Ebel 2012; Ebel et al. 2016; Moody et al. 2016). Pronounced alteration of soil aggregate stability and soil organic carbon content has been associated with high levels of burn severity (Mataix-Solera et al. 2011; Vega et al. 2013). Decreased soil aggregate stability has been related to increased susceptibility to erosion in unburned (Cantón et al. 2009) and burned soils (Fernández et al. 2016).

Pinus pinaster Ait., Pinus sylvestris L., and Pinus pinea L. are the most important forest species in the Iberian Peninsula, growing preferentially on acidic and well-drained soils (Ruiz del Castillo 1979). These forests are frequently affected by forest fires; in fact, almost a third of the area burned between 2006 to 2015 in Spain was pine forest (López and López 2019).

This study explored the relationship between soil burn severity and spectral indices by considering three soil properties related to the susceptibility to soil erosion after fire: saturated hydraulic conductivity, as a surrogate for soil water repellency; soil organic carbon content; and the mean weight diameter of soil aggregates. Quantification of these variables enabled their conversion into continuous variables and exploration of a set of spectral indices for determining the degree of alteration of soil properties after fire in three pine stands in central and southern Spain.

Methods

Study sites

The study was carried out in three wooded areas representative of the pine forests of central and southern Spain (Fig. 1). In all cases, they were pine forests with a shrubby understory of Erica L. sp., Cistus L. sp., and Cytisus Desf. sp. These pine stands were affected by wildfire in the summers of 2018 and 2019. In all cases, time since last fire was more than 20 years.

  1. 1.

    The Nerva Fire (Nerva) burned 1750 ha of forest and agricultural land in southern Spain from 2 to 10 August 2018. The field survey was carried out in forest stands mainly dominated by Pinus pinea L., covering 563 ha of public land. The soils are Eutric Cambisols developed on slates with sandy loam texture. Rainfall in the previous three months prior to fire was 39 mm.

  2. 2.

    The Gavilanes Fire (Gavilanes) burned 1415 ha of forest land in central Spain from 28 June to 4 July 2019, some of which (348 ha) was covered by Pinus pinaster Ait. stands. The soils are Cambisols and Leptosols developed on granite with sandy loam texture. Rainfall in the previous three months prior to fire was 60 mm.

  3. 3.

    La Granja Fire (La Granja) affected 370 ha of Pinus sylvestris L. stands in central Spain from 4 to 7 August 2019. The soils are Cambisols and Leptosols developed on gneiss with sandy loam texture. Rainfall in the previous three months prior to fire was 57 mm.

Fig. 1
figure 1

Location of wildfire-affected areas where study plots were located to assess alterations in soil properties using spectral indices in pine stands affected by crown fire in Spain, in 2018 and 2019

Full size image

Experimental design and field sampling

We established 25 sampling plots in areas where the vegetation was fully consumed by crown fire (minimum size 60 m × 60 m) and where stone cover (particle diameter >10 cm) was <50%: nine plots in Nerva, four plots in Gavilanes, and 12 in La Granja. Physiographic characteristics are compiled in Table 1.

Table 1 Mean physiographic characteristics of the study sites located in each burned area to assess alterations in soil properties using spectral indices in pine stands affected by crown fire in Spain, in 2018 and 2019. Range of variation is given in parentheses. N = north, S = south, E = east, W = west
Full size table

We assessed soil burn severity in plots of radius 20 m, centered in the larger plots. The soil burn severity was assessed with the aid of a 20 cm × 20 cm quadrat, which was placed at 80 systematically selected points along two perpendicular transects. The soil in each quadrat was classified using a modified version (Fernández and Vega 2016) of the soil burn severity index proposed by Vega et al. (2013). That classification considers six soil burn severity levels: (Level 1) very low soil burn severity: burned litter (Oi) but limited duff (Oe + Oa) consumption; (Level 2) low burn severity: Oa layer totally charred and covering mineral soil, possibly some ash; (Level 3) moderate soil burn severity: forest floor (Oi + Oe + Oa layers) completely consumed (bare soil), but soil organic matter not consumed and surface soil intact; (Level 4) high soil burn severity: forest floor completely consumed, soil organic matter in Ah horizon consumed and soil structure altered within a depth of less than 1 cm; (Level 5) very high soil burn severity: same as high soil burn severity but within a soil depth equal to or greater than 1 cm; and (Level 6) as Levels 4 and 5 and color altered (reddish).

In each burned area, we collected five soil samples (0 to 2 cm) for each level of soil burn severity identified. In addition, we collected five samples in unburned patches with similar soil characteristics. The samples were transported to the laboratory in plastic collars (10 cm diameter) to keep them intact. In all cases, we carried out soil sampling within two weeks following fire. Precipitation before sampling varied from 6 mm in Nerva and Gavilanes fires to 10 mm in La Granja Fire.

