Impact of Biomass Burning on Aerosol Size Distribution, Aerosol Optical Properties and Associated Radiative Forcing

The influence of biomass burning on aerosol size distributions, particle number and radiative forcing has been studied at a rural site in Spain. It has been found that air contaminated by aerosols from biomass burning presents four times the total number of particles registered in non-contaminated air. In the case of the smallest fraction of the fine mode, between 0.1 and 0.2 µm, the increase soars to over seven times the total number of particles. An analysis of the evolution of the count mean diameter in the fine mode (CMD f ) in the 8 daily measurements has revealed a decrease of over 25% in this parameter in the modified measurements when compared with measurements that were not contaminated by aerosols from biomass burning. In contrast, when the aerosol transport time is long, the increases detected in CMD f range between 15% and 100% when compared with measurements of air by non-aged aerosol from biomass burning. Shortwave radiative forcings have been calculated for these high loads of fine aerosols with GAME (Global Atmospheric Model) software. For the August event, the daytime average of surface radiative forcing is –66 (±30) W/m 2 , and at the top of the atmosphere the forcing is –32 (±12) W/m 2 . Induced daytime average of atmospheric radiative forcing reaches 34 (±20) W/m 2 . The study demonstrates that wildfires affect not only the number of particles and the size distribution, causing a clear increase in the number of aerosols in the atmosphere, but they are also responsible for altering the local radiative balance.


INTRODUCTION
Biomass burning, both natural and man-induced, is an aggressive phenomenon which releases large amounts of particles and gases into the atmosphere, altering its composition at local, regional and even global scales. The consequences that may be triggered in the atmosphere by biomass burning depend mainly on the type of fuel and its characteristics, the relative humidity and the wind conditions (Reid et al., 2005).
Together with fossil fuel burning, biomass burning is the main source releasing carbon aerosols into the atmosphere (IPCC, 2001). During combustion, the elements released cause an increase in the concentrations of particles belonging to the fine mode (particles less than 1 µm) (Cachier et al., 1995;Remer et al., 1998;Reid and Hobbs, 1998;Dubovik et al., 2002), since between 80 and 90% of the particles released by biomass burning correspond to this mode (Reid et al., 2005). These are mainly organic particles, containing mostly organic carbon (OC) and black carbon (BC) (IPCC, 2001), with inorganic traces of sulfates, nitrates, inorganic nutrients and metals, which all together amount to approximately 10% of the particle mass (Cachier et al., 1995).
The different aerosol-mixing states lead to important variations in the hygroscopicity of particles from biomass burning (Cocker et al., 2001;Vakeva et al., 2002;Aklilu et al., 2006;Engelhart et al., 2012). Engelhart et al. (2012) have observed a clear positive correlation between the hygroscopicity of these particles and the inorganic mass fraction of the aerosol. The values of the hygroscopicity parameter, k found range between 0.006 (slightly hygroscopic) and 0.6 (highly hygroscopic), where k represents a quantitative measure of aerosol water uptake characteristics. It is defined through its effects on the water activity of the solution.
Many studies have focused on the hygroscopicity of aerosols from burning processes and their efficiency as Cloud Condensation Nuclei (CCN) (Warner and Twomey, 1967;Eagan et al., 1974;Roberts et al., 2002;Martins et al., 2009). Warner and Twomey (1967), Eagan et al. (1974) and Roberts et al. (2002) have concluded that, in the case of supersaturations of over 0.5%, particles from wildfires may act as condensation nuclei. Petters et al. (2009) have found that aerosols from burning processes are CCN active irrespective of the changes they may undergo later in the atmosphere. Engelhart et al. (2012) have found a convergence of the efficiency of aerosols from biomass burning in the formation of CCN after photochemical ageing processes.
Indirect and semi-direct radiative forcings estimates are part of the main current concerns and it is expected that the uncertainties on estimated radiative and climatic impacts should be reduced and our understanding of it improved (IPPC, 2007). Direct radiative forcings have been also widely documented depending on aerosol sources (Hodzic et al., 2007;Mallet et al, 2008Mallet et al, , 2009Malavelle et al., 2011). The inferred significant decrease of solar energy at the surface may strongly modify the surface energy budget. Latent heat and sensible heat fluxes from the surface are consequently disturbed as shown by Feingold et al. (2005) or Pere et al. (2011). The reduction in surface latent and sensible heat fluxes associated with biomass burning impact could reduce cloudiness (Jiang and Feingold, 2006).
In addition to the climate changes that may be caused by aerosols generated by biomass burning, we must also consider the negative impact of aerosols onto human health as they result in poorer air quality. The entrance and deposition of these particles in the respiratory tract depends on the size/surface proportion of the particles and on the ageing conditions involved (Oberdärster et al., 2005). The smaller particles are the ones that enter the respiratory system more easily, reaching even the alveolar region (EPA, 2002). The conditions of the respiratory tract, with a relative humidity close to saturation, trigger a number of changes in the particles inhaled. Londahl et al. (2008) have studied the increase in size of aerosols from biomass burning as a result of the high levels of humidity in the respiratory system. It was found that the hygroscopic increase suffered by the particles from combustion processes coincides with the minimum size required for the deposition of these particles in the respiratory tract.
The current study shows the changes in aerosol size distribution registered in a rural area during four summer days with intense biomass burning events. The study events have been registered in the province of León, Spain, and the distance between the wildfires and the sampling point ranged between only a few km and a maximum of 90 km.
Once we identified the cases in which the smoke plumes from the various wildfires (a clearly defined and located source of aerosols) reach the sampling point, the study of the aerosol size distributions revealed the origin and the type of air mass that keeps the particulate matter in suspension before it reaches the sampling point (Calvo et al., 2010a). Afterwards, we estimated the optical properties of the aerosols using the Mie theory. Thus, the Single Scattering Albedo (SSA) and the asymmetry parameter (g) were estimated by the GAME (Global Atmospheric Model) radiative transfer model to compute the instantaneous clearsky direct radiative forcing (Calvo et al., 2010b).

