Friagem Event in Central Amazon and its Influence on Micrometeorological Variables and Atmospheric Chemistry

In the period between July 9 and 11, 2014 a Friagem event reached the central Amazon region causing significant changes in microclimate and atmospheric chemistry. On July 11, the southwest flow related to the Friagem converged with the easterly winds in the central Amazon region. The interaction between these two distinct air masses formed a convection band, which intensified over the Manaus region and the Amazon Tall Tower Observatory (ATTO) site. The satellite images show the evolution of convective activity on July 11, which lead to 21 mm of precipitation in the ATTO site. Moreover, 5 the arrival of the Friagem caused a sudden drop in temperature and a predominance of southerly winds, which could be seen in Porto Velho between July 7 and 8 and in Manaus and ATTO site from July 9 to 11. The results of ERA-Interim reanalysis and Brazilian developments on the Regional Atmospheric Modeling System (BRAMS) simulations show that this Friagem event coming from the southwest, carries a mass of air with higher O3 and NO2 mixing ratios and lower CO mixing ratio compared to the airmasses present at the central Amazon. At lake Balbina the Friagem intensifies the local circulations, 10 such as the breeze phenomena. At the Manaus region and ATTO site, the main effects of the Friagem event are: a decrease in the incoming solar radiation (due to intense cloud formation), a large temperature drop and a distinct change in surface O3 and CO2 mixing ratios. As the cold air of the Friagem was just in the lower 500 m the most probable cause of this change is that a cold pool above the forest prevented vertical mixing causing accumulation of CO2 from respiration and very low O3 mixing ratio due to photochemistry reduction and limited mixing within the boundary layer. 15 1 https://doi.org/10.5194/acp-2020-564 Preprint. Discussion started: 14 July 2020 c © Author(s) 2020. CC BY 4.0 License.

The data acquisition at the tower was performed by data loggers CR1000 and CR3000 (Campbell Scientific inc., USA), with instantaneous measurements taken every minute for meteorological variables and at high frequency for CO 2 (10 Hz) and O 3 (30 s) mixing ratio, subsequently processed every 30 min. The variables used in this study and their respective sensors are presented in more detail in Table 1. 85 The O 3 data at T3 site were obtained as part of the U.S. Department of Energy Atmospheric Radiation Measurement Program (ARM, http://www.arm.gov/measurements) during the GoAmazon 2014/5 project (Martin et al., 2016). O 3 -mixing ratios were measured with an ultra violet gas analyzer (TEI 49i model, Thermo Electron Corp, USA). The instrument was installed at a height of 3.5 m above the ground (Dias-Júnior et al., 2017). At T2 and T0z, O 3 -mixing ratios were also measured with the same analyzer model (Thermo 49i) at a height of 12 m a.g.l. and 39 m a.g.l., respectively.

90
The European Center for Medium-Range Weather Forecasts (ECMWF) ERA-Interim reanalysis was used at intervals of 6 h, with the objective of evaluating the evolution of the Friagem event investigated in this work. The ERA-interim model and the ECMWF reanalysis system present spatial resolution with 60 vertical levels, harmonic spherical representation for the basic dynamic fields, and reduced Gaussian grid with uniform spacing of approximately 79 km for the surface (Berrisford et al., 2011). Furthermore, enhanced images of the infrared channel of the GOES-13 satellite were used, with the purpose of 95 analyzing the formation and passage of convective systems in the study area.

