Spatiotemporal variability of Humidex Index over the Northeast Region of Brazil

Thermal comfort, Humidex, is a very important variable to assess the degree of thermal stress on living beings. Because it depends on climatic variables, Humidex has a space-time-intensity variation conditioned to various weather and climate attenuators. Thus, the objective of this work is to verify possible trends and spatiotemporal analysis of the Humidex index on the Northeast of Brazil (NEB). ERA5 data distributed by ECMWF was used for the period from 1990 to 2019. It was found that the MATOPIBA region was the one that stood out the most with positive discomfort trends over NEB, as well as the region with the largest Humidex indexes. While on the NEB coast there were low Humidex valleys which may be associated with oceanic influence. It was observed that the night time has a positive trend on NEB, especially on MATOPIBA. In addition to the hourly scale, it was also found that on the monthly, seasonal and annual scale, this region of MATOPIBA has a highlight. This is due to the fact that the replacement of the natural forest by agricultural production, in which, it directly affects the energy balance in this region. MATOPIBA ISSN:1984-2295 Revista Brasileira de


Introduction
Temperatures outside a comfort range affect the human health and are related to increased stress and mortality. Reports of the various Brazilian professional organizations (CPTEC, INPE) shows that global warming causes climate change expressed for example by the temperature increasing. In Brazil, and particularly, its Northeast region (NEB), the mean temperature has been increasing 25% f aster than the global average since 1910, implying that the population may be particularly exposed to the effects of increasing temperature relative to other populations (NOAA, 2017), posing risks for human health. In these circumstances, knowledge of the conditions of thermal discomfort it is necessary, especially in areas with high mortality risk due to excessive temperatures. The thermal comfort is difficult to be quantified because it is necessary to consider a range of environmental and human factors to decide what makes a person to feel thermal comfortable. The first factors can include air temperature, radiant temperature, wind speed and air humidity, while the later are attributed to so called "metabolic heat" (Brake and Bates, 2003).
According to World Health Organization, the continuous exposure of an individual to an environment whose thermal conditions are adverse to those he is accustomed causes the organism to undergo physiological changes to adapt to the environment (Buitrago et al., 2016). This is particularly important, as heat-wave episodes are became more frequent and may cause hyperthermia, dehydration, and respiratory diseases. In the other hand, extreme cold temperatures can result in ischemic heart, cerebrovascular, and respiratory diseases (Schwarz et al., 2010).
As global models predict significant changes in the climate of the NEB regions, it is expected that the increased temperature would directly affect the well-being of the population, since the variations of air temperature, precipitation, relative humidity, and wind speed are highly important when quantifying the thermal comfort of a region. More importantly, the NEB is exposed to several weather and climate attenuating systems that can lead to an increase in the level of comfort or thermal discomfort such as the Intertropical Convergence Zone (ITCZ), Breezes, High Level Cyclonic Vortexes (HLCV), stability lines, convective systems, wave disturbances, and less frequently, tropical cyclones (Hastenrath and Heller, 1977;Kousky and Gan, 1981;Molion and Bernardo, 2002;Fedorova et al., 2009;Rodrigueset al., 2010;Fedorova et al. , 2020). In addition, in El Niño/La Niña years, the precipitation regime decreases/increases in the Brazilian Northeast (NEB) and increases/decreases in the south region of Brazil (Grimm et al. 1998).
In that sense, as the climate conditions in the NEB tends to be worsening due to various phenomena at different spatiotemporal scales such changes will cause discomfort in the NEB population, and may require adaptation of the organism during these periods. In that sense, this study is an efforts to assess the degree of thermal comfort evaluate throughout the NEB based on Humidex index estimations. Evaluating the spatiotemporal variability of the indices and ocean-atmosphere forcings, will help to identify the potential correlation between external forcing and thermal conditions within the region.
It is known that the interaction between climate, weather and city configuration is complex (Darbani, et al., 2021;Matallah, et al., 2021;Pereira and Bourscheid). According to Pereira and Bourscheid (2021) although people who live in buildings located near the sea are also exposed to high temperatures, the fluctuation remains in a range classified as comfortable, mainly due to the shading and the sea breeze. However, the growth of cities has increased thermal discomfort due to the replacement of vegetation by concrete among other factors (Sharma et al., 2021). According to Meili et al. (2021) the increase in humidity due to vegetation restricts was regarded as the factor responsible for the reduction of the thermal average to 3 °C, as Silva, W. T. C., Cavalcante, G. H. well as, urban growth has an negative effect on thermal comfort microclimate (Darbani, et al., 2021). That been said, urban developments should develop strategies that seek mitigation and adaptation, taking advantage of the existing green infrastructure (Matallah, et al., 2021, Meili et al., 2021.
In that sense, this study focusing on understand the spatiotemporal trends of the Humidex index over the NEB, taking in consideration, how the change in land use can impact the thermal comfort indices? How does the land-ocean-atmosphere interaction may impact the Humidex indices?
Figure 1-Brazil and study area in red. Altimetry is represented by the colorbar.

