The Twangiza open pit gold mine is located in the Mwenga Territory, South Kivu Province, DR Congo, at 45 km south-southwest of Bukavu city, the capital of the South Kivu Province. The road access from Bukavu is at 85 Km, through the national road (RN2). Its lowest average temperature is 13 degrees Celsius while its maximum is 22 degrees with a hilly topography averaging between 1700 - 2500 m asl. However, the highest close by hill goes up to 3250 m asl ( Basima et al., in press ).

The mine area consists of a network of steep hills and valleys that result in a dense stream catchment area, draining in a northwesterly direction. The Mwana River to the west and the Lulimbohwe River to the east surround the ridge that the mine occupies. Both the Mwana and Lulimbohwe are fast flowing, perennial, rock mountain rivers. They converge immediately to the north of the exploration camp, at the Mwana Bridge (Bridge 6 for the mine), forming the Kadubo River that flows through Luhwindja and Burhinyi, Ngweshe, and Kasika to reach the Ulindi River, and ultimately the Congo River, of which Ulindi River is a tributary (affluent).

The following stations were selected to install plastic manual rain gauges: Exploration Camp, Plant site 5 km away; TMF area 7 km away; and Cinjira village 10 km away, as shown in Figure 1 .

The topographic characteristics of the four stations indicated in Figure 1 are presented in Table 1 .

a. Main meteorological and rain gauge station.

Daily rainfall data were collected using manual rain gauges at four stations (Cinjira, Plant site, TMF site and Exploration Camp). An automated weather station Davis Vantage Pro 2, installed at the Plant site station collected for over a quindecinnial, at a 30-minute frequency interval: rainfall, wind speed, wind direction, temperature, atmospheric pressure, heat index dew point, and humidity. The data collected was transmitted to the Davis Vantage Pro 2 Wireless Receiver console, which allowed for regular and automatic downloading of data to a PC, as text files that were later converted to Excel.

Air dust was collected using the DustWatch CC multidirectional buckets, that permits sampling of 4 different wind directions (North, South, East and West), as operated by the wind, the direction of which opens and closes the relevant sample bucket to the atmosphere.

The buckets were mounted on a frame at 2.2 meters above ground to prevent thermal interference with the precipitation phenomenon as well as the height to which particulate can be lifted with a wind of 3.0 ms⁻¹. The buckets were partially filled with distilled water, to which bleach was added to prevent algal growth. This media then catches and retains any precipitant reporting into the unit opening.

The dust collected was filtered through a sub-micronic pre-weighed filter using a vacuum Buchner filter arrangement. The Metler Toledo precision balance was used for weighing the filters. Once the wet filtrate has been desiccated by evaporation of any retained moisture, the filter was reweighed to calculate the collected mass.

The relevant mass classification refers to Table 2 for historical limits ( Kuhn & Loans, 2016 ).

The fall-out dust standards from South Africa ( SANS, 2011 ) are shown in Table 3 and Table 4 .

Continued

a. (mg/m²/day), 30 Day average.

a. (mg/m²/day), averaged over 30 days.

The cross-sectional area of the buckets is a standard constant in all of the calculations representing the area over which fall-out collection has been made (0.02545 m², the internal measured diameter of the standard bucket used is 17.5 cm).

The 30-minutes rainfall data collected from the Davis Vantage automatic weather station and the daily rainfall from rain gauges were computed over the quindecinnial 2010-2024 using the applications Jamovi 2.3.28, JASP 0.18.3 and Statistica 10. These rainfall data over the quindecinnial were used to compute the average year rainfall trends of the study area. The non-parametric one-way ANOVA (Kruskal-Wallis test) and the parametric one-way ANOVA (F-test) and T-test were computed on Jamovi 2.3.28 and JASP 0.18.3 to compare means of the quantitative weather parameters (temperature, humidity, wind, heat index, air density, dew and atmospheric pressure) by stations, septennia (2011-2017 and 2018-2024), years and seasons. The interactions between seasons and septennia were calculated as well. The Tukey test (unequal n) was used for post hoc comparisons of means. The significance level of 5% was used for decision. The decision tree (machine learning module) based on the application SPSS 23.0 was used to compute the distribution factors (independent variables) of the quantitative parameters (dependent variables). Time series were processed on Past 4.11 to analyze the trends of the quindecinnial rainfall daily data. The Man-Kendall trend test was run as well as the moving average and the smoothing plots. Multivariate analysis of the quantitative data (temperature, humidity, wind, heat index, air density, dew point and atmospheric pressure) was computed as well using PCA (Principal Component Analysis) by stations, septennia, years and seasons as grouping variables.

