The characterization of spatial and temporal patterns of wind is essential to several sectors, including energy, urban climate, and applied meteorology (Correia Filho et al., 2022). In addition, wind data are used in the calculation of hydrological models, with the objective of assessing the impacts of changes in land cover on the water balance of the Itacaiúnas River Basin, providing insights into the management of water resources (Pontes et al., 2019).

In the context of environmental risk, studying wind transport allows us to understand the potential for mass dispersion, such as gaseous pollutants, in a given region. In this way, Ye et al. (2019) found that the modulation between large-scale atmospheric circulation and regional transport explained the wet deposition of mercury in an urban environment from anthropogenic emissions outside the state of New York. In another study, Quaghebeur et al. (2019) observed high concentrations of arsenic from the interaction between rain and the roof in a rainwater harvesting system on the outskirts of Poopó Lake in Oruro, Bolivia, and reported that the source of arsenic in the dust may be natural, related to mining, or both.

In Brazil, the Itacaiúnas River Watershed (IRW), which is located in the Eastern Amazon (Figure 1), brings together urban areas, farms, forests, and mining activity. Hypothetically, the elaboration of strategic plans for production and security of productive sectors such as agrarian, mining and public management should include prior knowledge about regional wind transport since the propagation of all types of matter, such as plumes of smoke, gases, and dust, depends mainly on the trajectory of the prevailing winds and their intensity, which can vary considerably between seasons. In this context, Tavares and Silva Júnior (2019) discussed the role of regional circulation from a mining area in the municipality of Canaã dos Carajás in the state of Pará, Brazil, to understand the wind field in order to make immediate decisions regarding the supposed aerial displacement of pollutants to urbanized areas. The authors found that during the period studied, the predominant wind directions were Northeast (NE) and Southeast (SE); therefore, no pollutants were transported to urban regions.

Atmospheric circulation in the IRW region, presents variability caused by the composition of multiple-scale (space and time) phenomena, including local storms, squall lines, and the intertropical convergence zone (ITCZ) (Cohen et al., 1995; De Souza et al., 2017). In these cases, advanced studies on the theme of atmospheric variability (hereafter frequencies or periodicities) have successfully used the wavelet transform (WT) technique, which identifies the dominant periodicities and their interactions with other frequencies in a time series. Studies such as these are important in supporting environmental diagnoses, in addition to providing knowledge of small-scale mechanisms that influence local circulation and the regional climate (Bitencourt et al., 2016). Thus, the WT has been used in several studies on atmospheric phenomena (Echer et al., 2008; Blain and Kayano 2011; Moura and Vitorino, 2012; Vilani and Sanches, 2013; Rocha et al., 2018; Rocha et al., 2019), hydrological systems (Sassi et al., 2011; Santos et al., 2013; Guo et al., 2015; Sun et al., 2016), and tidal studies (Danial et al., 2019).

Knowledge of the variability of local systems, such as breezes and topographical forcing, is important when acquired during periods when the effects of the El Niño-Southern Oscillation (ENSO) are not present. Studies indicate that this large-scale phenomenon interferes with the local wind field (Berg et al., 2013; Mohammadi and Goudarzi, 2018). Watts et al. (2017) assessed the impact of ENSO on the wind speed used for energy production. They concluded that in the Canela area of Chile, El Niño significantly decreases the wind speed and energy generation from September to December, while La Niña increases these parameters in the same months.

The aim of this study is to describe the space-temporal variability of the local wind circulation considering six distinct areas, into same river watershed: forest, pasture, and the transitions forest/pasture (two areas), urban/pasture and mine/pasture. It is important to highlight that, to achieve this objective, we used data observed through six automatic meteorological stations strategically distributed in a 42,000 km² area watershed in the Eastern Amazon region (Figure 1).

Figure 2 summarizes wind speeds recorded during 2017 at each hydrometeorological station for daytime (from 6 a.m. to 5 p.m.) and nighttime (from 6 p.m. to 5 a.m.) periods. Regardless of the hydrometeorological station, the wind speed during the nighttime period tends to be lower than that measured during the daytime period. However, it is noteworthy that at the same weather station, in both periods, the wind speed has a similar range.

At station SL, for the daytime period, the mean wind speed is 2.66 ± .99 m/s (.28–7.93 m/s), and 75% of the observations are values lower than 3.23 m/s. In contrast, at the SB station, for the same period, a mean wind speed of .7 ± .49 m/s (0–2.83 m/s) is observed, and 75% of the observations are values lower than 1.1 m/s.

The prevailing wind patterns in the region show the composition in the wind rose for each of the stations, simultaneously indicating the frequency of occurrences and their respective intensities (Figure 3). The data for AbF (Figure 3A) and IF (Figure 3B) show preferred winds from North-Southwest (N-SW) and from North-North-Northeast (N-NNE), respectively. The highest frequencies of wind direction at the GD station (Figure 3C) are North-Northeast (NNE) and South-Southwest (S-SW), with the highest most frequent speeds between 3.5 m/s and 5.0 m/s.