STUDY ZONE
The province of León is the largest in the region of Castilla y León, Spain, with a total land area of 15,581 km 2 . The climate is Mediterranean, with Continental influences, and with a few traces of Atlantic influence too, thus leading to a wide range of different landscapes. In the summer, the province of León is one of the areas most widely affected by wildfires, considering both the number of fires and the total land area burned.
Most of these fires are human-induced. The statistics provided by the regional government Junta de Castilla y León state that 90% of all the wildfires registered are caused by man, be it intentionally or as a result of carelessness.
The measurements analyzed in this paper have been carried out in the district of Carrizo de la Ribera (42°35′6″N 5°49′53″O), 26 km to the north-west of the city of León, Spain (42°35′59″N 5°34′18″O). This is a rural area on the right bank of the River Órbigo, at an altitude of 873 m.
The weather station installed by the Spanish National Agency for Meteorology (AEMET, in its Spanish acronym) in this area (42°35′N, 5°39′W) reflects a Mediterranean but continentalized climate, with a clearly marked seasonality. More detailed information about this study zone has been reported by Alonso-Blanco et al. (2011).

Measurement Equipment
To determine the aerosol size spectrum in this rural area a probe was installed in a field about two km from the town (42°35′59.90′′N, 5°50′50.92′′W). This laser spectrometer, a Passive Cavity Aerosol Spectrometer Probe, PMS Model PCASP-X, measures particle optical diameter nominally ranging between 0.1 and 10 µm, by considering the light dispersion properties of the particles at a wavelength of 633 nm collected between the angles 35° and 135°. The probe classifies the particle diameter into 31 channels, i.e., 31 intervals of discreet particle sizes. The device was calibrated by the manufacturer using polystyrene latex (PSL) particles of a known size. The refractive index of latex beads (1.59-0i) is different from that of atmospheric particles, resulting in an aerosol size distribution that is "PSL size equivalent".
In this study, we are presenting PCASP-X size distributions adjusted using Mie theory and implemented with a computer code developed by Bohren and Huffman (1983). In the case of the measurements of air that was not contaminated by aerosols from biomass burning, the correction was carried out according to the refractive index typical for rural aerosol, whose values vary depending on the relative humidity (Kim and Boatman, 1990). And in the case of the measurements of air that was contaminated by aerosols from biomass burning, the object of this study, the adjustment was carried out considering a specific refractive index for aerosols from biomass burning.
The study period comprises 4 days, the 13 th and 14 th of August and the 3 rd and 4 th of September 2001. The particle size spectrum was measured 8 times daily, during 10-minute intervals, every 3 hours. The specific refractive index for aerosols from biomass burning was computed by interpolating the real and the imaginary part according to the relative humidity at the time of the measurement (values between 34% and 100%). For the measurements of air containing aerosols from biomass burning the refractive index used has been 1.49-0.013i, obtained by Calvo et al. (2010b) from the AERONET sunphotometer retrieval at 675 nm for the wildfire detected in Palencia (41°N, 4°W) on the 8th of August 2003. The refractive indices for aerosols from biomass burning reported in the literature take a range of different values. This is due to the fact that the refractive index depends on the size of the aerosol, the mixing state, the wavelength (λ) at which the index is estimated and the water content of the particles (Seinfeld and Pandis, 1998). In the case of aerosols from biomass burning, one of the most important factors to consider when it comes to estimate the refractive index is the relationship between EC/OC, which determines in turn the relationship between absorption/dispersion of the radiation by the aerosol (IPCC, 2001). Guyon et al. (2003) have found an average refractive index -at a value of λ of 545 nm-of 1.41-0.013i for particles from biomass burning in the Amazon rainforest. Dubovik et al. (2002) have estimated for biomass burning smoke in the Amazon forest -for the visible spectrum of wavelengths-a refractive index with a real part with values between 1.47 and 1.52 and an imaginary part with values between 0.00093 and 0.021. The refractive index found for aerosols in Palencia (1.49-0.013i), the one used in this study, is located within this range of values found by Dubovik et al. (2002).
In our case study, the choice of the refractive index was determined by the proximity of the study zone to Palencia, an area with a similar topography and vegetation, both natural and of anthropic origin, and with similar characteristics in the wildfires occurring in both areas.
It was also necessary to carry out several corrections on the number of counts sampled by the spectrometer in each channel. Firstly, the flow measurement value was set in relation to the altitude of the sampling point. In the present study the probe was installed at an altitude of 873 m and a correction factor of 0.905 was introduced. In addition, each measurement had to be adjusted with a correction factor to account for coincidence losses at high concentrations. These corrections are described in detail in Calvo et al. (2010b).
A weather station was installed next to the probe to register automatically data on precipitation, pressure, temperature, relative humidity and wind speed and wind direction. A wind profiler Sodar SR 1000 was also installed in the same field, with a pulse frequency of 5 tones, about 2150 Hz, a pulse power of 300 W, 8 s pulse-repetition time and a maximum range of 1250 m. This device provided an automatic recording of the wind vector in its three components. The data recorded by the weather station and the Sodar were stored every 30 minutes.