The rainfall recorded on a daily basis over a quindecinnial is presented in Table 5 , from 2011-2024 at the Twangiza Plant site’s rain gauge at an elevation of 2341 m asl. The rainy and the dry months are highlighted in color scale. The two rainy seasons are also depicted, the main one spanning from September - January, followed by the small one extended from February - April. The dry months extends from June - August.

Figure 3 shows a stationary and relatively stable trend with no statistically significant pattern of the rainfall over the 2011-2024 period at all of the study stations. The oscillatory pattern is by year (14 years) and seasons (dry and rainy). The Mann-Kendall trend test results are the following:

Visibly there are similar trends from 2011-2023, highlighting no difference between the two septennia. This is probably due to proximity of the huge Itombwe montane forest regarding the regional climate. Regarding the study site microclimate, a probable positive impact of the progressive afforestation is noted, boosted by the community-based afforestation supported by the mine ecological progressive rehabilitation program (100 ha so far), which provide sensitization, tree seedlings distribution and follow up on growth. This community-based afforestation is much larger in space and cover areas from Kaziba, Luhwindja, Burhinyi and Ngweshe chiefdoms accounting in total for 635,934 tree seedlings distributed and followed up, covering 757 ha within the general community ( Twangiza Gold Mine, 2023 ).

No significant difference of daily rainfall was observed between the four stations regarding both septennia and between both septennia about the main rainy season that spans from September to January. This is probably due to the close proximity of the stations, being within 2.5 - 5 Km distance, the two furthermost stations being 7 Km a part. Though the sites are in close proximity, their elevation varies from 1800 m asl for the Exploration Camp station and 2600 m for the Cinjira station, an 800 m drop in elevation.

However, a significant difference was depicted between both septennia regarding the second rainy season (B = February-May) occurred only at the Plant site ( $\bar{x}$ = 31.9 ± 28.71; versus 39.3 ± 29.07; df = 1; F = 2.86; p = 0.09). This significant difference is displayed in Figure 4 .

Also, a significant difference between both septennia regarding the dry season (June-August) occurred at 3 sites, more so at the Plant site ( $\bar{x}$ = 3.9 ± 7.94 vs 8.0 ± 13.18; df = 1; F = 4.6; p = 0.03); except for TMF. This is shown in Figure 5 .

Peaks in rainfall recorded during the dry season at the Cinjira station are displayed in Figure 6 and Figure 7 .

Peak 1 shows more rainfall during the dry season at Cinjira in septennium 2 as compared to Plant site (from 15 - 30). Peak 2 shows similar rainfall at Cinjira as compared Plant site during the 2 septennia (from 60 - 90). Peak 3 shows more rainfall at Cinjira as compared Plant site during the 2 septennia (from 91 - 106). Peak 4 shows significantly more rainfall at Cinjira as compared Plant site during the 2 septennia (from 110 - 135). The dry season had more rainfall in the second septennium (2018-2024).

Aside the manual rain gauges’ data, the microclimate parameters obtained from the automated weather station at the Twangiza mine site per septennium and season are presented in Table 7 .

Continued

The septennium grouping for the automated weather station parameters are highlighted in Figure 8 .

This grouping is summarized in Table 8 .

The barometric pressure, wind speed and humidity values were higher during the first septennium (2011-2017) and they decreased during the second septennium (2018-2024) ( Figure 8 and Table 8 ). However, the values of temperature, heat index and dew point increased during the second septenuum; e.g 25.3˚C max dry season 2018-2024 vs 23.4˚C dry season 2010-2017; thus differences of 1.9˚C for temperature and 1.5˚C for heat index were observed. This is an indicator of warming in the microclimate along the quindecinnial.

The Kruskal-Wallis test shows a significant difference of temperature between the two septennia (X2 =3.69; df = 1; p = 0.05). The parametric ANOVA also shows a significant difference between seasons (p < 0.001) and there is a significant interaction between septennium and season (p = 0.04).

There is a highly significant difference between the dry season and the main rainy season, the dry season and the second rainy season and between the 2 rainy seasons (p < 0.001). This highly significant difference between the three seasons is observed during the 2 septennia.

