Article pubs.acs.org/est
Influence of Climate and Land Use Change on Spatially Resolved Volatilization of Persistent Organic Pollutants (POPs) from Background Soils Jiří Komprda,*,† Klára Komprdová,† Milan Sáňka,† Martin Možný,‡ and Luca Nizzetto†,§ †
RECETOX (Research Centre for Toxic Compounds in the Environment), Kamenice 753/5, CZ-625 00 Brno, Czech Republic Czech Hydrometeorological Institute, Doksany Observatory, 411 82 Doksany 105, Czech Republic § Norwegian Institute for Water Research, Gaustadalleen 21, NO-0349, Oslo, Norway ‡
S Supporting Information *
ABSTRACT: The subject of this study is the assessment of the influence of climate and land use change on the potential re-emission of organochlorine pesticides (OCPs) from background and agricultural soils. A deterministic spatially and temporally explicit model of the air− surface exchange was created, fed with distributed data of soil and atmospheric concentrations from real measurements, and run under various scenarios of temperature and land use change for a case study area representative of central European conditions. To describe land use influence, some important features were implemented including effect of plowing, influence of land cover, temperature of soil, and seasonal changes of air layer stability. Results show that volatilization of pesticides from soil largely exceeded dry gas deposition in most of the area. Agricultural soils accounted for more than 90% of the total re-emissions both because of the generally higher soil fugacities (higher loads of chemicals and relatively low organic carbon content), but also due to physical characteristics and land management practices enhancing the dynamics of the exchange. An increase of 1 °C in air temperature produced an increase of 8% in the averaged total volatilization flux, however this effect can be neutralized by a change of land use of 10% of the arable lands to grassland or forest, which is consistent with projected land use change in Europe. This suggests that future assessment of climate impact on POP fate and distribution should take into consideration land use aspects.
■
INTRODUCTION Concentrations of persistent organic pollutants (POPs) in the environment continue to decline;1,2 however, due to their persistence and affinity for organic carbon, a significant amount of these pollutants is stored in soils and sediments.3−9 POPs can be released from these reservoirs back to the atmosphere and readdressed to long-range transport. The most important processes of remobilization are in fact volatilization from the oceans’ water and soil surfaces.10−13 In recent years, an increasing number of data sets of soil POP concentration became available thanks to the establishment of regional monitoring networks.14−18 These data allow a qualitative and quantitative evaluation of the distribution processes between soil and air to be performed at regional level.19−23 Several distribution models were recently used to identify the impact of climate change on the fate of POPs in the environment.24−27 The majority of the studies concluded that the mobility of POPs in the environment will increase due to re-emission from soils as a result of increased air temperatures. Long-term temperature trends will influence the total amount of pollutants that are released from secondary sources.24 This enhanced re-emission, however, may not be © 2013 American Chemical Society
consistent in different areas, because it depends on a combination of land cover characteristics,28−30 soil characteristics, land management practices, and physical−chemical properties of substances.31,32 Climate change can induce shifts in land use by affecting structures and functions of ecosystems or by affecting land management in a number of interconnected ways including both natural (changed hydrology, shift in nutrients availability, shifts of habitat distribution, etc.) and social−economical drivers (namely: human adaptation to climate change). Arguably, land use change implications for POP fate and distribution have not yet been under sufficient focus. The goal of this study is to assess, on a spatially resolved way, the combined influence of increasing temperature and shifts in land use and management on the re-emission of two Special Issue: Rene Schwarzenbach Tribute Received: Revised: Accepted: Published: 7052
November 29, 2012 March 6, 2013 March 18, 2013 March 18, 2013 dx.doi.org/10.1021/es3048784 | Environ. Sci. Technol. 2013, 47, 7052−7059
Environmental Science & Technology
Article
Soil Temperature. Fate models often assume soil to be in thermal equilibrium with air. This may lead to substantial miscalculations of pivotal parameters controlling partitioning and exchange of POPs between soil and air which are exponentially related to temperature (namely, equilibrium partitioning coefficients and diffusion coefficients). Soil temperature is the result of a complex heat balance in which air temperature is only one of the involved parameters. Deterministically solving this balance involves knowledge of a number of parameters which are often not available for spatially explicit approaches. However empirical approaches are available which allow estimating, in first approximation, soil temperature from air temperature in a range of different conditions. Here, soil temperatures were calculated using the SoilClim model.34 The model works with daily time step and requires six meteorological parameters: global solar radiation (MJ m−2 day−1), maximum and minimum air temperature (°C), precipitation (mm), water vapor pressure (kPa), and average wind speed (m s−1). Generally surface soil temperature is higher than air temperature especially on bare land and grassland. In contrast, forest soil is mostly colder than air, with the exception of particularly low air temperatures or in case of snow cover. The highest soil temperature is in arable land followed by grassland and forests in warmer months while the opposite trend is observed in cold periods. Model of the Air−Surface Exchange. Calculation of pollutant volatilization from soil is typically based on a threeresistances model.35 Coefficients describing the properties of soil (namely, water, air, and organic matter content), the properties of the pollutants (namely, the octanol−water (Kow) and air−water (Kaw) equilibrium partitioning coefficients and their dependence on temperature) and coefficients describing the dynamics of pollutant transfer across the phase boundary, are required. In this study the calculation was based on the fugacity approach.36 Default values for air and water content in soil, 30% and 20%, respectively, were used in this model according to Mackay.36 The description of diffusive exchange between air and soil and associated parameters is based on the scheme used in the model POPCYCLING-Baltic.37 Briefly, the three-resistances classical model is based on the assumption that a pollutant on its way through the phase boundary between air and soil must overcome an air-side resistance plus two parallel resistances for the diffusion into the soil. These can be described in terms of the following conductances: the air-side mass transfer coefficient (MTC) (U1, m h−1), the MTC describing diffusion through the soil water (U2), and the MTC describing diffusion through air in soil pores (U3), respectively, as described in the following set of equations.
organochlorine pesticides (namely: hexachlorobenzene (HCB) and dichlorodiphenyldichloroethylene (DDE)) regulated under the Stockholm Convention on POPs) from soils. The approach presented here is innovative since it considers in a unique frame spatially explicit concentration data of these compounds in soils classified based on land use and elevation; the use of geostatistic methods for POP soil distribution assessment; and a state-of-the-art fugacity model to predict air−surface exchange.
■
MATERIALS AND METHODS Model Inputs. Pesticides HCB and DDE (as the prevalent metabolite of DDT in soil) were selected for the purpose of analyzing volatilization fluxes. It can be assumed that reemission from agricultural and forest land represents the major secondary atmospheric sources for these compounds (due to their previous direct application), while inhabited areas will play a rather minor role.9 Pollutants such as PCBs and PCDD/F can be also of interest for this analysis but their concentration in background soils is low and secondary emissions from urban areas and industrial sites may still be dominant.33 The parameters used in the model have different spatial or temporal dependence. Soil Concentrations. A spatially explicit assessment of soil− air exchange of POPs requires knowledge of distribution of compounds in the soil and the atmosphere. Although, unfortunately, these data are not available for the continental or global scale in sufficient resolution, a regional data set allowed the compilation of detailed high resolution inventories for a range of POPs. In this assessment we used the Czech Republic as a case study scenario since for this region, a highly resolved (1 × 1 km) data set of concentrations of several POPs in soil (including HCB and DDE) is available.9 The data set also includes soil organic carbon content, land use, and soil porosity data which were used to parameterize the model. In addition the Czech Republic soil type, land use, and environmental gradients well represent conditions for all central Europe. The data set refers to average soil concentration in the upper 20 cm of the soil core. Air Concentrations. Concerning air concentration, time series of pesticides measured weekly during year 2006 in Košetice background air station (Central Czech Republic) were used. Košetice observatory is representative of rural areas for central Europe. Air concentration was treated as a timedependent, space-independent parameter. This assumption is supported by considering the relatively small spatial scale of this assessment (78 867 km2) and the short residence time of air in this region (according to prevalent wind speed this is estimated to be typically around 30 h, which is in the range of the time resolution set here for the model outputs (namely 8 h)). Air Temperature. Air temperature was treated as a timedependent, space-dependent parameter. Time series of 1095 measurements (3 per day) were used for different altitude (in 100-m resolution steps from 200 to 1600 m asl). The data from 150 meteorological stations distributed throughout the Czech Republic which covered more than 160 000 measurements of air temperature provided by the Czech Hydrometeorology Institute observatories were used for calculation of time series. The values from stations in the same altitude were averaged, and distributed to all areas within the same altitude range. The model was configured to provide 8 hourly resolved estimates of the air surface exchange of POPs, therefore, time dependent inputs were assigned to this time step.