Thermal Inversions, Backward Trajectories and Circulation Weather Type Classification
Three additional aspects have been analyzed to implement the study: the thermal inversions, the backward trajectories of the air masses, and the circulation weather type classification. The thermal inversions were computed from the data obtained by the radiosoundings carried out in La Coruña (43.36°N, 8.41°W, altitude 67 m), Madrid (40.50°N, 3.58°W, altitude 633 m) and Santander (43.48°N, 3.80°W, altitude 59 m) at 1200 and 0000 UTC, and provided by the University of Wyoming (http://weather.uwyo.edu/upperair/ sounding.html).
In order to identify the weather type associated with a particular synoptic situation, a circulation weather type classification (CWTs) was set up based on Jenkinson and Collison (1977) and Jones et al. (1993). These procedures were developed to define objectively Lamb weather types (Lamb, 1972) for the British Isles. The daily circulation affecting the Iberian Peninsula is described using a set of indices associated to the direction and vorticity of the geostrophic flow. The indices used were the following: southerly flow, westerly flow, total flow, southerly shear vorticity, westerly shear vorticity and total shear vorticity. These indices were computed using sea level pressure values obtained for 16 grid points distributed across the Iberian Peninsula. This method allows for a maximum of 26 different CWTs. Following Trigo and DaCamara (2000) in their study for Portugal, this study does not have a separate class for unclassified days, but has opted for disseminating the fairly few cases with possible unclassified situations (< 2%) among the retained classes. This classification has been used previously for the Iberian Peninsula with a number of different applications, such as the study of lightning (Tomas et al., 2004), splash erosion , aerosol size distribution in precipitation events Calvo et al., 2012) or in studies on precipitation (Fernandez-Gonzalez et al., 2012).
To establish the provenance of the air masses reaching the study zone, back trajectories of 120 h (5 days) have been used, computed at 500 m, 1500 m and 3000 m AGL with the HYSPLIT model -HYbrid Single-Particle Lagrangian Integrated Trajectory Model-by NOAA (Draxler and Rolph, 2003). The HYSPLIT model (Draxler and Hess, 1998) carries out a wide range of simulations related to the longrange transport, dispersion and deposition of air pollutants. The model has been developed by the NOAA Air Resources Laboratory in Maryland, USA. This study will be complemented by a description of these air masses.

Radiative Forcing: Global Atmospheric Model (GAME)
Aerosol-related optical parameters such as Atmospheric Optical Depth (AOD), Single Scattering Albedo (SSA) and the asymmetry parameter (g) are necessary to perform radiative transfer calculations. The last two parameters listed above are not measured directly, so we estimate them via the Mie theory by taking into account the measured aerosol size distribution and the estimated refractive index, representative of a mean chemical composition for biomass burning aerosol. In that way, optical properties are estimated as though the aerosol population was internally mixed. SSA and g are estimated at the seven wavelengths used in the GAME radiative transfer model. As for the AOD parameter, we have used MODIS data for August 13 th and 14 th 2001 and for September 3 rd and 4 th 2001 over the study region at a wavelength of 550 nm.
Once the necessary parameters have been gathered, the spectral optical properties are modeled. Table 1 gives the single scattering albedo and the asymmetry parameter for specific wavelengths (400 nm, 550 nm and 950 nm) and for total size distribution.
The instantaneous clear-sky direct radiative forcing has been estimated with the GAME radiative transfer model detailed in Dubuisson et al. (2004). GAME accounts for the scattering and absorption processes by particles and gases. Gaseous absorption (H 2 O, CO 2 , O 2 and O 3 ) is treated from the correlated k-distribution, using a line by line code (Dubuisson et al., 1996), and multiple scattering effects are treated using the Discrete Ordinates Method (DOM) (Stamnes et al., 1988). This method allows for accurate treatment of scattering and absorption by aerosols, clouds and molecules. Upward and downward net radiative fluxes are calculated over the spectral solar range, from 2,500 to 50,000 cm -1 , with a 100 cm -1 spectral resolution. Calculations of radiative fluxes integrated over the entire shortwave region are performed at the specific sampling time.
From these fluxes, we have computed the aerosol clearsky daily direct forcing at the bottom of the atmosphere (BOA), ΔFBOA, and at the top of the atmosphere (TOA, 20 km in this case), ΔFTOA, defined as mentioned below. The first one represents the effect of particles on the net shortwave radiation fluxes reaching the surface and the second one the radiation fluxes reflected back to space by aerosols. ΔBOA and ΔTOA forcings have been calculated as follows: where FBOA(w) ↓ and FBOA(o) ↓ are, respectively, the downward net radiative fluxes simulated at the surface with (w) and without (o) aerosols. FTOA(w) ↑ and FTOA(o) ↑ are, respectively, the upwards net radiative fluxes simulated at the top of atmosphere with (w) and without (o) aerosols. With this convention, a positive sign of ΔF implies a warming effect. Finally, we compute the atmospheric forcing, ΔFATM, by using the following relation: where ΔFATM represents the possible absorption of solar radiation due to the absorbing properties of fire particles within the atmospheric layer where aerosols are located.

Methodology
The Department for the Environment of the regional government Junta de Castilla y León provided a database with information on the number of fires per day, the district where each fire occurred, the date of detection and extinction (with the exact day and time), and the total land area affected in hectares (ha) with the type of vegetation burned. This information was used to draw maps of the province of León with all the districts affected by wildfires on the study days and the days immediately before, i.e., on the 12 th , 13 th and 14 th of August and on the 2 nd , 3 rd and 4 th of September 2001.
The maps and the data on the wind direction at surface level and a certain altitude registered by the weather station and the Sodar enabled us to determine which measurements carried out by the PCASP were affected by the transport of smoke plumes from the surrounding wildfires.
The particle records of the 8 daily measurements were compared with the life time of the wildfires, from the time of detection until the time of extinction. Considering the evolution in the number of particles, it was possible to establish the time at which the smoke plume arrived at the sampling point because the corresponding measurements presented very large numbers of particles, reaching a fourfold increase when compared with the other measurements carried out the same day but not affected by the smoke. Once the measurements that might have been affected by the smoke plumes had been identified, the first data selection was obtained by checking the wind speed and direction at that time and the distance between the wildfire and the probe. Subsequently, the second data selection was made with the non-parametric test of Mann-Whitney for a significance level of 0.05 (Essenwanger, 1986), with the purpose of identifying those measures of air affected by aerosols from wildfires. For that reason, the measures previous to those affected by aerosols from wildfires have been used as background.
The weather conditions on the study days have been characterized by analyzing the atmospheric stability considering the circulation weather type, thermal inversions and the corresponding air masses.
The aerosol size distribution on the study days has been analyzed in detail by checking the evolution of the count mean diameter in the fine mode and the number of particles corresponding to this size range in the 8 daily measurements. The clear-sky instantaneous direct radiative forcing of the aerosols released has been studied too, both at the bottom of the atmosphere (ΔFBOA) and at the top of the atmosphere (ΔFTOA), as well as the atmospheric forcing (ΔFATM).