Laboratory measurements

In each soil sample, saturated hydraulic conductivity (K) was estimated by means of the Decagon model S mini disc infiltrometer (Decagon Devices, Inc., Pullman, Washington, USA) according to the following formula:

$$ K=\frac{C1}{A}, $$
(1)

where C1 is the coefficient of the second-order polynomial term fitting the curve of the cumulative infiltration versus the square root of time (Zhang 1997), and A is the value relating van Genuchten parameters based on the specific soil type to the suction rate and radius of the mini disc infiltrometer (unitless; here A = 3.9099 given an estimate of the soil as a sandy loam as well as a 2.25 cm infiltrometer radius and a suction rate of −2 cm). Duration of the infiltration experiments varied with soil burn severity level, but it was stopped after the maximum run time of 10 min when necessary. The soil samples were air dried and carefully crumbled by hand into small pieces after infiltration experiments. Soil aggregates were dry sieved (Kemper and Rosenau 1986) and separated into size fractions (10 to 5 mm, 5 to 2 mm, 2 to 1 mm, 1 to 0.25 mm, 0.25 to 0.05 mm, and <0.05 mm). Percentage by weight of aggregates in each fraction was determined to enable calculation of the mean weight diameter. Soil organic carbon content of soil samples after sieving (2 mm) was determined by dry combustion in a LECO Elemental Analyzer (LECO, St. Joseph, Michigan, USA).

We obtained mean values of soil parameters (saturated hydraulic conductivity, mean weight diameter of soil aggregates, and soil organic carbon content) for each plot by multiplying the corresponding value associated with each level of soil burn severity by the frequency of each level in each plot.

Spectral indices

We calculated the spectral indices using data from Sentinel-2 satellite (Table 2) provided by the European Space Agency (ESA). Sentinel-2 is a constellation of two satellites providing high-resolution multispectral optical imagery (Drusch et al. 2012; Brown et al. 2018). All study images used were downloaded from the ESA website (Copernicus Open Access Hub, https://scihub.copernicus.eu/dhus/#/home; accessed 10 January 2019) and were Level 2A (bottom of atmosphere) reflectance images, which are atmospherically corrected (Kaufman and Sendra 1988).

Table 2 Sentinel-2 bands characteristics. Bands shown in boldface were used to assess alterations in soil properties in three study areas in pine stands affected by crown fire in Spain, in 2018 and 2019: Study areas, with the pre- and post-fire image dates used for each in parentheses, were: Nerva (31 Jul 2018 and 4 Oct 2018), Gavilanes (1 Jun 2019 and 31 Jul 2019), and La Granja (2 Aug 2019 and 6 Sep 2019)
Full size table

Different spectral indices have been specifically designed to discriminate between different levels of affectation in surfaces affected by fire (Table 3). From images available for each burned area, we selected those from dates as close as possible to the ignition and suppression of each wildfire. Although, a priori, the imagery was cloud free across the study area (both before and after fires), to avoid possible errors, we performed cloud correction using the Scene Classification Image, available in Sentinel-2 Level 2A, of 20 m spatial resolution. This cloud correction is based on a series of threshold tests using the Sentinel-2 bands and their ratios, providing a classification image that is applied as a mask to the original image (Muller-Wilm et al. 2013; Gascon et al. 2017).

Table 3 Selected spectral indices, based on Sentinel-2 bands (Table 2), used to assess alterations in soil properties in three study areas in pine stands affected by crown fire in Spain, in 2018 and 2019: Study areas, with the pre- and post-fire image dates used for each in parentheses, were: Nerva (31 Jul 2018 and 4 Oct 2018), Gavilanes (1 Jun 2019 and 31 Jul 2019), and La Granja (2 Aug 2019 and 6 Sep 2019)
Full size table

Statistical analysis

We used a general linear mixed model to test the effect of soil burn severity on soil properties. The values were normalized by obtaining the relative difference between the value of each soil burn severity level and the corresponding mean value for the control plots in order to compare soil burn severity levels across fires. We considered site as a random factor in the models. When significant effects (P < 0.05) were indicated, post hoc pairwise comparisons (with Bonferroni adjustment for multiple comparisons) were conducted to detect differences between the main effects of treatments and their interactions. Residuals were tested for autocorrelation, normality, and homogeneity of variance.

We used linear regression to detect any possible relationships between soil burn severity and spectral indices. We log transformed the data when necessary to achieve normality. We used the R statistical package (Core Team Development 2019) for all statistical analyses.

Results

Mean values of each soil property in each experimental site and for each soil burn severity level are listed in Table 4. No samples from soil burn severity Level 1 were identified in the burned areas. Soil burn severity Levels 2 and 3 were the most frequent in the field plots. Level 4 frequency ranged from 6% in Nerva and La Granja (soil perturbation depth = 0.5 cm) to 14% in Gavilanes (soil perturbation depth = 0.5 cm). In turn, soil burn severity Level 5 was identified in 6% of the quadrats in Nerva, 8% in Gavilanes, and 13% in La Granja. Only 1% of the quadrats were classified as Level 6 during sampling in the three study sites, with a mean soil perturbation depth of 1.5 cm.