The max wind speed variable distribution mainly varied by the septennium ( $\bar{x}$ = 13.3 ± 9.01 m/s; n = 132 357) followed by the month and year. During the first Septennium, the average high wind speed was 14.3 ± 9.26 m/s; n = 77 813. In the second Septennium, the average max wind speed was 11.9 ± 8.44 m/s; n = 56 544. The slowest max winds were noted in April, December and January 2024 ( $\bar{x}$ = 5.9 ± 4.21; n = 829) and in June, September 2023 ( $\bar{x}$ = 7.9 ± 6.07).

There is a highly significant difference between values of max wind speed between the 2 septennia, the months and an interaction between septennia and months.

Figure 9 shows the average high wind speed by season during both septennia.

During the dry season, the maximum wind speed values were highest 2014. In the main rainy season, the wind speed was highest in 2014 as well. The high wind speed peaked up during the second rainy season, only in 2011, 2014 and 2017.

The most frequent wind direction in the study area is ENE, NE followed by NNE during the first septennium (2011-2017). During the second septennium, the predominant wind direction is NNE followed by N.

Regarding dust fall-out precipitation, there was a highly significant difference between 2017 and 2023 (p < 0.001). For the difference between 2017 and 2023, the dust fallout has respectively 420.7 and 42.6 (mg/m²/day), thus a tenfold difference. A significant difference was observed for dust intensity between 2018 and 2023 (p = 0.05), respectively 308 and 42.6, thus a seven-fold difference. Mining activities were halted in December 2019, a reason of the decrease in dust fallout. A similar difference between 2017 and 2023 was observed in the southern direction, respectively 402.5 and 55.5, a seven-fold difference. A highly significant difference was observed between 2018 and 2023 with respectively 353.4 and 55.5, showing again a six-fold difference. In the eastern direction, dust fallout intensity in 2017 was highly significantly more than 2023, with a tenfold difference (497.5 vs 47.8). The trend was similar in the western direction, the dust fallout being 9 times more intensive in 2017 than 2023 (431.2 vs 46.0). The results show that when operational, the mine produced 7 to 10 times more dust that when not operational, which is attributed to mobile and mineral extracting equipment. However, the results show that whatever the situation, the dust level residential limit was not exceeded.

The correlation factors describing the meteorological parameters at the Twangiza site are presented in Table 9 .

The meteorological parameters are grouped into 3 assemblages as shown in Figure 10 : 1) Heat Index and relative humidity. Minimum, high and average temperature. 2) Dew point with relative humidity 3) Air pressure and Atmospheric pressure and air density. The Heat component negatively affected relative humidity, i.e. when temperature and heat increased, relative humidity diminished (γ = −0.52). Also when the heat and temperature increased, the air pressure diminished. When the air pressure increased, the dew point decreased (γ = −0.41). Temperature variance was utterly explained by component 1 (100%) alike heat index and maximum and minimum temperature, but negatively affected relative humidity. Factor 2 strongly explained the variance of the Dew point (94%) but less so for the relative humidity (41%). The air density and barometric pressure variances were both explained by component 3 (64% and 54.5%). Actually, the 3 components explained up to 92% of the overall variances.

In line with the current results, Allarakha (2024) stated that Temperature is another factor that affects air density. As the temperature increases, air molecules move faster and spread further apart when colliding. Higher temperatures and lower air pressure at high altitudes reduce air density. Increasing altitude decreases air pressure. Air becomes denser as air pressure increases.

In the Twangiza area, which extends from 1700 - 3000 m asl, the weather forecast around the mine site is unpredictable with the day weather changing frequently and suddenly to the extent that visitors and the general public always gear themselves as changes may be experienced at any time, as if they were experiencing all the seasons in just a day. In fact, it is known that a rise in barometric pressure is generally considered an improvement in the weather or clear skies, while low barometric pressure may mean worsening weather, and rain or snow, within a day or two. When barometric pressure is high, the air is heavy and dense and stays close to the earth’s surface, allowing the sun to shine. Low pressure allows the air to rise and form clouds, and produce rain, or snow if it is cold enough. Low pressure also produces high winds because air can move more easily, and when it happens over a warm ocean, it can generate a storm ( WMO, 2023 ).

Barometric pressure, wind speed and humidity values were higher during the first septennium (2011-2017) and they decreased during the second septennium (2018-2024). However, the values of temperature, heat index and dew point increased during the second septennium. Differences of 1.9˚C for temperature and 1.5˚C for heat index were observed. This is an indicator of warming in the area along the quindecinnial. Supporting this result, Liersch et al. (2014) predicted in neighboring Burundi that, although rainfall is expected to increase in the rainy season, the months leading up to season (August/September) are likely to be drier. Some of the models used predict a probable prolongation of the dry season. In previous decades, before 1984, May did not use to be a dry month as it was an integral part of the shorter rainy season and September has become drier. This depletion of May and September rainfall is a change in the study region ( Muhigwa, 1999 ).