Das =
A 1 U1ZA
+
1 U2ZA + U3Zw
(1)
U2 =
BA νA10/3 h (νA + νw )2
BA = 1.8 e−2 m 2 h−1
U3 =
Bw νw10/3 h (νA + νw )2
Bw = 1.8 e−6 m 2 h−1
(2)
(3)
where Das is the overall transfer coefficient of air-soil (mol Pa−1 h−1); A is the area (m2); Ba and Bw (m2 h−1) are compound molecular diffusivity in air and water, respectively; νA and νw are volume fractions of air and water in soil, respectively; h (m) is 7053
dx.doi.org/10.1021/es3048784 | Environ. Sci. Technol. 2013, 47, 7052−7059
Environmental Science & Technology
Article
the soil depth relevant for gaseous exchange; and Za and Zw (mol m−3 Pa−1) are the fugacity capacities of air and water, respectively. Here the diffusion path in soil was corrected by tortuosity using the Millington−Quirk formula. According to Wania et al.37 the following set of parameters was chosen: U1 was set to 2.08 m h−1 over agricultural soils, while it was reduced by a factor of 5 over forest soils in order to describe the effects of canopy induced aerodynamic resistance. h was set equal to 1 cm for agricultural soil and 5 cm for grassland and forests, U1 was assumed to represent the maximum value of the air side MTC over agricultural land. In principle U1 depends primarily on wind speed, but also air temperature and solar radiation.38 For the sake of simplicity and the lack of spatially resolved high resolution data of wind speed no explicit relationships were traced between these parameters and the MTCs, however an empirical variation of U1 was introduced by assuming a lower value (factor of 3) during the winter. This factor reflects the influence of the stability of surface cooling in winter and the reduced solar radiation. In synthesis, the atmosphere during winter is more stable. A correction for the stability factor during the spring and autumn was obtained by linear interpolation of U1 values between summer and winter. These assumptions may appear largely arbitrary, however it must be acknowledged that there is a very big uncertainty around the parameterization and the value of air−soil MTC for POPs. The selected values have to be taken as representative of the order of magnitude of the process kinetics, in fact, overdetailing MTCs parametrization may drive to unfruitful complexity and equally low accuracy in model outputs. In addition, the adoption of this simplification has to be considered in the light of the main scope of this study which is to compare, on relative terms, the air−surface exchange flux in different scenarios affected by environmental perturbation, rather than predicting real volatilization data, which, on the other hand, are not experimentally available for comparison purposes. The volatilization flux N (mol h−1) of the selected compounds through the air−soil interface is detailed according to the fugacity approach calculated by multiplying Das and the fugacity of compound in air and soil f (Pa), as follows: Nnetflux = Nsoil − air − Nair − soil = fsoil Das − fair Das
POP fate models, although it would probably be appropriate to introduce the concept of SAI (surface area index), which would be equivalent to LAI (leaf area index) currently adopted in some fate models.40 The gaseous exchange flux of HCB and DDE between arable soils and air in the periods of September− October was calculated by correcting fluxes considering plowing by a factor described by the following equation: fplow = 3 − log10(d)
where d is the number of days after plowing. Basically eq 6 states that immediately after plowing (set here as default at the beginning of September), the volatilization flux from arable soils is enhanced by a factor of 3 (Figure S1), while it declines to its default values in about two months following a logarithmic decay expressing settling of soil surface. This behavior is consistent with observations performed on short and medium term trends of CO2 efflux from agricultural soils after plowing.41 Model Configuration. The model was run for the Czech Republic scenario at a spatial resolution of 1 × 1 km (for a total of 72 000 