Humidex Index
The calculation of the Humidex index was based on the methodology adopted by Thom's (1959) andd'Ambrosio Alfano et al (2011), and are expressed in the equations below: Where, = 6.112. (10 7.5( 237.7+ ) ) 100 (2) To calculate the relative humidity (RH), the dry bulb ( ) and wet ( ) temperature were used according to equation (3) from ERA5. RH corresponds to the ratio of the actual vapor pressure ( (T)) and saturation ( ( )) at the same temperature, in percentage: The actual vapor pressure ( (T)) and the saturation pressure ( ( )) were calculated from equations (4) and (5): Where, ( ) and ( ) correspond to the actual vapor pressure at temperature T, and vapor saturation pressure at temperature , respectively.
( 0 )) is the vapor saturation pressure at a reference temperature 0 , = 2.5 × 10 6 / corresponds to the latent heat of water evaporation and = 1000 = 461.52 / . = 8.3144621 /( ) is the gas molar constant and = 18.01528 / is the molecular weight of water vapor. The reference temperature and vapor saturation pressure used were 0 = 273,15 K and ( 0 ) = 6,11 mb. The Humidex limits and the thermal comfort index are shown in Table 1.

Humidex range
Thermal discomfort level

Mann-Kendall test
To calculate the Humidex trend on the hourly, monthly, seasonal and annual scale, the Mann-Kendall (MK) non-parametric statistical test was used (Mann, 1945;Kendall, 1975). The S statistic is obtained by the sum of all counts, where the sign ( − ) is determined by (6): The MK test analyses data collected over time for consistently increasing or decreasing trends. The test gives one of two options: null hypothesis (no monotonic trend in the series) or alternative hypothesis (there is a trend and it can be positive, negative or null). The variance obtained from the MK statistic is used to define the value for a specific significance level.

Results and Discussion
Hourly and Monthly Climatology of the Humidex Index Figure 2 shows the spatiotemporal average of the Humidex indices for the NEB region from 1990 to 2019. It is observed that the thermal comfort index increases from 30 ºC to 24 ºC between 22Z (19h local) to 10Z (07h local), night time to middle morning period, following the decrease of air temperature (24 ºC to 36 ºC). Such pattern is more prominent over regions with high altitude such as the Chapada Diamantina region located over the central region of Bahia. Meanwhile, during daytime (8 am local time) to later afternoon (6 pm local time) period it was found peak Humidex indices of 39 ºC on Maranhão and Piauí regions which is correspondent to greatest discomfort according to Humidex classification, lasting for about 10 hours, on average.  According to the Humidex index, regions located at 5ºS latitude revealed indexes above 35ºC from October to May. It was found that the areas near the coast has a relatively constant index throughout the year. Can is associated with the influence of the ocean as a source of heat and humidity on the coast.
It was observed that between 10 ºS to 20 ºS the Humidex index declines from January (29 ºC) to August (> 23 ºC), mainly on the Chapada Diamantina and central region of the state of Bahia as well as in other regions of high altitudes such as in Pernambuco, Alagoas and Paraíba. This can be attributed to the position of the sun, altitude and weather and climate factors. For instant, studies showed that during El Niño, cold fronts are less frequent in the region inhibiting rainfall, and the other hand, La Niña is characterized by frequent cold fronts inducing intense rainfall and lower air temperature (Pereira et al., 2011).
Despite the temporal variability of precipitation that influences the energy balance, the NEB region reveals a constant energy balance throughout the year. According to Correia Filho et al (2019) they showed that the rainy season in the Amazon, Caatinga and Cerrado biomes occur between January and April, but varying in intensity, not seen in the Atlantic Forest. The regions with greatest comfort levels were Bahia and Pernambuco with value of 26 ºC to 23 ºC during May to September (Figure 3). Correia Filho et al (2019) show that the extreme south of the Caatinga, Cerrado and Atlantic Forest are influenced by frontal systems and, consequently, influence the elevated thermal comfort levels in the region.