Experimental design
The numerical simulations of the present study were made using the BRAMS (Brazilian Regional Atmospheric Modeling System) mesoscale model version 5.3 (Freitas et al., 2017). BRAMS represents a Brazilian version of the Regional Atmospheric Modeling System (RAMS) (Cotton et al., 2003) adapted to tropical conditions. This version of BRAMS contains the coupling 100 of the JULES (Joint UK Land Environment Simulator) (Best et al., 2011;Clark et al., 2011) and CCATT (Coupled Chemistry-Aerosol-Tracer Transport) models (Longo et al., 2010;Freitas et al., 2009), making BRAMS a new and fully-coupled numerical system of atmosphere-biosphere-chemical modeling, called JULES-CCATT-BRAMS (Moreira et al., 2013).
The integration time of the model was 72 hours, starting at 00 UTC on July 9 th , 2014. The numerical experiment was performed using only a grid whose horizontal resolution was 1.5 km, with 185 points on x, 140 points on y, and 39 points 105 on z. The vertical grid resolution was variable with the initial vertical spacing of 50 m, increasing by a factor of 1.1 up to the 1.2 km level, and from that point forward this spacing was constant to the top of the model (around 16 km). The domain covered by this grid, the distribution of the main rivers and topography can be observed in Fig. 2.
The initialization of the model was heterogeneous, using the ECMWF-ERA Interim reanalyses (www.ecmwf.int/en/forecasts/ datasets/reanalysis) every 6 hours in a quarter-degree spatial resolution. Seven soil layers were defined up to the depth of 110 12.25 m and the assumed soil humidity was heterogeneous, as described in Freitas and Freitas (2006). Soil texture data were originally obtained from the Food and Agriculture Organization of the United Nations (UN FAO) and were adapted for the Brazilian territory by INPE (Rossato et al., 2004).
In this simulation, cloud microphysics uses the Thompson cloud water single-moment formulation, which consists of the separate treatment of five classes of water that are then mixed in a single treatment for each type of cloud (Thompson et al.,  2008; Thompson and Eidhammer, 2014). In addition, it includes the activation of aerosols in the cloud condensation nuclei (CCN) and ice nuclei (IN), thus, it predicts the concentration of the number of water droplets in the clouds, as well as the concentrations of two new aerosol variables, one for CCN and one for IN. These variables are grouped into hygroscopic aerosols called "water friendly" and non-hygroscopic aerosols are "ice friendly" (Freitas et al., 2017).
The parameterization of the long and short wave radiation used was the Carma (Community Aerosol and Radiation Model 120 for Atmospheres) (Toon et al., 1989). This scheme solves the radiative transfer using the two-flux method and includes the main molecular absorbers (water vapor, CO 2 , O 3 and O 2 ) and treats the gas absorption coefficients using an exponential sum formula (Toon et al., 1989). The JULES-CCATT-BRAMS radiation schemes are coupled online with the cloud and aerosol microphysics models to provide simulations of aerosol-cloud-radiation interactions (Freitas et al., 2017). The physical and optical properties of the cloud in the radiative scheme of Carma were parameterized according to Sun and Shine (1994) and 125 Savijärvi et al. (1997); Savijärvi and Räisänen (1998) using liquid and ice water content profiles provided by the JULES-CCATT-BRAMS cloud microphysics scheme (Freitas et al., 2017).

