Do fire severity effects on soil change in space and time in the short-term? What ash tells us

FLAMMA, 5 (1), 23-27, 2014 ISSN 2171 - 665X CC Attribution-NonCommercial-ShareAlike 3.0 License 23 Do fire severity effects on soil change in space and time in the short-term? What ash tells us P. Pereira (1*), A. Cerdà (2), X. Úbeda (3), J. Mataix-Solera (4), A. Jordán (5), D. Martin (6), P. Mierauskas (1), V. Arcenegui (4), L.M. Zavala (5) (1) Department of Environmental Policy, Mykolas Romeris University, Vilnius, Lithuania (2) Department of Geography, University of Valencia, Valencia, Spain (3) Department of Physical Geography and Regional Geographic Analysis, University of Barcelona, Barcelona, Spain (4) GEA-Environmental Soil Science Group, Department of Agrochemistry and Environment, Miguel Hernández University, Alicante, Spain (5) MED_Soil Research Group, Department o Crystallography, Mineralogy and Agricultural Chemistry, University of Seville, Sevilla, Spain (6) USGS, 3215 Marine Street, Boulder, Colorado, USA *Corresponding author: pereirau[email protected] Keywords Abstract Ash redistribution Ash color Fire severity Fire severity index Spatial correlation In the absence of data, the impact of fire, especially wildfires, is measured analysing the fire severity. This post-fire assessment is very useful because allow to identify the degree of destruction imposed by the fire. Among the techniques used to determine fire severity, ash colour is often used, that permit identify the degree of organic matter consumption (darker ash uncompleted combustion, lighter ash completed combustion). The objective of this paper was observed if fire severity changes in space and time, according to ash colour analysis, applying an index. The ash colour analysis was carried out one and fifteen days after the fire. In this area we identified ash with four different colours, black (B) dark grey (DG), light gray (LG) and white colour (W) and some uncovered areas classified as bared soil (BS). Black and DG represent medium fire severity, LG and W, higher severity. The results showed that in the studied fire, the severity was high and a great part of the plot was uncovered by ash (BS). Fifteen days after BS increased as the fire severity index, from 6.05 to 6.45, showing that during this period the ash redistribution in a short period after the fire can influence the fire severity assessment. We did not identified significant differences between measurements and the coefficient of variation (CV%) remained the same. However significant differences were identified with the spatial correlation analysis with Global Moran's I and the spatial structure of fire severity index. This is evidence that ash color changed in this period in the space and the traditional statistical methods did not detected, only with spatial analysis. The analysis of fire severity using ash color some days after the fire can induce important errors, because wind can (re)mix ash and a particle produced in one area can be easily exported to other. Received: 24 October 2012 | Accepted: 14 May 2013 1 INTRODUCTION Fire severity is an indirect measurement of fire effect on the ecosystem. In absence of real data, difficult to be collected during wildfires, indirect estimations are frequently carried out, as soil organic matter content, soil hydrophobicity, minimum branch diameter, crown scorch, fine fuel combustion, among others (Pereira et al., 2012). Ash colour is a current method to estimate fire severity. The degree of combustion change ash physical and chemical properties. These changes are an indirect effect of the temperatures reached, which depends also of the plant species and ecosystem affected (Pereira, 2010). Ash is highly mobile, especially in the immediate period after a fire when can be wind redistributed uncountable times in a complex manner, especially in severe wildland fires, where ash is easily transportable due to the intense combustion (Pereira et al., 2013). Only after the first rainfalls ash can bind onto soil surface, infiltrate or erode. FLAMMA | Vol. 5 | 1 | 23-27 24 kewness (Skew) and Kurtosis (Kur). Differences between sampling periods were observed with the non-parametric Wilcoxon Matched Pairs Test. Spatial autocorrelation of fire severity in each period was assessed with Global Moran's I, that is similar to the Pearson correlation coefficient. The index values fall between -1 and +1. Negative correlations shown that, data was spatially dispersed and positive correlations that data was clustered. This index calculates also the Z-value. A positive Z-value represents that samples are clustered and negative that are dispersed. Lower positive and negative values represent a random pattern. Significant differences were considered at a p<0.05. Previous to data modelling, data normality and homogeneity of the variances