Meteorological Analysis
The meteorological characteristics of the 4 study days, 13 th and 14 th of August and 3 rd and 4 th of September 2001, are shown in Table 2. The average monthly temperatures are typical of the summer months. The highest values are found on the 13 th and 14 th of August with 22.1°C and 18.9°C, respectively. Lower values were recorded on the 3 rd and 4 th of September, with 16.5°C and 15.9°C, respectively. The maximum and minimum temperatures follow a similar trend: around 30°C and 15°C, respectively, for the days in August, and around 25°C and 7°C, respectively, for the days in September. In general, the temperature parameters show a decreasing trend along the summer, typical for this season in these latitudes.
The daily precipitation registered in the study period is very low: only 0.2 mm in the two study days in August. The average relative humidity ranges between 60 and 77%, with the highest values registered on the 3 rd and 4 th of September. The average wind speed was moderate, with less than 2 m/s on average every day. On the 4 th of September the wind speed registered remained under 1 m/s.

Circulation Weather Types, Thermal Inversions and Air Masses
The CWT classification shows that during the study days of August and September (Table 2), the CWTs were three: the purely northern type (N) on the 13 th of August, the purely north-western type (NW) on the 14 th of August, and the purely north-eastern type (NE) on the 3 rd and 4 th of September. The air masses that reach the Iberian Peninsula on the study days came from the north. However, on the 13 th and 14 th of August these air masses had a clearly maritime influence, and on the 3 rd and 4 th of September they had a clearly continental influence.
On the 14 th of August no thermal inversions were registered in the soundings carried out in Madrid and Santander at 0000 UTC and there are no data on this issue in the sounding carried out in La Coruña at 0000 UTC. In contrast, in the other 3 study days thermal inversions are common, both radiative and subsidence inversions at altitudes under 1000 meters AGL. There are even moments when both types coexist, thus generating an isolating layer that delays the dispersion of the smoke plumes from the various wildfires. Consequently, there is a high concentration of aerosols in the layer closest to the surface. As for the soundings carried out at 1200 UTC in La Coruña, Madrid and Santander, no data for the study days are available.
The back trajectories ( Fig. 1) have drawn for the study days at three altitudes (500, 1500 and 3000 m AGL) reveal different situations. On the 13 th of August the high pressure systems around the Iberian Peninsula generate a saddle point in the study zone. On that day, the back trajectories show a clear atmospheric instability ( Fig. 1(a)) and the aerosols registered at two altitudes, 500 and 1500 m, come from the surface (in the 6 hours immediately before the meaurement). This indicates that the aerosols from the wildfires joined the air mass that reached the probe. On the 14 th of August the situation tends to stabilize; the air masses bring to the study zone marine and continental aerosols, and at 500 m there are also surface level aerosols in the 6 hours immediately before the arrival of the air mass to the sampling site ( Fig.  1(b)). On the 3 rd and 4 th of September (Figs. 1(c) and 1(d)) the air mass reaching the Iberian Peninsula is a maritime polar (mP) air mass. This air mass brings with it maritime aerosols and is characterized for being cold and wet, thus explaining the decrease in temperature and the increase in the relative humidity registered on these days when compared to the study days in the month of August.

Characterization of the Wildfires
This section describes the wildfires registered during the study days (13 th and 14 th of August and 3 rd and 4 th of September), as well as the ones registered on the previous days, the 12 th of August and the 2 nd of September, respectively. These wildfires altered the local atmospheric composition and some of them contaminated the aerosols measurements carried out on the 13 th and 14 th of August and on the 3 rd and 4 th of September.
On the 12 th of August, 10 wildfires were registered and a total land area of 244 ha was burned. Of these fires, 5 were very small (the burned surface is S < 1 ha), and 5 were medium-sized (500 ha > S > 1 ha). Three of the mediumsized fires originated to the north-west of the study zone, within a radius of 90 km, and burned 239 ha, 98% of the land area affected that day. All the fires detected on the 12 th of August were extinguished that same day.
On the 13 th and 14 th of August, 19 and 18 wildfires were registered, respectively. A total of 37 fires burned 517 ha. Of these, 26 were very small and 11 of a medium size. Of the medium-sized fires, 2 started within a radius of 75 km to the south-west of the study zone and burned 93% of the surface affected those days, a total of 479 ha. Both fires were detected on the 13 th of August and were extinguished on the 14 th .
On the 2 nd of September, the day immediately before our two study days (3 rd and 4 th of September), 7 wildfires burned a total area of 143 ha. Of these, only one was of small size and the other 6 were medium-sized. Two of the mediumsized fires started within a radius of 95 km to the north-west of the study zone and burned 134 ha, 94% of the surface burned that day. Both fires started and were extinguished on the 2 nd of September, coexisting in time.
On the 3 rd and 4 th of September there were a total of 11 wildfires: 9 on the 3 rd of September and 2 on the 4 th of September, burning together 65 ha. Of the total number of wildfires, 5 were small and the rest medium-sized. Two of the medium-sized fires burned 65 ha, 80% of the surface burned those days. Both fires were detected on the 3 rd of September, within a radius of 75 km, to the south-west of the study zone. The medium-sized fire affecting a smaller area, 17 ha, was extinguised on that same day, but the other one affected a much larger area, 35 ha, and could not be extinguished until the 4 th of September.
In general, the wildfires detected during the study days ranged during daytime, between 1200 UTC and 2200 UTC. Few fires are detected outside this time range. The wildfires registered on these days are mainly located to the west and south-west of the study area. The zones affected are mostly in the mountains, at altitudes of around 1000 m, surrounded by valuable landscapes.
The smoke plumes of the medium-sized wildfires affecting considerable amounts of land may be carried long distances by the wind (Fraile et al., 2006), at least to a regional scale. In our study they are responsible for the important increases in the number of aerosols registered in the sampling point, installed in a rural area.