Environmental characteristics in the Amazon basin scale
From the ECMWF ERA-interim reanalysis the evolution of the horizontal wind and air temperature near the surface, in the 130 north region of Brazil, between July 6 th and 11 th , 2014, at 12 UTC (Local Time = UTC -4 h) (Fig.3) can be obtained. On the 6 th it is observed that the mean temperature was of the order of 24 • C in three places of interest of this work, being: Porto Velho; Manaus and ATTO site (Fig. 3a). The dominant wind direction was from East in practically the entire Amazon region.
The surface temperature and wind direction represent the standard normally found in this region (Fisch et al., 1998;Pöhlker et al., 2019). However, on July 7 th , the dominant wind direction becomes South-Southeast in the region of Porto Velho, as is 135 evidenced by the presence of a mass of air with a lower temperature (around 18 • C) approaching this city (Fig. 3b).
In the course of the days, between July 8 th and 9 th , the mass of cold air advanced even more towards Porto Velho, just as the dominant wind direction changed to South in all western regions of the state of Amazonas, as well as to the southern regions of Manaus and the ATTO site (Fig. 3c, d). On July 10 th , the southerly winds arrive in the Manaus region and the ATTO site, characterizing the arrival of Friagem in the area of interest of this work (Fig. 3e, f). For this period, the CPTEC 140 technical bulletin reported the penetration of a polar air mass in the subtropical and tropical Brazilian region that advanced in the Southeast-Northwest of Brazil, giving origin to the cold waves of the South, as well as causing the Friagem phenomenon in the Amazon (http://tempo.cptec.inpe.br/boletimtecnico/pt). Therefore, the arrival of the Friagem phenomenon in the Southwest and central regions of the Amazon is evidenced, that produced abrupt drops in the values of temperature and change in the wind direction. Similar results were also found by other 145 authors (Marengo et al., 1997;Fisch et al., 1998;de Oliveira et al., 2004).
The wind behavior throughout the Amazon basin before and during the Friagem event is represented in Fig. 4a and 4b, respectively. Interestingly, at the time the Friagem was present in the Manaus and ATTO site region, there was convergence of the easterly winds with the westerly flow associated to the Friagem (Fig. 4b). The easterly flow carries humidity from the Atlantic coast to the central region of the Amazon, while the southerly flow, associated with the Friagem event, transports 150 masses of dry and cold air from high latitudes to the Amazon region (Marengo et al., 1997). Figure 5 shows the satellite images before and during the Friagem event in the central Amazon. Convection in the confluence between Amazonas and Tapajós rivers region was observed at dawn, on July 11 at 07 UTC (Fig 5a). This convection propagated in the West direction, arriving in the ATTO site region at 13 UTC (Fig 5c). Since this convective system is not associated to the squall lines that form along the coast (Cohen et al., 1995;Alcântara et al., 2011;Melo et al., 2019) it is possible to state that 155 this convection has its formation associated with the convergence of these two air masses with different properties (Fig 4). It is noteworthy that during the propagation of this convection on July 11 th , it intensified and caused the highest rainfall (starting at 12:30 UTC) registered at the ATTO site during the month of July 2014, with a record rainfall of 21 mm.
The evolution of the horizontal wind and O 3 -mixing ratio near the surface (both from ECMWF ERA-interim reanalysis), during July 7 th and 11 th , 2014, at 18 UTC can be seen in Fig. 6. On July 7 th onward the Friagem event carries air rich in O 3 160 northwards (Fig. 6a). This airmass reaches the state of the Amazonas on July 8 th (not show here). On July 11 th at 12 UTC (not shown here) the air mass influenced by the Frigaem has the shortest distance from the study region (ATTO-site). On July 11 th at 18 UTC the Friagem begins to dissipate (Fig. 6b). However, it should be noted that this mass of air rich in O 3 did not reach the Manaus region and the ATTO-site. It is believed that the presence of the cloud cover in central Amazonia on 11 th , July production. The rain forest canopy is a strong sink for ozone Fan et al., 1990;Rummel et al., 2007). Therefore, the low O 3 mixing ratio in the Manaus region and the ATTO-site during the 11 th July (Fig. 6-f) would be associated with cloudiness and prolonged transport over forested regions. Marengo et al. (1997) also showed that one of the effects of Friagem in central Amazonia is the induction of the cloudiness.

170
They did not show the impact of Friagem on O 3 levels but showed that the values of income short radiation that reach the surface are greatly reduced during Friagem events. In this work we also find a strong reduction in income short radiation in central Amazonia that will be discussed in Fig. 10 3.2 Air temperature during the Friagem event  15 • C) between July 7 th and 10 th , whilst in Manaus region and at ATTO the decrease was in the order of 10 • C and 7 • C, respectively, between the 9 th and 11 th of July. Therefore, the temperature fall in Manaus region and ATTO occurred around 180 one day after the temperature fall observed in Porto Velho.
At Porto Velho, both the maximum and minimum values of air temperature were substantially reduced during the presence of the Friagem. However, at Manaus region and the ATTO site, the decrease was mainly observed in the maximum temperature values. Although the decrease was not so evident at the time of the diurnal minimum (at least on the 10 th and 11 th ) the whole diurnal cycle was disturbed with (much lower) minima than the average at different times of the day.

