Determination of vulnerability areas from the simulated deposition of atmospheric pollutants using LOTOS-EUROS chemical transport model in North West South America

This work presents the implementation of the LOTOS-EUROS regional atmospheric Chemical Transport Model (CTM) on Northwestern South America. The impact of land use and orography update in the model was analyzed to identify potential vulnerable natural areas by quantifying atmospheric deposition pollutants. CTMs allow simulating the physical dynamics of trace gasses and aerosols, including processes such as emission, chemical reactions, transport, and deposition. The deposition 5 of atmospheric contaminants like nitrogen dioxide (NO2) and ammonia (NH3) induces chemical fluxes in natural ecosystems, with potential subsequent severe impacts like biodiversity loss. Due to the vast geographical diversity present in the study area, the LOTOS-EUROS model was updated for the land and topography inputs to simulate more representative conditions for the study region. Depositions were very sensitive for the change of land cover maps used in the model, and on the other side, topography update impacts more in the high layer of the model above harsh terrain. Additional simulations for the updated 10 scenario using point sources were performed to identify the deposition area’s spatial extent for the principal Colombian cities.

This paper is a non-peer reviewed preprint submitted to EarthArXiv.  The model follows a mixed-layer level approach with a 25 m surface layer, a mixed layer with a top at the boundary layer height taken from the meteorological input, and two reservoir layers of at least 500 m. When the mixing layer is very thick, 95 such as over mountainous terrain in the tropics where elevations can regularly reach heights of more than 3500 m, the top of the model is extended to accommodate the minimum thickness of the reservoir layers. The deposition process is dominated by the interaction of the static surface layer. The default elevation model for LOTOS-EUROS is obtained from the ECMWF meteorological data, which has a resolution of 0.07 • (≈ 7 km). An updated elevation model used for the region was obtained from the Global Multi resolution Terrain Therefore emissions were taken from EDGAR (Emission Database for Global Atmospheric Research) 4.2 for 2008. However, This paper is a non-peer reviewed preprint submitted to EarthArXiv. previous studies have shown a significant gap in knowledge for the Colombian territory in the EDGAR inventory (e.g., (Gonzalez et al., 2017;Pachón et al., 2018;Nedbor-Gross et al., 2018)), this database was at the time the only one accessible with all the species required to operate the model in the selected domain for the time of the simulations. Biogenic emissions were taken from the MEGAN 2.1 model. The MACC/CAMS GFAS global fire assimilation system from Kaiser et al. (2012) was used with a time resolution of 1 hour to account for the occasional fire events. The chemical mechanism used was the Carbon 110 Bond mechanism 5 (CB05), and the sea salt emissions were parameterized according to formulations for the fine and coarse aerosol modes from (Monahan et al., 1986) and (Mårtensson et al., 2003). This paper is a non-peer reviewed preprint submitted to EarthArXiv. 2.1.1 Land Cover/Land Use data For modeling the deposition dynamics, LOTOS-EUROS requires a map with deposition properties per grid cell. Land use characteristics are relevant for the CTM deposition dynamics because they define the parameters of the terrain roughness and 115 canopy altitude of each category that determine the velocity at which the component will be deposited, dependent on the vegetation type. The default land use/land cover (LU/LC) input data for LOTOS-EUROS were derived from the Global Land Cover (GLC2000) project (Fritz et al., 2003). GLC includes 23 categories consistent with the FAO (Land Cover Classification System of the Food and Agriculture Organization) classification (Di Gregorio, 2005). For South America, the mapping of these categories at spatial resolutions of 1 km x 1 km was done in (Eva et al., 2002), with contributions from some regional experts 120 based on multi-resolution satellite data. In this work, the LU/LC data was updated with the 2009 Land Cover Climate Change Initiative (CCI) dataset (Defourny et al., 2017). CCI has 38 categories with a horizontal resolution of 300 m x 300 m. Figure   3 compares the default and updated LU/LC models for Aburrá Valley. The mapping of the 39 (CCI) and 23 (GLC) LU/LC categories to the nine classes of the DEPAC deposition model is illustrated in Figure 4. The descriptions of each category are presented in Table 1. The mapping from CCI to GLC took into account the similar morphological characteristics between 125 categories and the aseasonality in this tropical region. The mapping from GLC to DEPAC is the standard scheme constructed for LOTOS EUROS. The model defines each grid cell the fraction covered by each of the LU/LC classes used by the DEPAC module and calculates each fraction's deposition.

Experiment description
Simulations were conducted with the following inputs: Experiment 1 (Exp-1), default elevation model and default LU/LC data; 130 Experiment 2 (Exp-2), default elevation model and updated LU/LC data; Experiment 3 (Exp-3), updated elevation model and default LU/LC data; and Experiment 4 (Exp-4), updated elevation model and updated LU/LC data. For each of these four experiments, the total deposition (wet and dry) nitrogen dioxide (NO 2 ) and ammonia (NH 3 ) was calculated for the entire 2016.