Influence of Biomass Burning on Aerosol Measurements Aerosol Size Distributions
The multi-log-normal function was used to characterize the size distributions of aerosol particles (e.g., Whitby, 1978;Hoppel et al., 1994). It thus becomes very easy to compare several data sets of aerosol particles. The multilog-normal concept is thoroughly described in the literature and the overall outcome has proven to be useful whenever parameterizations are required (e.g., Mäkelä et al., 2000;Birmili et al., 2001). It was found that the size distributions analyzed by the PCASP-X were bimodal: with a fine and a coarse mode. Fig. 2 shows the aerosol size distributions on the study days. During the wildfires registered on the study days, we observed a clear increase in the number of particles smaller than 0.2 µm in the measurements contaminated by the smoke plumes. Increases of up to seven times were recorded in this size range. The same situation was found by Janhall et al. (2010) in fresh smoke from biomass burning,   recording an increase in the number of particles in the fine mode by an arithmetic mean with a standard deviation of the Geometric Mean Diameter of 117 ± 13 nm. Kleeman et al. (1999), in a study on particle mass distributions of fresh smoke from wood burning, estimate a mode centered in a range of particle size between 0.1-0.2 µm. The application of non parametric test of Mann-Whitney measures allowed us to identify those ones affected by aerosols from wildfires. Since the increases observed in the number of particles correspond to minor sizes of 0.2 µm, the test has been executed for the first five channels corresponding to this size. Thanks to these results, it has identified five measures of air contaminated on the 13 th of August, one measure on the 14 th of August and one and three measures on the 3 rd and 4 th September, respectively.
During the event comprising the 13 th and 14 th of August, the measurements affected by wildfires are the ones recorded during the night on the 13 th of August (0100 UTC and 0400 UTC) and on the 14 th of August (0100 UTC) and the central hours of the 13 th of August (1000 UTC, 1300 UTC and 1600 UTC). Particles concentration average increases of 300% were registered in air contaminated by aerosols from biomass burning when compared to previous uncontaminated measurements. At 0100 UTC on both days the measurements exceeded a 400% increase in the particles concentration ( Fig. 2(a)). These increases are even more important in the case of particles in the size range between 0.1 and 0.2 µm, exceeding 700% in the particles concentration in some of the measurements. Of particular importance are the two processes of contamination by aerosols from biomass burning registered on the 13 th of August. The first of these events occurred in the early hours, at 0100 UTC and 0400 UTC, and the figures registered were 35,492 and 19,882 particles/cm 3 , respectively. Three hours later, at 0700 UTC, the figures registered were 1734 particles/cm 3 . This decrease reflects clearly the evolution of the smoke plume from biomass burning, from its arrival to its dispersion and the subsequent return to normal levels of atmospheric particulate matter.
A similar situation is observed in the second contamination process on the 13 th of August, between 0700 UTC and 1600 UTC. At 0700 UTC the number of particles registered is 1734 particles/cm 3 . The arrival of a smoke plume to the probe causes a considerable increase in the number of particles at 1000 UTC: 7728 particles/cm 3 . The number of particles remains constant at 1300 UTC, with 7712 particles/cm 3 , but at 1600 UTC there is a significant decrease in the number of particles: 2546 particles/cm 3 . The number of particles of sizes between 0.10-0.12 µm starts increasing at 1000 UTC, with 4873 particles/cm 3 , more than four times the number registered in that size range at 0700 UTC. At 1300 UTC this increase in the number of particles spreads to the ones between 0.1 and 0.2 µm, with 2308 particles/cm 3 , 40% more than the ones registered at 1000 UTC in the same size range. The return to normal begins at 1600 UTC, with 2444 particles/cm 3 , a drop of 67% in the number of particles smaller than 0.2 µm (Fig. 2(b)) when compared to the measurement carried out at 1300 UTC.
On the 13 th of August, the measurements were affected by the wildfire in Villagatón, in the district of Astorga. This fire started on the 12 nd of August at 1346 UTC and lasted until 2159 UTC the same day. During this period the wind direction was mainly N with an average speed of 1 m/s. The smoke plume took approximately 8 hours to reach the sampling point after the beginning of the fire. In contrast, the measurements affected by aerosols from biomass burning registered on the 14 th of August corresponded to a wildfire which started in Santa Colomba de Somoza, also in the district of Astorga, on the 13 th of September at 1930 UTC. This fire was extinguished on the 14 th of August at 2000 UTC. The average wind speed registered during the fire at the measurement site is 1.7 m/s, with three dominating directions: S, N and NNE. The smoke plume took around 6 hours to reach the probe after the fire started (Table 3).
Figs. 2(c) and 2(d) represent the situation on the 3 rd and 4 th of September. On the 3 rd of September, two measurements were contaminated by aerosols from biomass burning, the one at 0400 UTC, with 9005 particles/cm 3 , and the one at 2200 UTC, with 32,961 particles/cm 3 . The increases in the number of particles were of 60% and 500%, respectively, when compared to the previous uncontaminated air measurements. The measurement carried out at 2200 UTC includes aerosols from a wildfire in Castrillo de la Cabrera, in the district of Truchas (Table 3), which began on the 3 rd of September at 1145 UTC and was extinguished at 1800 UTC that same day, burning a total of 17 ha. During the active life of the wildfire, at surface level the average wind speed is of 3 m/s, mainly of the directions NNE and WNW, so the smoke plume reached the probe in around 6 hours from the beginning of the fire. At 0400 UTC the measurement was affected by aerosols from biomass burning, but it was not possible to determine which fires contaminated that record. It may be the case that aerosols from other fires were registered here too, including fires from the 2 nd of September around the study zone, within a radius of 38 to 92 km, since the number of particles measured on that day, excluding the other two contaminated measurements, is of 4147 ± 1394 particles/cm 3 , the highest value in the four study days.
On the 4 th of September, the wildfire in Castrillo de la Cabrera, in the district of Truchas, is the reason for the high number of particles measured in the early hours of that day, with increases of over 500% when compared with previous measurements, at 0100 and 0400 UTC, registering 34,171 and 12,090 particles/cm 3 , respectively. At 0700 UTC, the measurement immediately after the one at 0400 UTC, there were only 7017 particles/cm 3 . In other words, as on the 13 th of August, the evolution of the number of particles at these times (0100 and 0400 UTC) indicates the arrival of a smoke plume from biomass burning and the later return to the normal levels of particulate matter.
The considerable increases in these days in certain measurements cannot be explained by the presence of a high atmospheric stability, nor by the simultaneous occurrence of radiative and/or subsidence thermal inversions. The only reason is the arrival of smoke plumes from the wildfires registered on those days.
In general, the measurements affected by the wildfires during the study period were the ones carried out during the night (between 2200 UTC and 0400 UTC) and in the central hours of the day (between 1000 UTC and 1600 UTC). The results indicate that the wildfires release to the atmosphere particles of the submicrometric aerosol fraction (fine mode), as shown by Rissler et al. (2006). A direct relationship was found between the occurrence of certain wildfires and the increase in the number of particles in some of the measurements. We claim that in the four study days the important wildfire activity is influencing the concentration of the number of particles at a local level, altering more or less the 8 measurements carried out during those days.
The graphs corresponding to the surface and volume size distributions (Fig. 3) and the calculation of Volume Median Diameter (VMD) and Surface Median Diameter (SMD) have been made. In general, it has been observed that the distribution area and volume of air measures affected by aerosols from fires were increased for particles < 0.2 µm and the SMD and VMD are shifted to smaller sizes, indicating that the particles emitted by wildfires correspond to small sizes.