cells) consistently with the resolution of HCB and DDE distribution data.9 Time resolution for the calculation was set at 8 h and corresponded to the resolution of the air temperature measurements. The appropriate set of input parameters was assigned to each cell for a given time step, and the final value of the net flux was obtained by subtracting volatilization flux with dry gaseous deposition flux. It is important to note this model only addresses the calculation of soil−air exchange fluxes but does not integrate to solve soil or atmosphere mass balance of contaminants. In other words, soil concentrations are assumed to be constant and air concentrations are assumed not to be significantly affected by air−soil exchange at the regional scale. The assumption on soil concentrations stands if we consider that the fraction of these compounds effectively exchanged with the atmosphere generally represents a very small portion of the total burden of contaminants in the soil surface layer.42 As a result sensible soil concentration changes due to air−surface exchange occur over time scales of several decades even in situations in which soils received direct application of the chemicals (such as in the case of HCB and DDT).11,43 This assumption was also assessed in this study by comparing the calculated exchange fluxes integrated over one year and the overall pool of the contaminants in the soil. Pools of pesticides were calculated for top soil layer (0.20 m) based on concentration of pesticides,9 bulk density, and thickness of soil horizons according to Č upr et al.44 Case Study Area. The Czech Republic represents a relatively small area of 78 869 km2 but with a heterogeneous character covering many environmental gradients typical of the entire central European area. Altitude gradients spread between 115 and 1602 m asl with most of the area included between 250 and 700 m asl (average = 450 m). Mean annual air temperatures mostly vary between 5.5 and 9 °C with winter minima around −20 °C and maximal summer values of 30−35 °C. The Czech Republic presents a relatively large portion of the land covered by forest (32.6%) which is nearly comparable to arable land (41.8%). The remaining area is covered mostly by grassland (17.4%) and water, urban, and industrial land (8.1%). Typical soil reference groups for both forest and agricultural soils are Cambisols and Stagnosols. Other important soil types are Podzols and Regosols for forests, and
(4)
One of the most influential factors controlling volatilization from soil is soil temperature by means of the exponential function linking equilibrium partitioning coefficients on temperature.36 In this study a regression model was used to describe temperature dependence of partitioning coefficients in the form ln K xy = A + B /T
(6)
(5)
where T is soil temperature (K) while parameters A and B were taken from Mackay.39 Effect of Plowing. Another process included in the model, which can significantly affect the volatilization flux, is plowing. It has been observed that for certain compounds (including organochlorine pesticides), atmospheric concentrations show maxima not only in summer, which corresponds to the temperature maximum, but also in spring and autumn.2 This is likely due to the influence of land management and in particular plowing. This process basically results in increasing the soil specific surface available for evaporation as well as its porosity. Effect of plowing is not included in most of current 7054
dx.doi.org/10.1021/es3048784 | Environ. Sci. Technol. 2013, 47, 7052−7059
Environmental Science & Technology
Article
Figure 1. Spatial distribution of monthly integrated volatilization fluxes of DDE and HCB for the coldest and the warmest months (January and July).
instead describe flux changes normalized on the relative proportion of land use representation and they are therefore independent from land use size and represent pure effect of change. Standardized values are more useful to the scope of transposing results to other areas.