185
Similar behavior was observed by Marengo et al. (1997) for the Southwest and Central Amazon regions during an episode of Friagem. Therefore, it is noted that due to the occurrence of the Friagem, the southernmost regions of the Amazon present more intense reductions in temperature values, compared to the regions located more in the center of the Amazon basin.
Additionally, the ATTO site is located in a forest region, 58 km from the Balbina dam lake and Manaus region is under the influence of intense urbanization (de Souza and Alvalá, 2014) and is located in the proximity of rivers. Thus, there is evidence 190 that both the ATTO site and Manaus region may be under the influence of lake (Moura et al., 2004) and rivers breezes (dos Santos et al., 2014), respectively, which could offer them a greater thermal inertia.

ATTO site wind direction
In addition to the changes observed in the daily air temperature cycle at the ATTO site, changes were also observed in the local wind direction during the Friagem period (Fig 8). Before the arrival of this phenomenon, between July 6 th and 8 th , it was 195 observed that the direction of the horizontal wind was predominantly Southeast and Northeast. On the other hand, on July 9 th the wind direction was well distributed among the four cardinal points, and on July 10 th and 11 th the wind flow had higher frequencies of West, North and Southwest, when the Friagem arrived at ATTO site. The general wind directions before and after the Friagem are consistent with long term observations at ATTO (Andreae et al., 2015). The low frequency of observed wind directions from the westerly directions (based on 2.5 years of data) led to the conclusion that effects of local circulation 200 (due to Uatumã River ≈ 12 km and Balbina Lake ≈ 58 km) are not important or could not be observed (Andreae et al., 2015).
At least not on regular basis.
Silva Dias et al. (2004) showed that during the arrival of a Friagem event in the Santarém-PA region (East of the Amazon), the atmospheric pressure at sea level increased, resulting in a pressure gradient force pointing in the opposite direction than the trade winds, which would be consistent with a slowdown of the easterly winds. In this way, these authors were able to observe 205 with greater clarity the occurrence of river breeze circulations in this region. Following this hypothesis, the behavior of the wind at the ATTO site was analyzed every two hours, during the period in which the Friagem was active in this region Fig. 9.  UTC it is observed that the prevailing wind was from the West, indicating a deviation from the general flow, which would normally be from the East. In the early afternoon (16 UTC), the wind changed to the North direction until the early morning (12 UTC) of July 11 th . This change in wind direction to the West and to the North observed during the morning of July 10 th and 11 th , respectively, does not correspond to the expected direction during the occurrence of the forest breeze towards Lake Balbina. Therefore, it is believed that the flow related to the Friagem phenomenon overlapped with that of the breeze 215 circulation observed by Moura et al. (2004), or that the forest-lake breeze circulation does not present the capacity to reach the micrometeorological tower of the ATTO site in 58 km distance (In line with results from Andreae et al. (2015)). This aspect will be discussed in the next section where the results of the simulation with JULES-CCATT-BRAMS model will be analyzed.
3.4 Radiation, ozone and CO 2 during the Friagem event Figure 10 shows the values of incident short wave radiation (SW in ), O 3 and CO 2 measured at the ATTO site, between July 220 6 th and 11 th , 2014, respectively (black line). The SW in values decrease during the morning of July 11 th when Friagem arrives at the ATTO site (Fig 10a). Moreover, the maximum value (≈ 450 W m −2 ) of SW in occurred at approximately 19 UTC (15 LT), whereas the average monthly daily maximum SW in (orange line) usually occurs at 16 UTC (≈ 800 W m −2 ).  The reduction of the incident short-wave radiation values observed on the 11 th (Fig 10a) was possibly associated to the 235 presence of convective systems in this region, as shown in Fig 5. It is known that cloudiness is a determinant meteorological factor in the daily O 3 cycle (Gerken et al., 2016).
It is interesting to note that the rain event during July 11 th did not result in an increase of near surface O 3 as observed by others authors (Betts et al., 2002;Gerken et al., 2016;Dias-Júnior et al., 2017). It is believed that the convective cloud formed during the Friagem event was not as deep as the clouds investigated by Betts et al. (2002) and Gerken et al. (2016), which, 240 through their downdrafts, transport O 3 from the high troposphere to the surface. molar fraction around 420 ppm approximately at 10 UTC and minimum of less than 390 ppm (de Araújo et al., 2010). However, on July 11 th at 14 UTC, a significant increase of CO 2 (≈ 470 ppm) was observed in relation to the monthly average. This 245 increase may be related to the incident radiation attenuation due to increased cloudiness which reduces the efficiency of the forest in absorbing CO 2 gas via photosynthesis (Ruimy et al., 1995). Also limited vertical mixing as discussed below is a potential reason.