were tested. Several data transformations were considered, as neperian logarithm, Figure 1. Distribution of ash colour a) one day after the fire b) fifteen days after the fire. The ash mobility can be a problem to assess the fire severity, since the ash analysed in an area could be not produced in the same place. However, it is also an opportunity to study the impacts of fire severity (different ash characteristics) on soil cover and understand that the same ash particulate can have implications in different parts of the soil profile and surface. From this point of view, fire severity can change in space and time. The objective of this work is to study the ash colour changes in an experimental plot in the immediate period after a fire. 2 METHODOLOGY 2.1 STUDY AREA AND SAMPLE COLLECTION A wildfire occurred in July 26 of 2010 and affected an area of 100 ha near the urban area of Quinta do Conde (Portugal), located at 38° 57’ N, 09° 05’ W and 115 m of altitude. The geological substrate was composed by Plio- Pleistocene low cementation dunes, and the soils are classified as Podzols (IUSS Working Group WRB, 2006). The mean annual temperature is 14.8 oC and the annual precipitation of 639.2 mm. The burned forest was mainly composed by Pinus pinaster trees. One day and fifteen days after the fire, ash colour was observed visually in four parallel transects with 20 m separated by one meter in a south faced uniform slope with an angle of inclination of 17%. In each transect, ash colour were observed with a resolution of 50 cm. A total of 200 samples were collected per sampling period. This qualitative assessment, do not allow to quantify the degree of fire severity observed in each point. Thus, fire severity was (re)classified. A previous study by Pereira (2010) showed that ash produced at low severity has a reddish colour (R), at medium severity, black (B) and dark grey (DG) colour, and at high severity light gray (LG) and white colour (W). Also, after high severity fires patches of bare soil (BS) were found, due the high combustion temperatures that consume all the litter. In the present case this is true because in the contiguous area not affected by the fire, continuous thick accumulations of litter were identified. The ash identified was B, DG, LG and W. Some areas were bare. To quantify fire severity effects on soil cover, some indexes were applied, 1 to B, 3 to DG, 5 to LG, 7 to W and 9 to BS (no protection). 2.2 STATISTICAL AND SPATIAL ANALYSIS Some descriptive analysis were performed, mean (m), Standard deviation (SD) Coefficient of variation (CV%), FLAMMA | Vol. 5 | 1 | 23-27 25 These results highlight to the necessity of spatial statistical analysis that detect changes, not identified by traditional statistical analysis (paired comparison tests and CV%) as it will be observed in semi-variogram parameters in the Table 2. One day after the fire the best-fitted model was the Exponential and a Nug/sill ratio of 76.46% that represents a weak spatial dependency and a random distribution. Fifteen days after the fire, the best-fitted model was the Gaussian and the Nug/sill ratio was 44.40% showing a moderate spatial dependency and a lower spatial variability of fire severity. The coefficient of determination (R2) was also higher fifteen days after the fire were higher fifteen days after the fire, indicating that the spatial correlation was high in the second measurement, as observed previously with the Global Moran I index. The nugget effect represents the error of the variance that can be attributed to the small scale variance, outliers or reduced number of samples. In this study, the number of samples was representative of the studied area, the outliers were not possible to minimize (even after several data transformations) and the small scale variability of fire severity is an intrinsic characteristic of burned areas, where combustion conditions change very quickly due to different fuel conditions (e.g.: moisture, type, density, package, connectivity, etc.). In this study the outliers were not removed because they are of major importance for analysis, and were attributed to the heterogeneous burning conditions. One day after the fire the small scale variability of fire severity was higher than fifteen days later, as shown the high nugget effect. The reduction of the small scale variability is due the ash redistribution and homogenization of areas with a determined severity index. Here, there was an increase of the BS areas, which means that soil became unprotected in some patches across the plot. In other areas ash was deposited by the wind as it is represented in Figure 2. One day after the fire the small scale variability of fire severity was observed in some areas of the northwest, central and south of the plot. There is no spatial pattern and the map of fire severity confirms the