Daily Evolution: Count Mean Diameter (CMD f, ) and Number of Particles in the Fine Mode (N f )
The role of biomass burning in the characteristics of the particulate matter on the four study days was studied by comparing the evolution along the 8 daily records of the count mean diameter (CMD), and the number of particles. As mentioned above, biomass burning processes cause changes in size distributions, mainly in the smaller fraction of the fine mode. The estimations carried out to determine the daily number of particles in the coarse mode (1-10 µm) indicated that there were very few particles in this size range, around 3 particles/cm 3 ; therefore, the results focus exclusively on data corresponding to the fine mode (in this study, with PCASP-X only particles between 0.1-1 µm), considering the evolution of the CMD f and the total number of particles (N f ) (Figs. 4 and 5).
On the 13 th of August, the concentration of the number of particles was higher during the night than during the day. The particles registered during the evening also had a larger CMD f , reaching its maximum at 1900 UTC with a size of 0.15 µm. In daytime the minimum is reached at 1600 UTC with 0.07 µm. All the records presented a high geometric deviation, i.e., particles of very different sizes coexist on the same day.
As for the number of particles in the fine mode, the first measurement contaminated by aerosols from biomass burning was registered at 0100 UTC. In this record the estimation is N f = 94,025 particles/cm 3 . At this time, the CMD f drops from a value of 0.12 µm at 2200 UTC on the 12 th of August to a value of 0.09 µm at 0100 UTC on the 13 th of August.
In the following measurement at 0400 UTC the CMD f was 0.10 µm, with a decrease in N f of 98% (Figs. 4(a) and 4(b)). Between 0700 UTC and 1900 UTC the second contamination process occurred with the arrival of particles from biomass burning. The estimations for N f were 2716, 3271, 13,629, 7507 and 828 particles/cm 3 for the measurements taken at 0700 UTC, 1000 UTC, 1300 UTC, 1600 UTC and 1900 UTC, respectively. At 0700 UTC and at 1000 UTC the CMD f was 0.10 µm, later dropping to 0.08 and 0.07 µm at 1300 UTC and 1600 UTC, respectively. By 1900 UTC there was a clear return to normal values, with 0.15 µm.
The data clearly illustrate the fact that the measurements were affected by two smoke plumes from biomass burning, and that they later returned to normal background levels in the content of particulate matter after the smoke plumes dispersed. In addition, the measurements contaminated by aerosol from biomass burning presented a decrease in the CMD f of more than 25%. Other studies on the CMD of aerosols from biomass burning found slightly higher values. Le Canut et al. (1996) found a CMD of 0.125 ± 0.02 µm in emissions of fresh smoke from wildfires in the savannahs in northern Africa. Janhall et al. (2010) found a CMD of around 0.120 µm in fresh smoke for three fuel types: forest, savanna and grass. The size of the aerosols during the combustion processes depends mainly on the type of fuel, the availability of oxygen, the temperature of the combustion, and on the phase in the combustion process (flaming phase or smoldering phase) (Reid et al., 2005). The uncertainty generated by instrumental techniques must also be taken into account. On the 14 th of August, during the night, the most important oscillations both in the number of particles and in their sizes were found. The concentration of particles in the fine mode remained more stable during the daytime (0700 UTC to 1900 UTC) (Figs. 4(c) and 4(d)). The minimum CMD f was reached at 0100 UTC, estimating a value of 0.09 µm, coinciding with the main increase in N f that day: 11,749 particles/cm 3 . The high concentration of particles in the fine mode, together with the low values of the CMD f , which dropped from 0.12 µm at 2200 UTC on the 13 th of August 2001 to 0.09 µm at 0100 UTC on the 14 th of August (a 25% decrease) points towards a clear air contamination by the presence of aerosols from biomass burning.
The maximum CMD f was registered at 0400 UTC with 0.17 µm. During the whole day there was a marked heterogeneity in the size of the particulate matter, with large and small particles coexisting. But at 0400 UTC the low geometric deviation (σ g ) with respect to the previous measurements (1.39 µm) pointed towards more uniform sizes, with the larger fraction of the fine mode dominating. The reason for this situation may have been a greater atmospheric stability during the night leading to the ageing of the aerosols from biomass burning registered at 0100 UTC.