Ngarukiyimana et al. (2021) found that the spatial distributions of T_max and T_min represent a significant warming trend over the whole country notably since the early 1980s. Surprisingly, T_min increased at a faster rate than T_max in R3 (0.27˚C vs. 0.07˚C/decade in March-to-May) and (0.29˚C vs. 0.04˚C/decade in October-to-December), resulting in a significant decrease in the DTR. This is another confirmation of warming in Rwanda from 1961 to 2014 within 3 elevation regions: 1000 - 1500 m; 1500 - 2000 m and ≥ 2000 m).

Engdaw et al. (2021) located the increase in temperature and heat in Africa around 2016 as compared to 2010. According to them, most data sets agree in identifying 2010 as a peak heat year over Northern and Western Africa while Eastern and Southern Africa experienced the highest heat wave occurrence in 2016. Their Results clearly reveal that heat wave occurrences have emerged from natural climate variability in Africa. In addition to temperature, also humidity, wind and incident radiation play a role in altering the occurrence and characteristics of heat waves ( Engdaw et al., 2021 ).

Significant differences occurred between both septennia regarding the dry season (June-August), with more rainfall in the second septennium (2018-2024). This confirms a microclimate change in the study area, even at a range of 3 - 10 km distance, indicating intermittent rains. Van vyve (2006) , who studied the spatial variation in rainfall in southern Niger (non-arid) tropical climate, showed over a radius of 1 - 20 km that the best way to manage rainfall-related risks in agriculture is to spread the fields around the village; since the probability of rainfall in two places a few kilometers apart is highly random. From a spatial point of view, she observed a great variability in annual totals; e.g. for the year 2000, she noted a zone of around 300 mm rainfall spaced about 5 km apart from an area where annual rainfall exceeded 500 mm. For some years (2000, 2003 and 2004), She observed a limited number of large wetter and drier zones. In 2001 and 2005, on the other hand, the annual totals were organized into limited wetter and drier zones, which are very heterogeneously organized. She noted that farmers had been spreading their fields around the village for a long time. They have always been aware of the risks associated with variability of rainfall in their region.

This difference of rainfall levels between nearby places is commonplace in various regions and climates, as shown by local or national high-resolution precipitation analyses, especially at the boundary of any neighbor isohyets. da Silva et al. (2020) illustrated such intermittence of rains in a reservoir catchment in Brazil.

A microclimate is a small area that has different atmospheric conditions than the surrounding area ( Camus, 2017 ). Ellis & Eaton (2021) indicated that microclimates can be found in most places but are most pronounced in topographically dynamic zones such as mountainous areas, as the case of Twangiza, but also on islands, and coastal areas. Agele (2000) argued that opencast mining activities, alike in Twangiza, do cause soil disturbance and produce changes in the concentrations of toxic elements, dusts, biogenic gases and microclimatic gradients. These gradients in microclimate have implications for leaf to air vapor pressure differentials that control plant water relations and energy balances. There is therefore high irradiance, soil water deficits, high temperatures and vapor pressure deficits, variable substratum depths and toxic levels of some minerals ( Agele, 2000 ).

The results show that when operational, the mine produced 7 to 10 times more dust that when not operational, which is attributed to mobile and mineral extracting equipment. Momoh et al. (2013) observed that suspended particulate matter in the air ranged from 60.25 to 1820.45 µg/m³ at the Mukula mine in Limpompo, South Africa. Such measure includes wet drilling and blasting, sprinkling of water on the mine roads and planting of vegetation around the mines and neighboring communities. Dayda (2024) indicated that during many of the extraction and processing processes related to the production and processing of rock and mineral products, dust particles are created and can become airborne. There are three primary sources of dust in mines: drilling and blasting, loading and hauling, and crushing and grinding. As shovels scrape rock and trucks cross complex terrain, this activity generates a lot of dust. The movement of these trucks and transferring material to conveyor belts continues the cycle of dust creation, putting workers and anyone nearby in danger. Crushers break down rocks into smaller fragments, whereas grinding means the reduced size of the crushed materials, which gives fine products than crushing. Each step releases a fine dust cloud, contaminating the air and posing inhalation risks.