previous spatial autocorrelation and semi-variogram analyses. Fifteen days after the fire it is observed that fire severity effects on soil cover is more “patched”. In some areas at south and north, ash was removed and in others accumulated as in southeast and in the central-north part of the plot (Figure 2). box-cox and square root. However, none of them was able to normalize data. In this case, non-transformed data was always more close of the normality and homogeneity of the variances were higher than the transformed, in all the methods. Thus for semi-variogram modelling and data interpolation we used non transformed data. In order to observe the spatial dependence of the fire severity index it was calculated the nugget/sill ratio. If the ratio is less than 25%, the variable has strong spatial dependence, between 25% and 75%, the variable has moderate spatial dependence, and greater than 75%, the variable shows only weak spatial dependence (Chien et al., 1997). In this work data was interpolated with the ordinary kriging method. Statistical analyses were carried out with Statistica 7.0 and interpolation with Surfer 9.0. 3 RESULTS AND CONCLUSIONS Immediately after the fire the majority of the ash identified was LG and a great part of the plot was uncovered, showing that fire had a high severity (Figure 1A). Fifteen days after the fire the area with BS increased, as the mean fire severity index (6.05 to 6.45), showing that during this period, wind had redistributed the ash and therefore the fire severity effects and the potential estimation of fire severity through ash colour (no rainfall occurred between the measurement periods). Despite this redistribution, the CV% remains the same and no significant differences between fire severity index were observed (Wilcoxon Matched Pairs Test p=0.084) (Figure 1). The Global Moran's I showed that one day after the fire the spatial autocorrelation was 0.13 and not significant at a p<0.05, indicating that fire severity had a random distribution. Fifteen days after the fire the spatial correlation was 0.35, and significant at a p<0.05, that represents an important modification of the spatial distribution and fire severity (Table 1). The Z-score also increases importantly, confirming the previous hypothesis. Table 1. Results of Global Moran’s I spatial autocorrelation test. One day after burn Fifteen days days after burn Moran's Index 0.13 0.35 z-score 1.92 5.03 p-value 0.06 0.0001 FLAMMA | Vol. 5 | 1 | 23-27 26 Table 2. Best fitted semi-variogram models of fire severity and corresponding parameters. Day Model Nugget effect Partial sill Sill Nug/sill ratio (%) Range (cm) R2 1 day Exponential 4.45 1.15 5.60 79.46 100 0.52 15 days Gaussian 2.66 3.33 5.99 44.40 160 0.71 Figure 2. Ordinary kriging maps of fire severity index a) one day after the fire and b) fifteen days after the fire. ACKNOWLEDGMENTS The authors appreciate the support of the project “Litfire”, Fire effects in Lithuanian soils and ecosystems (MIP- 048/2011) funded by the Lithuanian Research Council, Spanish Ministry of Science and Innovation for funding through the HYDFIRE project CGL2010-21670-C02-01, FUEGORED (Spanish Network of Forest Fire Effects on Soils http://grupo.us.es/fuegored/) and to Comissionat per a Universitats i Recerca del DIUE de la Generalitat de Catalunya. However, ash mobility is very complex to evaluate and in this stage further studies are needed to have better understanding of ash dynamic after the fire. In ash studies there are many uncertainties that further studies need to be focused. One of the most important is the fact that after the fire, wind and water can mix ash produced at different severities, producing ash with a new colour and different severity from the previous ones. This can be a source of error in this work that has to be accurate in the future. New methodologies are being developed to have a better picture of ash mobility. FLAMMA | Vol. 5 | 1 | 23-27 27 REFERENCES Chien YL, Lee DY, Guo HY, Houng KH. 1997. Geostatistical analysis of soil properties of mid-west Tailand soils. Soil Science 162: 291–297. IUSS Working Group WRB. 2006. World reference base for soil resources 2006. A framework for international, classification and communication. World Soil Resources Report No 103. Rome. Pereira P. 2010. Effects of fire intensity in ash chemical and physical characteristics of Mediterranean species and their impact in water quality, Phd thesis. University of Barcelona, Barcelona. Pereira P, Cerdà A, Úbeda X, Mataix-Solera J, Martin DA, Jordán A, Burguet M. 2013. Spatial models for monitoring the spatio-temporal evolution of ashes after fire – a case study of a burnt grassland in Lithuania. Solid Earth 4:153–165. Pereira P, Úbeda X, Martin DA. 2012. Fire severity effects on ash chemical composition and water extractable elements. Geoderma 191: 105–114.