Introduction 

It is recognized that wildfires can seriously alter the hydrological response of watersheds to precipitation, with increases in peak discharges and sediment yields. These changes can be attributed to several physical processes occurring during fires, which result in the reduction of interception and infiltration rates due to the destruction of canopy cover, the alteration of the hydraulic proprieties of soils, the creation of ash and consequent water repellence phenomena ([10], [20], [11], [35], [39], [33], [25], [13], [41], [31]).

Post-fire measurements show that changes in peak discharges are usually larger than changes in runoff volumes ([17], [28], [29], [7]). For this reason, the unit-area peak discharge is considered to be the most sensitive parameter for the description of the modified watershed response after a wildfire (e.g., [36]). More specifically, experimental analyses conducted both in Europe and in the United States indicate that the annual peak discharge in post-fire conditions can increase by a factor generally ranging from 1.2 to 6.5 ([19], [1], [17], [24], [9], [37], [34], [38]) and can even exceed 100 in some cases ([6], [3], [32]). However, there have been some case studies ([5], [2], [4]) in which little or no increase in post-fire discharges were observed.

In general, the first years after the event are the most critical in terms of hydrological instability ([26], [8]), even though in some cases an increased runoff discharge has been observed up to a decade following a wildfire ([20], [27]). Rowe et al. ([36]) analyzed how the ratio of unit-area peak discharge changed in pre- and post-fire conditions over 70 years in a case study in southern California; they measured as little as 2-fold increases for infrequent storms and as much as 40-fold increases as a consequence of frequent storms. Such ratio tended to diminish year by year after fire occurrence, being the time required to recover to pre-fire conditions depending on fire severity, affected forest type and post-fire meteorological conditions ([21], [22], [30], [12]). In light of these findings, it is evident that the standard rainfall-runoff methods applied in engineering practices need to be modified in order to model more accurately the change in watershed response following a fire ([14]).

In the present study, the effects of a wildfire on the hydrological parameters used for calculating runoff are assessed for a small urban basin in central Italy (San Giuliano, L’Aquila). Specifically, peak discharges in both pre- and post-fire conditions were calculated using the SCS-CN method, which estimates runoff by considering soil types, land cover and the influence of previous soil moisture conditions. Fire effects on peak discharges were then modeled by adjusting CN values according to existing approaches reported in the literature ([15]).

A Geographic Information System (GIS) - a powerful tool in hydrological modeling due to its capacity to handle large amounts of spatial and attribute data - was used for creating, managing and generating different layers and maps of aggregate hydrological parameters from different sources.

  Methods and materials 

Study area

The San Giuliano basin is part of the L’Aquila District hydrological system, bordered on the north by the Monti della Laga mountain range, on the east by the Gran Sasso Massif, on the south by the Monte Sirente range and on the west by the Monte Velino range. These form a natural barrier against atmospheric perturbations originating from the Adriatic and Tyrrhenian seas, resulting in a significant reduction in precipitation.

The basin can be divided in two sub-basins, whose characteristics are shown in Tab. 1: drainage area A, urbanization extent A_urb (expressed as a percentage of the drainage area), mean height H and time of concentration T_c, calculated using Giandotti’s formula ([16]).

A regionalization procedure was applied to overcome the problem related to the limited availability of hydrological measurements through space and time. For this reason, seven rain gauge stations located in the L’Aquila area were considered to determine the regional rainfall depth-duration equation h = at ⁿ (coefficients a and n are given in Tab. 2) for three return-period storms (50, 100 and 200 years).

A topographic map (scale 1:5000), land-cover map (based on Corine Land Cover classification) and soil map were used to delineate watershed borders, identify types of land-use and extract soil information, respectively.

pre-fire conditions

The SCS-CN ([42]) model is one of the commonly used methods for estimating the surface runoff from watersheds. The infiltration losses are combined with surface storage by the equation (eqn. 1):

where Q [mm] is the accumulated runoff or rainfall excess, P [mm] is the rainfall depth and I_a [mm] is the initial abstraction, which includes surface storage, interception and infiltration prior to runoff in the watershed, as expressed by the empirical equation (eqn. 2):

The term S is the potential maximum retention, given by (eqn. 3):

where CN [dimensionless - ranging from 0 to 100] is derived from data reflecting land cover, hydrologic soil group (A, B, C and D) and antecedent soil moisture condition (AMC, an index of soil wetness - [42]).

Area-weighted CNs for the two sub-basins were computed using GIS software, according to the scheme shown in Fig. 1, by combining information from the land cover map (Fig. 1a) and the hydrologic soil group map (Fig. 1b), which was produced by assigning a particular hydrologic soil group (A, B, C or D) depending on the soil’s minimum infiltration rate. Subsequently, to create a CN map (Fig. 1c), the hydrologic soil group field from the soil theme and the land use field from the land cover theme were selected for intersection. After this operation, a new polygon shapefile indicating the merged soil hydrologic group and land cover themes was generated. In accordance with USDA ([42]), the appropriate CN value was assigned to each polygon of this new map under the hypothesis of AMC II, which is essentially an average moisture condition. Data extracted from the new polygon attribute table were used to compute area-weighted CNs of the study area, equal to 69.4 for sub-basin 1 and 70.0 for sub-basin 2.

Fig. 1 - Use of GIS for the calculation of area-weighted CN in the San Giuliano basin. (a) Land cover map; (b) hydrological soil group map; (c) CN map.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

post-fire conditions

A serious wildland-urban interface fire recently destroyed 30% and 10% of the forest area in sub-basins 1 and 2, respectively. As shown in Fig. 2, about three years after the event, the largest difference between burned and unburned areas is represented by the lack of grass cover and underbrush.