Hourly, Monthly, Seasonal and Annual Trend of the Humidex index
It can be seen that Figure 4 shows that the positive hourly trend occupies a large part of the NEB region. During the night time period (22Z to 04Z), the MATOPIBA region (Figure 4, T.22 to T.04), presented the highest trends on temperature in NEB with a positive rate varying from +0.008ºC/Hour/Month to 0.01ºC/Hour/Month On the coast region, there is a slight negative trend during the night time period, which may be associated with increased sea breeze intensity and local precipitation. Studies have shown that sea breeze winds have significant cooling potential, impacting thermal comfort levels (He et al., 2020a, He et al., 2020b, Liu et al., 2020. The monthly Humidex index trend is shown in Figure 5. Again, the MATOPIBA region reveals a very significant trend varying from +0.04ºC/month to +0.08ºC/month, mainly in March, October and November. The small Humidex trends are seen in the coastal region which may be associated with oceanic influence (Moura et al., 2009;de Medeiros et al., 2020).  An increasing seasonal trend is observed throughout a large area covering the NEB region ( Figure 6). The states of Maranhão and Piauí registered trends exceeding +0.06 ºC/month. Approximately 2.41% of the NEB shows a trend equal to or greater than +0.06 ºC/Year and 24.21% of the NEB with a trend equal to or greater than +0.05ºC/Year. The increased seasonal trend observed in the São Francisco river basin +0.04ºC/Year and MATOPIBA ( Figure 6) can be attributed to the change in land use with intensification of soybean and other agricultural crops, causing a direct effect on the local and regional climate. According to Cunha et al.
(2015a), they showed that changes in land cover influenced the climate processes at a large scale due to the changes in surface energy and carbon balance. Moreover, Angelini et al. (2017) showed that the latent heat flow in the Cerrado may reach levels 60% higher than in agronomic crops, while the sensitive heat flow can increase up to 50% higher in agricultural areas compared to Cerra do areas. Overall, these results play a role on the cause-effect of replacing vegetation, and observed thermal discomfort increase in the study region.  The annual trends showed a similar variability compared to the seasonal, where the later indicated values exceeding 0.06 ºC/Year, and more intense in the Maranhão and Piauí states (Figure 7). Approximately 3.13% of the NEB shows a trend equal to or greater than 0.06 ºC/Year, expanding to 26.57% of the NEB with a trend equal to or greater than 0.05 ºC/Year. This represents an increase of 1.8 ºC to 1.5 ºC, respectively, in 30 years. Again, the change in land use directly affects the energy balance, rainfall and air temperature in the region (Narisma et al., 2003;Cunha et al., 2015a;Hirota et al., 2011;Angelini et al., 2017, Gaspari et al., 2018Silva et al., 2018;Cabral Júnior et al. 2019;Marçalet al., 2019).
According to Costa et al. (2020) the NEB region shows unmistakable signs of warming, in which, in the last decades, there have been predominant tendencies for a decrease in the percentage of cold nights (−8.4% days/decade), an increase in the number of hot nights (10.6% days/decade), as well as in the number of days per decade in which the minimum temperatures exceeded the threshold of 20 °C, and a tendency to increase heat waves events. These changes are due to some factors that, according to Carvalho et al. (2020) are related to the reduction in the number of precipitation/days, mainly along the coastal zone, as well as reduction in precipitation rate within 150 to 300 km from the coast, and in the Sertão region (Vieira et al., 2020; Marengo et al., 2020). In addition to climatic factors, the change in land use with replacement of native vegetation by soy has altered the energy balance. According to Vieira et al. (2021), they showed that approximately 4677 km 2 (0.63%) of MATOPIBA region is under strong signs of degradation, mainly in pasture areas and fields, commonly used for pasture. Consequently, in addition to the gradual replacement of native vegetation, solar radiation directly affects the soil, further warming this region. Table 2 shows that minimum 30.4°C and maximum 39.6 °C values of Humidex index (HI) were found in Maceió/AL and Teresina/PI, respectively, while peak average of 36.7 °C was found in Teresina/PI. Annual HI trends were statistically significant throughout the cities with the exception of João Pessoa/PB and Recife/PE, and indicates an overall positive tendency ( Table  2). The highest annual trends were on cities located in the northern sector of NEB, represented by the city Balsas/MA with 0.061 °C/Year. According to studies performed by  the maximum and minimum temperatures and the extreme temperature indexes showed a strong positive Silva, W. T. C., Cavalcante, G. H. trend in practically all series (p<0.05) over the MATOPIBA region, Maranhão state, Tocantins, Piauí and Bahia. The authors also pointed out that this region has been under a significant warming process in the last 3 decades.  (Figure 8). The results indicate that the growing trend of environmental discomfort is associated with the increase in soybean cultivation on Balsas and Tasso Fragoso (Figure 8). These two municipalities are the largest producers of soybeans in the state of Maranhão.
We applied pearson's correlation between the Humidex index and the soybean area in both municipalities and found correlations of 0.82 (p = 2.879e-08) (Balsas) and 0.83 (p = 1,101e-08) (Tasso Fragoso), which confirms that the gradual replacement of natural vegetation by soy reflected positively on the thermal discomfort over the region of Maranhão and Piauí at approximately 2 ºC (Figure 8). In order to verify the influence of largescale climate such as El Niño and La Niña events on the Humidex index oscillation, kernel density analyses was applied over both data series. It can be seen in Figure 9 that the 2016 El Niño was more intense than the 1998 El Niño with a Humidex index exceeding 40 ºC. Cunha et al. (2015b) showed that the drought from 2012 to 2013, affected about 85% of the Brazilian semiarid region, affecting precipitation, soil moisture, recovery and the effects of vegetation memory. Kogan confirms that El Niño2015/2016 was much stronger than the extreme event of 1997-1998 affecting different regions of the globe. It was observed that in the La Nina event in 2011, the Humidex was thermally more comfortable than La Niña of 2018, similar to the El Niño in 1998. Silva

