Vegetation Index and Air Temperature Behavior in Tefé-Amazonas , Brazil

Índice de Vegetação e Comportamento da Temperatura do Ar em Tefé-Amazonas, Brasil R E S U M O O objetivo deste trabalho foi analisar a modificação do uso da terra no comportamento da temperatura do ar no município de Tefé-AM. Para isso, foram coletados os dados de temperatura mínima e máxima do ar da estação meteorológica do INMET em Tefé do período de 1993-2012. Os dados foram tratados com técnicas estatísticas descritivas como: média, máxima, mediana, coeficiente de variação, amplitude, quantis e desvio padrão. Além disso, foi realizado o processamento de dados para elaboração de índice de vegetação da diferença normatizada (NDVI) para avaliar as transformações do uso da Terra em Tefé. Os resultados mostraram a substituição da área de floresta com a expansão da malha urbana, em o balanço de radiação superfície-atmosfera tem sido alterado, demonstrando o aumento temporal das temperaturas mínimas e máximas devido o processo de apropriação da natureza em escala local. Concluise que é necessário que o ordenamento territorial leve em consideração as condições físico-naturais da paisagem amazônica. Palavras-chave: temperatura do ar, NDVI, uso da terra, Tefé. Introduction Air temperature increases observed worldwide indicate serious socio-environmental problems to be faced. Studies performed in the Brazilian Amazon have shown that changes in land use caused by deforestation and replacement of forest areas with pasture and monocultures are leading to an increase in the surrounding temperature and, at a regional scale, to a decrease in rainfall and prolonged drought periods (Bagley et al., 2014; Buarque et al., 2010; Nobre, 2001). Many doubts persist mainly on the reliability of climate models used in studies of climatic variability in the Brazilian Amazon since the data series analyzed are not long enough and there is a shortage of observational climatic data and scientific research in parts of this region. Rainfall is high in the region, with pluviometric totals above 3500 mm per year in some localities. According to Marengo and Nobre (2009), about 50% of the water vapor added to rainfall formation returns to the atmosphere by evapotranspiration, demonstrating the importance of forest conservation and public policies for land use planning, integrating the dynamic factors of the landscape. According to Marengo and Nobre (2009), Revista Brasileira de Geografia Física v.11, n.03 (2018) 864-876. 865 despite the important feedback of precipitation and temperature due to the high evapotranspiration in the Amazon, assessing the hydrological cycle is difficult due to a lack of surface stations and frequency of space-time measurements. We should also consider that different types of vegetation in a multiple biodiversity are responsible for evapotranspiration in a higher or lower quantity. Thus, the assessment of land use and biota is important for climate analysis in the region. The Amazon region has a little thermal amplitude. Due to its location near the equator, it receives a high amount of solar energy during the entire year, which is reflected in high temperatures regardless of the season, except for cold spell days. A temporal-spatial analysis of temperature by means of data series allows the diagnosis of the society’s role reflected in changes in landscape characteristics, influencing climate to a regional and local extent. Climate elements, including air temperature, are intrinsically related to the balance of energy in the Earth-atmosphere system, derived from the input and output of ultraviolet and infrared radiations. However, air temperature is also spatially influenced by geographical factors of climate such as latitude, altitude, relief, continentality, maritimity, vegetation, and anthropic activities. In studies on the behavior of a microclimate, a previous knowledge about some of the elements and climatic factors of the place under analysis is necessary. Land-use changes from natural to agricultural systems in the Amazon may result in a number of socio-environmental problems, such as loss of biodiversity, climate change at mesoand microscale, among others related to the processes of appropriation of nature. These aspects might be closely related to the changes of the Amazon landscape, altering the radiation balance and convective process associated with rainfall formation. Thus, the main environmental problems related to land use are observed in areas of incompatibility between use and physical-natural characteristics, which correspond to areas of deforestation and burnings. These areas are normally used inadequately for the development of activities such as livestock, agriculture, and logging, where their physical-natural limitations are not respected. Moreover, territorial expansion of urban areas towards the Amazon region in a disorderly and unequal way interferes with the land use and occupation and hence its microclimate. In this context, the aim of this study was to analyze the effect of land use changes on the air temperature behavior in Tefé, AM, Brazil. Methodological procedures Series of monthly and daily air temperature data in the region of Tefé were collected from the Meteorological Station of the Brazilian National Institute of Meteorology (INMET). The INMET conventional meteorological station, which measures the maximum and minimum air temperature, started its operation on January 4, 1929. However, the historical data available in the Historical Database platform (BPMEP) begins from the 60/70’s. Daily, monthly, and annual data presented many gaps, being considered for the analysis of this study a 20-year series from 1993 to 2012, which is the most representative period and of better quality. Fault fills were performed in months with missing data. Monthly data with gaps were treated by means of to fill in the missing data, using the data of the nearest meteorological station located in Coari, which also has a low altitude. These data were statistically treated by statistical descriptive techniques such as mean, median, coefficient of variation, standard deviation, maximum and minimum values, amplitude, and quantiles. The normalized difference vegetation index (NDVI) was also obtained by processing the data. This index consists of a parameter to establish reflectance levels of a healthy vegetation, allowing the analysis of areas of dense and dispersed vegetation, as well as over-time changes in the forest such as replacement with pasture areas, deforestation, agriculture, and urban expansion. We selected Landsat 5 TM satellite images from 1991 to 2011, which encompass the vicinities of the urban area of Tefé. In this sense, bands 4 and 5, which comprise the visible red (R) and near-infrared (NIR) channels of Landsat 5 TM images, were used to determine the vegetation indices, which were then processed by the SPRING software, using the following arithmetic operation: C = Gain*((A+B)/(A−B)+Offset Wherein: C is the NDVI, A is the nearRevista Brasileira de Geografia Física v.11, n.03 (2018) 864-876. 