Fig. 2 - Effect of fire in San Giuliano basin. (a) Unburned areas; (b) burned areas.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

Peak discharge was expected to increase as a consequence of the fire. Post-fire conditions were then modeled by adjusting pre-fire CNs to post-fire CNs as a function of burned area and fire severity (low, moderate or high).

Despite the general consensus on the issue that peak discharge tends to increase as a result of fire, at present there is no consistent methodology to estimate post-fire CNs, since the analysis of the hydrologic response of watersheds to wildfire is still a topic under investigation ([40]) and limited to few studies have been carried out in the United States. Currently, existing approaches are not supported by post-fire data analysis and are mainly based on practical rules and/or experience ([15]). For example, some of these approaches are reported in the Burned Area Emergency Response (BAER) manual ([15]) and include the following:

(1) Wildfire Hydrologic Impact (WHI - [23]): post-fire CNs are calculated as a function of the WHI classification (severe/moderate/low, related to the percentage of the sub-basin characterized by high soil burn severity) and pre-fire CNs.

(2) BAER Hydrology Special Report ([18]):

• High burn severity: CN_post = CN_pre + 15
• Moderate burn severity: CN_post = CN_pre + 10
• Low burn severity: CN_post = CN_pre + 5

where CN_post is the post-fire CN and CN_pre is the pre-fire CN.

(3) BAER Design Storms ([15]):

• High burn severity with water repellent soils: CN_post = 95
• High burn severity without water repellent soils: CN_post = 90 to 91
• Moderate burn severity with water repellent soils: CN_post = 90
• Moderate burn severity without water repellent soils: CN_post = 85
• Low burn severity: CN_post= CN_pre + 5

It is worth noting that these rules are site-specific, strictly applicable to the regions in which they were developed, or for areas characterized by similar precipitation regimes and pre-fire CN values.

For the case study under investigation, since neither a detailed burn severity map nor site-specific studies for the evaluation of post-fire CNs were available, all three approaches listed above from the BAER Manual were used, considering each level of burn severity, to calculate the modified CNs (under the simplifying hypothesis that the whole burnt area would have been affected to the same degree of severity); this was done in order to perform a sensitivity analysis for post-fire conditions while examining the effects of the variability in model input parameters (estimates of post-fire CNs according to existing rules from the literature) upon model output (peak discharges related to different return period storms in post-fire conditions).

Subsequently the GIS procedure shown in Fig. 1 was updated by overlaying the burned area map on to the land cover map and adjusting CNs for those polygons affected by the fire: this resulted in the generation of new post-fire CN maps based on the application of the various approaches for estimating post-fire CNs (Fig. 3).

Fig. 3 - Use of GIS for the calculation of CN in post-fire condition. (a) Overlay of burned area map on land cover map; (b) Maps with post-fire CN values estimated according to the different approaches.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

  Results and discussion 

Tab. 3 shows the results of peak discharge calculations using the SCS-CN method in sub-basins 1 and 2 for the three return period storms in pre-fire conditions. The same procedure was applied for calculating the expected change in post-fire discharges, with CN values estimated using the different approaches reported in the previous section. Tab. 4 and Tab. 5 present the calculated area-weighted post-fire CNs and peak discharges, as well as the fire effect ratio (calculated by dividing post-fire peak discharge by pre-fire peak discharge), which can be considered as a global parameter for describing the altered watershed response to fire.

Results showed that area-weighted CNs in post-fire conditions vary depending on the approach used: Wildfire Hydrologic Impact (WHI), Baer Hydrology Special Report (BAER HSR) or BAER Design Storm (BAER DS), for each different fire severity. For sub-basin 1, the CNs ranged (considering low to high burn severity) from ~71 to ~78, while they ranged from ~70.5 to ~72 for sub-basin 2. This variability was also reflected in fire effect ratios, which ranged from ~1.2 to ~2.8 (for sub-basin 1) and from ~1.1 to ~1.5 (for sub-basin 2) for the 50-year return period storm, diminishing to ~1.2-2.3 (for sub-basin 1) and ~1.1-1.4 (for sub-basin 2) for the 200-year return period storm.

It should be noted that these estimates refer to the first years after the fire, since post-fire CNs are expected to attenuate during the recovery period, gradually approaching the pre-fire values, correlated with a proportional reduction in the fire effect ratio. Obviously, the recovery period is tightly related to the rate of vegetation recovery and further depends on the types of plant species existing in pre-fire conditions, on the hydrological characteristics of the area and on fire severity.

The high variability of the results obtained using different approaches indicates the urgent need to perform validation analyses and/or new experimental observations in order to increase the accuracy of the estimated post-fire CN values and consequent runoff.

  Conclusions 

The effects of wildfires on the hydrological response of watersheds can be modeled by adjusting the hydrological parameters involved in traditional rainfall-runoff models to account for changes induced by the fire (e.g., CNs or runoff coefficients adjusted as a function of burned area and fire severity). Various approaches have been reported in the literature; however, due to the lack of any consistent and verified guidelines, the selection of the method to use and the way in which parameters are adjusted to account for post-fire conditions entails a substantial degree of subjectivity, affecting the reliability of the model output.

For the investigated area, since local studies allowing post-fire CNs to be calculated with a high degree of certainty were not available (as in many other areas affected by wildfires), we analyzed the variability in the estimates of post-fire discharge obtained using the various approaches reported in the literature for modeling post-fire conditions. Given the subjectivity involved in choosing one approach or another, the potential range of results obtainable considering the different approaches was assessed. The range of fire effect ratios calculated using the aforementioned methodologies was very large (from 1.1 to 2.3), indicating the unreliability of estimates when it is not possible to verify the output in the field. Further experimental research is necessary in order to reduce the gap in the knowledge required to accurately estimate and validate post-fire hydrological parameters.

  References