On the 3 rd of September the CMD f was higher during the night than in daytime, with the maximum CMD f recorded at 2200 UTC, with a value of 0.14 µm. The minimum CMD f was found between 1000 UTC and 1600 UTC, with a value of 0.10 µm. It was in this time interval that the lowest geometric deviations of the fine mode were registered (1.53 µm). In other words, the particles found in the atmosphere in these hours corresponded to the smallest fraction of the fine mode and had more homogeneous particle sizes. With respect to the number of particles in the fine mode, there were two clearly distinct time intervals: one during the night (from 2200 UTC to 0700 UTC), with a concentration of particulate matter around 1161 ± 163 particles/cm 3 , and another one during the day (from 1000 UTC to 1900 UTC), with a higher number of particles estimated at 2971 ± 679 particles/cm 3 (Figs. 5(a) and 5(b)).
The estimated N f for the measurements contaminated by aerosol from biomass burning (0400 UTC and 2200 UTC) were 1254 and 1221 particles/cm 3 , with a CMD f of 0.12 and 0.14 µm, between 15% and 100% higher than the CMD f values in the other contaminated measurements in the study period. As mentioned before, on the 3 rd of September there was a high concentration of particulate matter during the whole day. The only direct influence observed in the measurement taken at 2200 UTC was the wildfire in Castrillo de la Cabrera, which started on the 3 rd of September. The low number of particles in the fine mode, together with the high values of the CMD f , seems to indicate that the aerosols measured suffered a longer transport from wildfires that occurred on previous days. The atmospheric stability during the night favored aerosol ageing. Some studies focus on characterizing aerosol growth by ageing, associated with long-range transport of the smoke plume. Radke et al. (1995) found that during the transport of the smoke plume from the wildfires, the geometric median diameter (GMD) of the aerosols in the fine mode increases to more than 0.25 µm after 25 h of transport. A more recent study by Eck et al. (2003) found that, when the smoke plume remains isolated during the transport and this transport lasts for several days, aerosol ageing processes occur and the values of the peak volume modal radius (r) registered lie around 0.20 µm.
On the 4 th of September the CMD f oscillated during the day, with an increasing trend towards the evening. The minimum CMD f was registered at 1000 UTC with 0.10 µm, and the maximum CMD f at 2200 UTC with 0.16 µm. The geometric deviation of the fine mode indicated that during the day the particulate matter had more uniform sizes (Fig. 5(c)).
As in the case of the other two study days, the 13 th and 14 th of August, there was an increase in the number of particles in the fine mode at 0100 UTC, with 12,749 particles/cm 3 . This increase coincided with a decrease in the CMD f in this measurement. The values changed from 0.14 µm at 2200 UTC on the 3 rd of September to 0.10 µm at 0100 UTC on the 4 th of September. This value tended to increase, and we found a CMD f of 0.12 µm at 0400 UTC and 0.13 µm at 0700 UTC, with an estimated N f of 2488 particles/cm 3 and 1907 particles/cm 3 , respectively ( Fig. 4(d)). At 0100 UTC and 0400 UTC there was a decrease of around 20% in the CMD f when compared with the measurement taken at 2200 UTC on the 3 rd of September, which was not contaminated by aerosols from biomass burning. The CMD f values obtained that day for aerosols from biomass burning were similar to the ones found by Reid et al. (2005) in their study on fresh smoke from various types of biomass burning. This study found that the CMD of the aerosols ranged between 0.1 and 0.13 µm. As mentioned above, the evolution of the CMD f and the N f on this day revealed the arrival of a smoke plume from biomass burning and the subsequent return to normal levels afterwards. During the rest of the day the number of particles remained stable with a mean value of 1490 ± 620 particles/cm 3 .
The four study days presented very similar trends, with a clear boundary between nighttime (from 1900 UTC, when dusk set in, until 0700 UTC, when the next day dawned) and daytime (from 0700 UTC to 1900 UTC). In the four study days the nocturnal CMD f presented an increasing trend, whereas during the day it tended to decrease. The same situation was found in a previous study by Alonso-Blanco et al. (2011), and it may be linked to a higher atmospheric stability, thus favoring aerosol ageing processes as the particles remain in the atmosphere for longer periods of time.
This situation changed only when the measurements are contaminated by aerosols from biomass burning. At that point, the number of particles estimated for the fine mode (N f ) soared to the highest values registered those days. This situation coincided with average decreases in the CMD f of over 25% when compared with previous uncontaminated measurements. After the dispersion of the smoke plume, the CMD f returned to the usual values registered before the arrival of aerosols from biomass burning. In other words, the smoke plumes from the wildfires were releasing into the atmosphere particles belonging to the smallest fraction of the fine mode. With prolonged aerosol transport times, as on the 3 rd of September, we found increases in the CMD f between 15% and 100% when compared with uncontaminated measurements and with shorter transport times (13 th and 14 th of August and 4 th of September).