The data for the SL station (Figure 3D) show a higher frequency of Southeast (SE) and East-Southeast (ESE) direction winds and more intense and frequent winds than the other stations. In contrast, station SB is characterized by the highest percentage of calm in relation to the total records in its wind time series, reaching a total of approximately 60% (Figure 3E).

In AbF, the predominance of winds during the day is N and NE, with few frequencies of SW (Figure 4A). At night, there are much more frequent SW winds. However, the predominance of N winds remains at night (Figure 4B).

Hourly WT analyses are applied to the three-dimensional (scale, time, and energy intensity) components of the mean wind speed during 2017. In the present study, the x-axis shows the temporal length (in days) of the series, corresponding to a year of hourly data and totaling 8,760 h, and the y-axis represents the frequency (in hours) of the phenomena, i.e., the multiple frequencies that exist in the result. At all stations, there is a frequency between 16 h and 32 h with more intense signals in the center of the data period, which corresponds to the daily wind cycle (24 h).

The prevailing wind patterns in the region (wind speed and wind direction) were influenced by differences in land cover, land use and topography. For example, the SL station present higher frequency of Southeast (SE) and East-Southeast (ESE) direction winds and more intense and frequent winds than the other stations, which is possibly associated with the topographic aspect; the SL station is located at a high altitude (613 m) in relation to the other stations in this study, in addition to being located in a mine/pasture transition area. The highest percentage of calm recorded in station SB may be because it is in a forested area, where two factors can contribute to these results: a) a weak horizontal temperature gradient at the surface, common in wooded areas, and b) the very roughness of the forest that imposes a natural reduction in wind speed. Dos Santos et al. (2013) carried out a study in a dense tropical forest in the Amazon and found that the temperature at the top of the canopy gradually decreases to ground level, with a redistribution of energy throughout the day, important factors that characterize the vertical and horizontal fluxes and modulate the forest microclimate.

The predominance of winds is also influenced by other factors. The same is true for AbF station, in which the Northern Flow, both during the day and at night, may be associated with a larger scale segment, probably of a mesoscale circulation. According to Figures 4C, D and considering that the IF station is located east of the AbF station (Figure 1), the same flux of N is observed acting on these stations. This corroborates the existence of a predominant flux on a larger scale that overlaps the local effects of heating and cooling, daytime and nighttime, producing circulating breezes. On the other hand, the GD station is in a more isolated orographic condition (Figure 1), with predominant N winds. In the GD station, the local effect of the breeze is pronounced because during the day, the predominance is N-NE (Figure 4E), and at night, it decreases in frequency and its direction is reversed, switching to S-SW (Figure 4F), with weaker winds between .5 and 2 m s⁻¹.

The SL station is located between a higher region in the North and lower in the South (Figure 1), with predominant winds in the North during the day (Figure 4G). At night, there are more intense prevailing winds from SE and S-SE (Figure 4H) because when the winds blow from the SE, there are no barriers, while N winds are intercepted by the relief of the highest part of the mountain. Therefore, there is a local breeze system. The predominant winds at the SB station region during the day are SE and weak winds because SB is in an area of forest (Figure 4I). During the night, the winds become weaker and often negligible. The same is true for the frequency of nighttime winds at the SS station (Figure 4J).

In the WT analysis, the continuous line in the shape of a cone varying in both axes of graphics, which is called the cone of influence, encompasses areas with significant variances at a 95% level; i.e., the periods outside this cone should be disregarded because they do not have adequate statistical confidence (Vilani and Sanches, 2013). The highest frequencies are at the top of the y-axis, and low frequencies are found at the bottom of the graph (Figures 6A–E). It is possible to observe the signal between 8 h and 16 h, in which the wind variability is on the order of 12 h; i.e., every day and every 12 h during the dry period (June to October), there is a dominant variability in the wind field. It is suggested that this signal is associated with local breeze circulation. Teague et al. (2014), using wavelet analysis, identified a wind force associated with the sea breeze in the medium circulation of the Northwestern Gulf of Mexico, Texas. Likewise, in another study carried out on the southwest coast of Portugal, greater wind speed energy was observed in the daytime period in the analysis of wavelets (Morlet) (Lamas et al., 2017).