Conclusions
This study aimed to study the Humidex index on different scales such as hourly, monthly, seasonal and annual and their respective trends. It was noticed that the MATOPIBA region was the region that stood out the most with positive discomfort trends, as well as the region with the highest Humidex indexes. In that, it was observed that the night time has a positive trend on NEB, especially on MATOPIBA. In addition to the hourly scale, it was also found that on the monthly, seasonal and annual scale, this region of MATOPIBA has a highlight. This is due to the fact that the replacement of the natural f orest by agricultural production, in which, it directly affects the energy balance in this region. It was observed that the NEB's east coast region was influenced by the Atlantic Ocean.
It was found that during the daytime, the northern region of the NEB has the highest rates of thermal discomfort and the central-southern region of the NEB, during the nighttime, has the most thermally comfortable rates.
It was observed that the Humidex index in the range of 5ºS presents from October to May more uncomfortable with indexes above 35ºC. Note that the NEB coast has a relatively constant index throughout the year due to the influence of the ocean. Between 10ºS and 20ºS, on the continental NEB region, the Humidex index declines from January (29ºC) to August (> 23ºC), mainly on the Chapada Diamantina and central region of the state of Bahia and some other h igh altitude regions as in Pernambuco, Alagoas and Paraíba, returning from September. This is due to the position of the sun, altitude and weather and climate factors.
It was found that in the La Niña period it was considered relatively more comfortable thermally, whereas in the El Niño period it was reversed. The reduction and increase in precipitation on the NEB due to El Niño and La Niña led to the attenuation of the Humidex index.
Therefore, the Humidex index was studied and the attenuation of this index in relation to the time-spatial scale was altered due to meteorological systems, change due to the modification of land use and attenuation due to the El Niño and La Niña phenomena. Silva, W. T. C., Cavalcante, G. H. Behaviours Affecting Students' Learning in Lecture Theatres in the Humid Tropics. Buildings, 6, 18.