Fate of urban contaminants experiments
To explore the fate of nitrogenous atmospheric species emitted from the main Colombian cities, the grid cells housing the centroids of the urban area for Bogotá, Medellín, Cali, and Barranquilla were assumed as artificial point sources of emissions.
The simulations were conducted with the updated elevation model and updated LU/LC scheme detailed above, for a total of 140 10 days in four different times of the year: March 1-10, June 1-10, September 1-10, and December 1-10. After a 2-day model spin up, the point source was from 08:00-18:00 of day 3 of the simulation, emitting a total of 1000 kg/hour NO 2 , which is the amount of daily NO 2 emissions reported for Medellń (UPB and AMVA, 2017). The artificial emissions were monitored during This paper is a non-peer reviewed preprint submitted to EarthArXiv. seven additional days, during which time all of the emitted species had either deposited or transformed. Similar simulations were conducted but without the point source's activation to estimate the background deposition values for each grid cell.

145
A second experiment was conducted as above, but focusing on either Medellín or Rionegro. The latter city is located at the This paper is a non-peer reviewed preprint submitted to EarthArXiv. 3 Results

Updated elevation model
LOTOS-EUROS interpolates the input elevation data within each grid cell according to the simulation's resolution (Fig. A1).

155
Changing the input elevation model can generate changes in the outcome of variables such as the vertical layers' temperature profiles. The effect of an updated elevation model depends on the desired simulation grid resolution. Figure 5 shows a transverse cut at a latitude of 6.6 • North for the simulation at a horizontal resolution of 0.09 • x 0.09 • , illustrating the impact of the change This paper is a non-peer reviewed preprint submitted to EarthArXiv.  Table 1. of input elevation information through the Aburrá Valley. The most significant temperature changes occurred in the upper layers above the rugged terrain, reaching differences of up to 5 • C degrees in top layers.  Changes were noticeable for categories like grasslands where the deposition decreased with the updated land-use configuration.
For categories such as deciduous forest, arable, coniferous forest, and permanent crops, mostly located in the Eastern region of the country, the simulations showed an increase in depositions related to the change in the fraction percentage of these 170 categories. The Amazon region (SE) presented minimal changes in deposition between the two LU/LC scenarios, primarily due to the negligible changes in LU/LC between the two data sources. The highest changes were found along the Andean cordilleras related to deposition in deciduous and coniferous forest, which were higher in the updated LU/LC scheme.
This paper is a non-peer reviewed preprint submitted to EarthArXiv.  This paper is a non-peer reviewed preprint submitted to EarthArXiv. use and changed to mosaic cropland, shrub, and grass cover in the new one. In the MFB image, overestimating respect to the Exp-1 conditions is seen in how the dominant land use was changed. Figure 9 shows the comparison between the deposition velocity and flux for the nine simulation categories. The comparison was made to compare a pixel of a particular zone of interest in a paramo ecosystem named "Paramo de Belmira" located in the northwest of Medellín city. It is possible to see that the change of seasons is significant for the velocity and flux deposition

Point sources experiment
The LOTOS-EUROS emission module explains the discharge of tracers and aerosols from various sources (anthropogenic, biogenic, marine, airborne dust, fires) that can be configured to define emissions in specific point sources to simulate scenarios. Figure A2 shows the simulation of the total deposition (dry and wet) for Nitrogen taking into account the emission of the 200 four principal cities in Colombia. Cities here were assumed to be point emission sources, which works to determine this city's influence area. Figure (10,A) shows the contours generated with an increment of 5 g/ha between contour level and bias correction of +2 g/ha to avoid negative values that appears as a numerical noise from the rest of the reference run minus the punctual perturbed emission run. The more rounded contour delimits the impact zone for each of the cities.
It is possible to identify some wind direction trends for the four different times of the year 2016 (1-9 days of March, June,

205
September, and December). The influence of Barranquilla to faraway zones is perceivable due to the close location of this city to the Caribbean coast, where intense wind conditions exist and flat topography that drives the transport dynamics far away.
More of the depositions from Barranquilla is going to the ocean direction and, in other time to the southwest of the city reaching inclusive the other cities deposition areas. For the other cities, the impact area is more limited but with higher deposition values due to the mountainous terrain's roughness and the less magnitude of the wind patterns presented. It is also interesting to see  This paper is a non-peer reviewed preprint submitted to EarthArXiv.
Code availability. One of the results of this work is the new land use generated for this study domain. the netcdf file is added to the submission of this work Data availability. TEXT

270
Appendix A Figure A1 shows the comparison between the two orography scenarios input to the model (the images at the left) against the four different run simulation resolution (0.12 • ,0.09,0.06,0.12). Although the CTM is terrain following, the simulations are insensitive to the orography change map when the simulation needs to be computed in high resolution. For no necessary high resolutions for the simulation, the change in orography is not relevant because the interpolation softens the differences.