Regarding the frequency between 16 h and 32 h present in all stations, it is noted that these signals are weak at the beginning of the year due to the interference of larger-scale phenomena in the rainy season, while during the dry season the daily wind cycle strengthens, becoming more prominent. The dynamics of large-scale phenomena, such as ITCZ and SACZ, may be the explanation for the attenuation of this signal during the rainy season. De Souza et al. (2017), studying the climatic patterns of precipitation in the Eastern Amazon, suggests that both ITCZ and SACZ are the main responsible for the increase in rainfall volumes, when present over the region. These phenomena produce extensive cloud cover and frequent storms (De Souza et al., 2005; De Souza and Rocha 2006; Ferreira et al., 2015; Santos et al., 2015; De Souza et al., 2017; Zeri et al., 2018). Therefore, the highest frequency of intense rain and gusts is expected in the entire region under study. Thus, we strongly suggest that during the rainy season in the region, the circulation of local breeze is attenuated due to the combined effect of two forcings: the attenuation of insolation, due to the increase in cloudiness, which weakens the surface temperature gradient between forests, pastures and cities; and the increase in rainfall volumes characteristic of storms produced more frequently as a result of the presence of the ITCZ and/or SACZ. In addition, we cannot avoid considering that the large-scale flow may also be modulating the microscale wind variability.

This study corroborates with that carried out by Oliveira and Costa (2011), who found important changes in the behavior of the wind due to the influence of the ITCZ and the anomalies of Sea Surface Temperature (SST), as well as strong events of El Niño and La Niña in Northeast Brazil. Pascual et al. (2010) analyzed the connections between the wind speeds and the large-scale atmospheric field. These authors showed that large-scale atmospheric patterns dominate the wind field in Spain. Thus, large-scale phenomena can attenuate or intensify smaller-scale phenomena, and in the case of the stations in this study, the signal of the daily wind variability is weakened. At the SB station (Figure 5E), in the rainy season, this signal suffers this interference to a lesser extent than the other stations, possibly because station SB is in a forested region and the wind dynamics are different. The mean wind intensity in this region is lower due to the influence of dense vegetation, as previously shown in Figure 3E.

In the TW analysis of the meridional wind field of the GD station (Figure 6F), which is installed in a transition area, this variability is stronger because there is a predominance of pasture in the North and there is a forest in the South; therefore, the variation in the wind in the meridional direction is more evident both in the diurnal cycle and in the intraseasonal variability, while it is barely perceptible in the zonal component. On the other hand, station SS (Figure 7), also a transition area, has a weaker meridional component due to the distribution between the predominant NE and SE winds throughout the year, as shown in Figure 3F, resulting in the variability in the wavelets of the stronger zonal field (Figure 7). The SL station stands out, presenting strong intraseasonal variability in its zonal component and evident variability in the meridional component (Figure 7).

At all stations, the marked signal of wind variability in the dry season can be explained by the increase in the surface temperature. For example, in the analysis of the TW of the average air temperature for stations GD and IF (Figure 8), a strong daily cycle (24 h) signal is noted, as is noted for the wind speed (Figure 5). This signal is due to the alternation of heating and surface cooling, weaker in the beginning of the year due to the rainy season and more intense in the dry period because the surface reaches higher temperatures. Even the superficial thermal contrasts of forest/pasture (GD) and urban/pasture (IF) in this period are accentuated. This behavior (WT mean air temperature) is reproduced at the other study stations. These results corroborate those shown by Moura et al. (2014), which were obtained for Santa Rita Island, which is located 15 km south of the city of Maceió, Alagoas state, Brazil; the authors found that the magnitude of the wind, with a persistent signal at the 24 h scale during the dry period, is caused by surface heating, where the land-sea thermal contrast is higher and results in intense breezes in the region.

In this study, the surface wind variability was investigated over a hydrographic basin composed of six different environments: forests, pastures and the forest/pasture, urban/pasture and mine/pasture transitions.

The prevailing wind patterns in the territory of the IRW are influenced by vegetation cover, land use and topography, with the SL station, in the mine/pasture transition area, presenting the highest values of wind speeds and the SB station, in an area forest, the lowest values.

The most evident breeze circulation is identified at GD and SL stations (transition areas), signaled by the reversal of wind direction during the night. These stations are distinguished by relief, GD in an isolated orographic condition and SL located in a higher region. The same N wind flow is observed at AbF and IF stations, predominantly during the day and night, overcoming the effects of the local breeze.

In the wind speed WT analyses, more marked frequency signals are identified in the 24 h wind cycle (daily) and in the 12 h wind cycle associated with breeze circulation in all seasons, with different signal strengths.

The data show the interference of large-scale phenomena, such as the ITCZ and the SACZ, which operate in the region in the first months of the year. The results suggest that these systems modulate wind variability on a local scale.

The limitation of the time series is that the installation of the stations of interest is quite recent. It is hoped that this work can incentivize the refinement and broadening of studies on environmental impacts in the region.

In summary, it is recommended to incorporate data from non-neutral years and additional meteorological stations in further studies, in order to have a broader characterization of the wind field in the IRW.