Chernozems, Fluvisols, and Luvisols for agricultural land. Prevailing soil texture is loamy. Concerning historical usage of pesticides, the Czech Republic (former Czechoslovakia), as many other European countries was a very active user of DDT and HCB and also an important producer of organochlorine pesticides until the early 1970s. Scenario Description. The relative influence of increased temperature and land use change on POP volatilization was studied using artificial model scenarios. Results were evaluated for combinations of three altitude classes identified as follows: lowland (L) < 500 m asl, highland (H) 500−800 m asl, high highland (HH) > 800 m asl, and individual land use types, including arable land, grassland, and forest. In total 7 scenarios were evaluated. Three were based on temperature change: scenarios T1, T2, and T3 where air temperature was increased by 1, 2, and 3 °C, respectively. The other four scenarios considered shifts of 10% in land use and included evaluations of short-term effects (e.g., organic carbon content in soil (Corg) does not change with land use, including scenarios AG, arable land to grassland, and GF, grassland to forest) and long-term effects (Corg changes to the average value of the new land use, including scenario AG_C, arable land to grassland, and GF_C, grassland to forest). These scenarios are consistent with projected European land use change.45 To simulate land use change the following parameters were manipulated: soil temperature parameterization, effective soil depth h, absence/presence of plowing (for land use shifts involving arable land) and, Corg (for the AG_C and GF_C scenarios). Refer to Supporting Information Table S1 for more details on parameter values. The percentage of change in volatilization flux calculated for the various land use scenarios over the total study area was calculated both in standardized and nonstandardized version. The nonstandardized values reflect conditions valid for the Czech Republic and they are influenced by proportional representation of land use in this region. Standardized values
■
RESULTS AND DISCUSSION Air−Surface Exchange in the 2006 Default Scenario. Simulation results showed that volatilization fluxes exceeded dry gaseous depositions for all land uses and during most of the time. As a result soils of the selected regional scenario serve as a net secondary source of HCB and DDE for the atmosphere. For the sake of clarity we define here the downward air-to-soil exchange as dry gaseous deposition carrying a negative sign; the upward soil-to-air exchange as volatilization carrying a positive sign; and their difference as net exchange. About 37.5 kg of DDE and 241 kg of HCB were calculated to be volatilized from the soils of the Czech Republic, annually under default conditions. Volatilization flux of HCB is nearly 6 times higher than that of DDE in spite of 1 order of magnitude higher DDE concentrations in soils. The total pool of HCB in the Czech soils is calculated to be 43 tons in contrast to 308 tons of DDE. If volatilization is considered as the only process driving pesticide clearance from soil it will take 178 years to minimize 2 orders of magnitude of the HCB content or 8200 years in the case of DDE. If we consider time of application of HCB and DDT (ca 1960−1975) in combination with other factors such as the development of inaccessible fraction of pollutant in soil and degradation time it appears that for HCB, volatilization is a relatively more important “clearance” mechanism in comparison to DDE. In addition DDE is a metabolite; therefore a net active input of this contaminant is represented by the biotransformation of DDT currently present in soil. Spatial variability of volatilization fluxes is presented in the form of maps for July and January (Figure 1). 7055
dx.doi.org/10.1021/es3048784 | Environ. Sci. Technol. 2013, 47, 7052−7059
Environmental Science & Technology
Article
Table 1. Quantitative Results of Volatilization, Concentration, and Pool of HCB and DDEa DDE volatilization land use grassland forest arable land total
HCB pool
concn
volatilization
pool
concn
N %
mean g/km2
sum CZ kg/year
sum CZ %
sum CZ kg
sum CZ %
mean ng/g
mean g/km2
sum CZ kg/year
sum CZ %
sum CZ kg
sum CZ %
mean ng/g
19 36 45
0.01 0.02 1.12 0.52
0.21 0.49 36.75 37.45
0.6 1.3 98.1
112623 75761 220526 307558
3.7 24.6 71.7
2.60 18.37 20.38 16.30
0.19 0.04 7.23 3.34
2.63 1.10 237.25 240.98
1.09 0.45 98.45
3576 4403 35100 43080
8.3 10.5 81.5
0.83 1.10 3.28 2.04
N%, percentage of individual land use in the Czech Republic; sum CZ (kg/year), sum of volatilization flux from area of land use of the Czech Republic for one year; sum CZ (kg), sum of pesticides pool in soil from area of land use of the Czech Republic; sum CZ %, percentage of sum of volatilization flux per year or sum of pools in soil for different land use; concn (mean ng/g), mean concentration in soil; volatilization (mean g/km2), mean of volatilization flux from 1 km2. a
Figure 2. Time series of daily mean volatilization flux of HCB in different types of land use.