Simulation of local circulation and its effect at the ATTO site
In order to better understand the local circulation and its role on the measurements made at ATTO site region, this section 250 presents the results of a numerical simulation made with JULES-CCATT-BRAMS coupled model. Figure 12a shows the vertical profile of the horizontal wind at a grid point near the ATTO site (02 • S -59 • W ) during model integration. At low levels (near 80 m), the Easterly wind is observed until the first hours of July 10 th . Then the wind has a predominant West-Northwest direction until the afternoon of July 11 th and afterwards the wind comes from the South. Therefore, it is observed that the simulation captured the horizontal wind behavior measured at a height of 73 m at the ATTO site, as shown in Fig 9. In addition, 255 above 500 m the flow is essentially from the East during the whole period of integration of the model. Apparently, the Friagem changes only the flow within a small layer adjunct to the ground. Figure 12b shows the values of the boundary layer height (BLH) obtained form ERA5 at a grid point near the ATTO site (02.10 • S -59.06 • W ). It is possible to note that before the Friagem event the maximum BLH values were greater than 1000 m. However, during the Friagem event, the maximum BLH value was around 600 m.

260
The large temperature drop Fig. 7a together with the information that the cold air of the Friagem was just in the lower 500 m (Fig. 12), points to the formations of a cold pool above the forest that prevents vertical mixing. As incoming solar radiation was low (Fig. 10a) the surface heating might not be sufficient to break the inversion or at least a very shallow boundary layer was formed as evidenced by the ERA5 data (Fig.12b). This would explain high CO 2 (accumulation of soil emissions) and very low O 3 (limited transport from aloft) at the same time at the ATTO site in addition to the reduced radiation (see section 3.3).  July 11 th . Between 03 and 11 UTC, the air temperature is higher on Balbina Lake compared to that above the forest area. This temperature gradient induces the formation of a forest breeze towards the lake with the wind converging towards the center of the lake (Fig 13a-e). At 13 UTC the temperature gradient reverses its direction and induces the formation of the lake breeze towards the forest that at 15 UTC is more clearly defined along the southeastern shores of Balbina Lake (Fig 13g).

270
Another interesting aspect is the entry of cooler air through the Northwest quadrant starting at 03 UTC, which is transported in Southeast direction. From 3 to 11 UTC a corridor of warmer air is established from Lake Balbina to the Southeast quadrant of the domain along the Uatumã River whose width is less than 1 km and can not be captured by the horizontal resolution in this simulation. The gradual drop in temperature and predominance of Northwest winds shown in this simulation at the grid points near the ATTO site agree with the observational data from this site ( Fig. 7 and 8).