Table 4 Mean values of selected soil properties according to the level of soil burn severity in each burned area affected by crown fire in Spain, in 2018 and 2019. The values of adjacent unburned controls are also presented. The levels of fire severity varied from low (Level 2) to extreme (Level 6) following the classification by Fernández and Vega (2016). Standard error in parentheses
Full size table

The saturated hydraulic conductivity of unburned soils was around 3 mm h−1 in all cases (Table 4). In soils affected by moderate levels of soil burn severity (Levels 2 and 3), a similar and negative relative change was observed (Fig. 2). In the most severely burned soils (Levels 4, 5, 6), the infiltration increased gradually and significantly with soil burn severity.

Fig. 2
figure 2

Relative difference between the value of each soil burn severity level and the corresponding mean value for the control plots for (A) soil saturated conductivity, (B) mean weight diameter of soil aggregates, (C) and soil organic carbon measured in the set of field plots located in the burned areas affected by crown fire in Spain, in 2018 and 2019. Mean values are represented for each study site and for the average of the three study sites. Mean values followed by the same letter are not significantly different (P < 0.05). The levels of fire severity varied from low (Level 2) to extreme (Level 6) following the classification by Fernández and Vega (2016). Vertical bars indicate the standard error

Full size image

The mean weight diameter of soil aggregates decreased with fire severity (Fig. 2). There were no significant differences between the most severely burned soils (Levels 5 and 6). Soil organic carbon decreased gradually and significantly with soil burn severity (Fig. 2).

The significant correlations between the spectral indices and the mean soil parameters are listed in Table 5. The relativized burn ratio (RBR) was the best predictor of saturated hydraulic conductivity and mean weight diameter of soil aggregates. The relationships between the indices and soil organic carbon were weaker in most cases (Table 5), but the best predictor was the differenced mid-infrared bilateral index (dMIRBI), with R2 and root mean square error equal to 0.548 and 0.087, respectively. The NBR spectral index performed similarly for the three variables analyzed (Table 5).

Table 5 Relationships between spectral indices and each soil property considered (transformed in logarithms) for the set of field plots located in the burned areas affected by crown fire in Spain, in 2018 and 2019. RMSE = root mean square error. See Table 3 for spectral index names
Full size table

Discussion

Our study demonstrated that the proposed soil burn severity categories, based on visual signs, are useful for inferring the degree of alteration of three important soil physical properties. In terms of indicating a difference relative to the unburned soil, the best performance was observed for soil organic carbon content, as previously observed in northwest Spain by Vega et al. (2013). As soil organic matter is the principal binding agent in these coarse-textured granitic soils (Benito and Díaz-Fierros 1989; García-Corona et al. 2004), high soil burn severity levels are expected to be associated with decreased soil aggregate stability, as observed in our study. Regarding saturated hydraulic conductivity, our findings are similar to those of Wieting et al. (2017), who reported that this parameter was almost seven times higher in severely burned soils than in soils affected by a low level of burn severity. Soil saturated hydraulic conductivity mimicked the changes in soil water repellency, confirming that this parameter reaches maximum values at lower levels of burn severity (Benito et al. 2009; Fernández et al. 2019).

Information that would enable comparison of the correlations between spectral indices and soil parameters is scarce. Moody et al. (2016) observed significant and positive correlations between differenced Normalized Burn Ratio (dNBR) and some soil hydraulic properties for a burned area in Colorado, USA. In contrast to the observations made in our study, these authors found that soil infiltration decreased with increased fire severity. In our study, RBR performed better for the selected variables than dNBR, as observed by Parks et al. (2014) with CBI as an indicator of fire severity. The Mid-Infrared Burnt Index (MIRBI) and the Burned Area Index for Sentinel-2 (BAIS2) indices have been successfully used to determine differences in canopy cover as a consequence of fire (Filipponi 2018; McCarley et al. 2018). The findings of our study show that dMIRBI may also be a good index for quantifying changes in soil properties after fire, although further research is necessary to confirm this. The findings indicated that relativized metrics are better at detecting high severity effects in large geographic areas, as suggested by Parks et al. (2014).

Conclusions

Our study findings showed that the proposed categories of soil burn severity, based on visual signs, are useful for reflecting gradual changes in some soil variables that play a critical role in post-fire soil erosion susceptibility. The visual indicators were better for inferring the degree of degradation of the soil organic carbon content.

The close relationship between the RBR and cumulative infiltration and mean weight diameter of soil aggregates, and between the dMIRBI and soil organic carbon content appears to be a consequence of the information provided by the Short-Wave Infrared (SWIR) spectral band.

Our study findings show the potential ability of spectral indices to reflect changes in some properties related to soil erosion susceptibility, although more data from different ecosystems is required to confirm the performance of each.

However, the variability found between sites in both soil properties and spectral index values seems to indicate the difficulty of establishing overall ranges of variation.