grassland soils have a homogeneous structure in the upper 20 cm therefore we expect pesticide distribution to be relatively well mixed in the upper core. Seasonality of Air−Surface Exchange. Simulations show the occurrence of intense seasonality in volatilization fluxes which are very similar for both pesticides (Figures 2 and S2). During winter months volatilization fluxes are generally very low or even absent since air and soil appeared to approach thermodynamic equilibrium. Between May and October intensive temperature driven enhancements of volatilization fluxes appear in all types of land use. Volatilization fluxes from arable soils show maxima during June−July but also in September. The September peak is an effect of plowing which appears to be a land management practice potentially having important effects on POPs revolatilization. Influence of plowing in the model takes place only during September and to a much lower extent October, however, during this period about 6.5% of the annual overall emission of HCB and 7% of DDE in the Czech Republic is emitted. This value is much larger than the sum of contributions from grassland and forest during the whole year. Although the parameterization adopted here to describe plowing in the model is very simplistic, these results show the convenience of further developing the parameterization and inclusion of this process into chemical fate models. Although our data refer to the Czech Republic scenario (which has a higher fraction of arable soils (45%) compared to the average of European countries (24%),46 USA (20%), China (15%), Brazil (7%), or the world as a whole (11%), the
Volatilization of both HCB and DDE is predominantly associated with arable lands (representing ca 98% of total revolatilization), contributions of grassland and forest are 1.1% and 0.5% for HCB and 0.5% and 1.3% for DDE, respectively. A summary of the results is reported in Table 1. Notably these values are not correlated to the land use distribution since arable soils in the Czech Republic cover a surface slightly larger than forests and only twice as large as grasslands. Enhanced emissions from farmlands are instead to be associated primarily with the higher fugacity in soils (which historically received direct inputs of these contaminants); with the effect of plowing and finally with the higher soil temperatures in summer calculated for arable soils compared to other land uses. It must be acknowledged that fugacity in soils was calculated here considering soil concentrations averaged along a soil depth of 20 cm. Distribution of POPs may not be homogeneous with depth of soil due to the different properties of different soil horizons. However this is true only for forest in which soil horizons are clearly distinct. Typically the bulk of POPs in forest soils is present at the Oa horizon. The fugacity of soil layers overlying this horizon and engaging in gaseous exchange with air may be different from that in the Oa horizon, and this can drive to a relative underestimation of volatilization. However since re-emission from forest represents only a minor fraction of total volatilization fluxes, this bias only has a minor effect on the outcome of this study. For agricultural soils, in fact, management actions such as plowing disrupt the buildup of soil horizons and homogenize pesticide distribution in the soil upper core, preventing the problem. Similarly 7056
dx.doi.org/10.1021/es3048784 | Environ. Sci. Technol. 2013, 47, 7052−7059
Environmental Science & Technology
Article
It must be noted, however, that the parameterization for bare soils was only applied when arable lands were standing cropfree (i.e., during the period between plowing and the new growing season). Whenever standing crop was present, arable soils temperature was parameterized consistently with grasslands in which predicted soil temperatures were very similar to air temperatures in all of the environmentally relevant range (e.g., −2 to 32 °C). Temperature changes had a little less influence on forest soils in comparison with other land use types. This is probably due to the higher content of organic carbon and lower soil temperature in summer months when compared to grassland and arable land. Other experimental studies also confirmed the significant influence of temperature on volatilization.47 However, it is difficult to make a sound comparison between results from field and laboratory volatilization experiments with model predictions, since in the first case artificial soil and highly contaminated or artificially spiked soils were used.11,47−49 In addition, methods used for experimental determination of air− surface are not yet well established and some approaches may even result in artifacts. Influence of Land Use Change on POP Emission. The AG scenario had strong influence on volatilization fluxes resulting in a decline of secondary emission of −7.5% in all classes of altitude. This scenario does not include appropriate development of Corg to the values typical of grasslands. If we take into account also soil organic carbon development (AG_C scenario) re-emission fluxes will further decrease to a −8.5% for both POPs (Figure 3). Standardized results of AG scenario are nearly identical to AG without correction on distribution of land use.
predominant influence of croplands in controlling re-emission of persistent organochlorine pesticides (which have been widely used on the global scale), suggests that this land use type is key for the inventory of secondary sources to the atmosphere also in other regions. Volatilization dominantly exceeded dry gas deposition on all land use typologies basically during most of the year and for both the studied pesticides. Forests and grasslands behaved more similarly, however some differences in seasonal trends can be observed. In forests, dry gaseous deposition of HCB exceeded or balanced volatilization in 10% of the total area (with higher percentage in colder months and/or higher elevations) (Figure S3). For grasslands, instead a much smaller percentage of the total area behaved as sink for atmospheric HCB (typically