275
Although the Balbina lake breeze was established, it did not reach the ATTO site until 15 UTC (Fig 13g). In addition, precipitation in the simulation occurred in the following hours, similar to that observed in satellite images (Fig 5), which in turn disrupts the environment propitious to vigorous breezes that could reach the ATTO site. Although the Friagem phenomenon causes the weakening of the trade winds, which in turn would allow the establishment of more intense breezes as proposed by Silva Dias et al. (2004), the cooler and drier air mass flow of Friagem in the central region of the Amazon was dominant  O 3 -mixing ratios are higher above the lake and its surroundings, for both times shown ( Fig. 14c and d). The O 3 -mixing ratio within the limits of the simulation domain are mostly below 11 ppbv, whereas above the lake these mixing ratios exceed 20 ppbv at certain points, especially at 02 UTC. The effect responsible for higher O 3 -mixing ratio both during the day e night 290 may be associated with the fact that deposition is very much reduced over the open water compared to the forest (Ganzeveld et al., 2009). It can also be seen that the Friagem extended in the direction of ATTO, but probably due to the onset of rain ( Fig.   5) was not clearly detected at ATTO.
In regards of CO gas, it can be observed that its concentration on the center of the lake at 02 UTC (Fig. 14e) is higher than in the regions near the margins of the lake, however, calls attention at this time the transport of CO arriving with the South and 295 Northeast winds, approaching the ATTO site. However, it is noted that the entire region of the simulation domain presents low CO mixing ratio at the time the Friagem is dissipated (Fig. 14f). Apparently, the Friagem event "expels" the polluted air mass in the South and Southeast of the ATTO site (around Manaus city), "cleaning" the atmosphere, or preventing this pollution from reaching ATTO site and Balbina lake.
NO 2 gas is an important precursor of O 3 , and is mainly related to emissions from fires and vehicles. The emission of 300 precursor gases in the formation of O 3 mixing ratio can increase of this trace gas to levels harmful to the forest, since the ozone can damage the stomatal functions of the leaves (Pacifico et al., 2015). In spite of this, it is observed that the higher NO 2 mixing ratio at 02 UTC ( Fig. 14g) seem to have their origin in the region where higher O 3 mixing ratios are found and presented lower NO 2 during the time of dissipation of the Friagem (Fig. 14h).

305
In the period of July 9 th to 11 th , 2014 a Friagem phenomenon reached the central region of the Amazon. Through the ECMWF ERA-interim reanalysis it was possible to verify that this phenomenon ventured the Amazon region from Southwest to Northeast, bringing a strong cold, dry, ozone-rich air mass in the West quadrant, which dominated the wind field in the central region of the Amazon.
Through the observational data it was possible to verify that the passage of the Friagem in central Amazon had its most 310 significant effects on July 11 th , in region of the Manaus city, such as: Balbina Lake; ATTO site and others sites (T2, T3 and T0z).
From the observational data collected at the ATTO site, it was observed that the 11 th was marked by a sudden fall in air temperature, a weakening of the typical East flow and a predominance of South, West and North winds. In addition, on the 11 th the interaction between the Friagem air mass and the trade winds flow gave origin to convection bands, which in turn caused 315 a significant reduction of the incident short wave radiation, besides a record rain of the month. With the BRAMS simulations we found that the cold air of the Friagem was just in the lower 500 m. These information leads us to the conclusion that there is a cold pool above the forest that prevents vertical mixing and consequently a increase in CO 2 mixing ratio and abrupt drop in O 3 mixing ratio is observed above the forest canopy.
Also, through the simulations of the JULES-CCATT-BRAMS it was possible to evaluate the main impacts that the Friagem 320 phenomenon caused both in the thermodynamic characteristics and in the atmospheric chemistry of the central region of the Amazon. In general, the model reproduced satisfactorily the main changes that the phenomenon brought to the environment of interest. In addition, the breeze circulations between Lake Balbina and the forest were well represented in the simulations, however, it was not possible to verify the influence of this breeze in trace gas concentrations at the ATTO site.
With the observational results and the simulations, it can be concluded that the Friagem phenomenon can interfere deeply in