Aerosols from Biomass Burning in Air Quality Control Regulations
The concentration of PM 10 particulate matter was estimated taking 1.35 g/cm 3 as the density of particles from biomass burning (Reid and Hobbs, 1998), and assuming that ambient air particles in rural areas have a density of 1 g/cm 3 , as found in previous studies by Thatcher and Layton (1994) and Riley et al. (2002), who analyzed particulate matter indoors and outdoors in urban and rural areas. In measurements that are not contaminated by aerosols from wildfires the estimated concentrations were under 20 μg/m 3 and in contaminated measurements the estimated concentrations may be of up to 90 μg/m 3 , higher than the daily maximum threshold value (24 hours) of 50 μg/m 3 established by European and Spanish Regulations on Air Quality Control (Real Decreto 102/2011 de 28 de enero). This concentration should not be exceeded more than 35 times per year. Barnaba et al. (2011), in an exhaustive study carried out in Europe on the contribution of wildfires to atmospheric particulate matter, establish a preliminary estimation of the average monthly contribution of aerosols from biomass burning during the months with more wildfires at levels of 10 µg/m 3 . This indicates that wildfires may influence legal controls on air quality.

Influence of Aerosols from Biomass Burning on Aerosol Optical Properties and Associated Radiative Forcing
First, considering the optical properties of aerosols (Table 1) we note the spectral dependence of SSA with high values for short wavelengths. It follows then that the aerosol is mainly light scatterer rather than absorber. Nevertheless, SSA values vary all day long, so the absorption property of the aerosol also varies. Indeed, positive atmospheric radiative forcings (Table 4) result in atmospheric heating by absorbing particles.
Our simulations enabled us to investigate the surface forcing during the fire events. The instantaneous surface forcing estimated by GAME reached up to -98 W/m 2 and -82 W/m 2 during the August and September fire events, respectively (Table 4). The daytime average was -66(±29) W/m 2 for a mean AOD of 0.5 and a refractive index of about 1.48-0.005i during the August event, and it was -57 (±23) W/m 2 for a mean AOD of 0.4 and a refractive index of about 1.45-0.004i during the September event. Such values are consistent with those simulated by Formenti et al. (2002), from the STAAARTE MED 1998 experiment. Indeed, for an AOD of about 0.3, for a refractive index of 1.55-0.025i, over sea surface and over sea plus vegetation surface, daytime averages of direct radiative forcing at the surface are, respectively, -55 (±20) W/m 2 and -49 (±16) W/m 2 . These average values were of the same order. The lower aerosol loadings in the column 'atmosphere' of the STAAARTE MED experiment (represented by the lower AOD) seems to be compensated by a more absorbing aerosol as the imaginary part of the refractive index is higher than ours. Instantaneous ΔFTOA (Table 4) were computed by GAME, reaching up to -51 W/m 2 and -42 W/m 2 during the August and September fire events, respectively. The daytime average of direct radiative forcing at TOA was about -32 (±12) W/m 2 and -30 (±10) W/m 2 , respectively. For comparison, these instantaneous values were found to be higher than those obtained by Formenti et al. (2002), in the case studies described above, with daytime averages of direct radiative forcing at TOA, respectively, of -21 (±7) W/m 2 and -12 (±5) W/m 2 . This difference might be due to the fact that in our case SSA values stand for scattering aerosols. For most of our case studies the aerosol forcing was negative (cooling effect) and much higher at the surface than at the TOA due to absorption.
Besides, positive instantaneous ΔFATM given by GAME illustrate the absorption of solar radiation in the atmosphere by the smoke aerosol considered (Table 4). It reached up to +60 W/m 2 and +45 W/m 2 during the August and September fire events, respectively. These instantaneous values were consistent with those found by Sicard et al. (2012) for biomass burning in Barcelona (+60 W/m 2 ). Our daily average values (Table 4) were consistent with those reported by Hodzic et al. (2007), who found that this forcing could reach daily average values of +40 W/m 2 in the fire source region.
Concerning the daily variations of these radiative fluxes, we observed that the direct atmospheric radiative forcing reached maximum values when the fire events occurred at a time of day when the solar radiation available is high. For example, the aerosol number concentration is 6 times higher at 1300 UTC than during the morning for the fine mode on the 13 th of August 2001. These high loads of aerosols in the fine mode interacted with shortwave radiation through absorption and scattering processes, which was translated by a lower downward radiative flux reaching the surface (ΔFBOA highly negative). The daily trends of radiative fluxes of our case studies were consistent with those found by Lyamani et al. (2006), with both radiative forcings at TOA and BOA increasing (absolute values) with a solar zenith angle of up to about 60° and decreasing absolute values with a solar zenith angle greater than 60°.

CONCLUSIONS
In rural areas wildfires may be considered the main sources of high loads of particles in the atmosphere and clearly alter particle size distributions. The present study has registered increases in the total number of particles of over 4 times when compared with uncontaminated measurements. This increase corresponded mostly to particles in the fine mode, mainly to sizes between 0.1 and 0.2 µm, where the values recorded were up to seven times the values in measurements not contaminated by aerosols from biomass burning. In other words, wildfires release into the atmosphere particles belonging to the smallest fraction of the fine mode.
Excluding the measurements of air contaminated by aerosols from biomass burning, in the 4 study days the count mean diameter of the fine mode (CMD f ) showed a clearly decreasing trend in the registers taken during daytime. In contrast, during the night there was an increase in the CMD f . It could be due to the processes of ageing or hygroscopicity and deliquescence provoked by the increase of the relative humidity during the night. This difference between daytime and nighttime measurements may be motivated by actual atmospheric dynamics or thermodynamics. During the day the convective processes are more intense, whereas during the night an increased atmospheric stability results in less ventilation, favoring aerosol ageing processes. This situation is only altered with the arrival of aerosols from wildfires, resulting in an increase in the number of particles in the fine mode and a decrease in the CMD f of more than 25% when compared with the measurement taken immediately before the arrival of the smoke plume. It was also observed that when the aerosols from biomass burning remained in the atmosphere for longer periods of time, the ageing processes and the aerosol hygroscopicity and deliquescence, resulted in increases of the CMD f between 15% and 100% when compared to contaminated measurements where no ageing processes were detected.
Wildfires are causing atmospheric pollution in rural areas, with PM 10 concentrations of up to 90 μg/m 3 , exceeding by far the daily threshold value of 50 μg/m 3 established by the European and Spanish Regulations for Air Quality Control.
Calculations of surface, top of the atmosphere and atmospheric radiative forcings with the GAME radiative transfer model reveal a strong attenuation of the shortwave incoming solar radiation at the surface during fire events. Due to high loads of fire aerosols in the atmospheric column related to high AOD and increases in fine mode aerosol concentrations, the instantaneous atmospheric radiative forcings reach high positive values (between 26 and 34 W/m 2 ) translating an important absorption. These different forcings of direct radiative effects may affect the heating rate in the atmosphere and induce major thermodynamic and microphysical feedbacks. Latent and sensible heat fluxes at surface as well as microphysics in clouds could be modified.