866 infrared (NIR) channel, and B is the visible red (R) channel, as in Figure 1. After applying the arithmetic operation, an image is generated by the SPRING application in the Geographic Information System environment. This image contains the vegetation index, whose variation of digital levels of pixels is from −1.0 to 1.0. Values near −1.0 represent areas with a low vegetation index such as bare soil, deforestation, grassy vegetation, alluvial deposits, urban area, and water bodies, while indices close to 1.0 represent areas with arboreal vegetation such as forest areas. The NDVI of the processed images allowed a landscape analysis where positive values (> 0) stand for areas of green dense vegetation with arboreal size, whereas the null values (or close to zero) are surfaces with no vegetation and, finally, areas with negative values represent water bodies and clouds. Figure 1: Organization chart to elaborate the NDVI. Landscape and air temperature changes in Tefé, Amazonas state, Brazil Most of the deforestation concerns in the Brazilian Amazon are focused on the agricultural frontier known as the “Arc of Deforestation”, among the states of Rondônia, Mato Grosso, Pará, and southern Amazonas. However, deforestation process is less intense in the interior of the Amazon when considering only the deforested area. This process is equally harmful to local populations and biodiversity, both from an environmental and social point of view. The replacement of green/forest areas is associated with changes in local microclimate, leading to alterations in quality of life of the population living in the surrounding cities. Tefé is located in the Middle Solimões region, in the Amazonas state (Figure 2), about 520 km far from the capital Manaus in a straight line. This city presents an equatorial climate with high temperatures and wide rainfall volumes (ALEIXO and SILVA NETO, 2015a; CORDEIRO and FREITAS, 2016). The highest total monthly rainfall is concentrated between January and May, a period known as the flood season in the Central Amazon, and the lowest total monthly rainfall occurs from June to December, characterizing the dry season (ebbing) (ALEIXO and SILVA NETO, 2015b). The analysis in a local perspective reveals significant changes in the landscape near the urban area and in road margins passing through Tefé. Revista Brasileira de Geografia Física v.11, n.03 (2018) 864-876. 867 A temporal analysis using images from different periods, within about two decades, allowed observing that, by means of NDVI, the analyzed cities of the Middle Solimões region presented a representative loss of areas of dense and healthy vegetation. In this sense, in the vicinities of urban areas of the analyzed cities, forested areas have been replaced with areas of deforestation, burning for agricultural purposes, logging, and livestock (Figures 3 and 4). The main change in NDVI was an increase, in 2011, of areas with values betwee


Introduction
Air temperature increases observed worldwide indicate serious socio-environmental problems to be faced. Studies performed in the Brazilian Amazon have shown that changes in land use caused by deforestation and replacement of forest areas with pasture and monocultures are leading to an increase in the surrounding temperature and, at a regional scale, to a decrease in rainfall and prolonged drought periods (Bagley et al., 2014;Buarque et al., 2010;Nobre, 2001).
Many doubts persist mainly on the reliability of climate models used in studies of climatic variability in the Brazilian Amazon since the data series analyzed are not long enough and there is a shortage of observational climatic data and scientific research in parts of this region.
Rainfall is high in the region, with pluviometric totals above 3500 mm per year in some localities. According to Marengo and Nobre (2009), about 50% of the water vapor added to rainfall formation returns to the atmosphere by evapotranspiration, demonstrating the importance of forest conservation and public policies for land use planning, integrating the dynamic factors of the landscape.
According to Marengo and Nobre (2009), despite the important feedback of precipitation and temperature due to the high evapotranspiration in the Amazon, assessing the hydrological cycle is difficult due to a lack of surface stations and frequency of space-time measurements.
We should also consider that different types of vegetation in a multiple biodiversity are responsible for evapotranspiration in a higher or lower quantity. Thus, the assessment of land use and biota is important for climate analysis in the region.
The Amazon region has a little thermal amplitude. Due to its location near the equator, it receives a high amount of solar energy during the entire year, which is reflected in high temperatures regardless of the season, except for cold spell days.
A temporal-spatial analysis of temperature by means of data series allows the diagnosis of the society's role reflected in changes in landscape characteristics, influencing climate to a regional and local extent. Climate elements, including air temperature, are intrinsically related to the balance of energy in the Earth-atmosphere system, derived from the input and output of ultraviolet and infrared radiations.
However, air temperature is also spatially influenced by geographical factors of climate such as latitude, altitude, relief, continentality, maritimity, vegetation, and anthropic activities. In studies on the behavior of a microclimate, a previous knowledge about some of the elements and climatic factors of the place under analysis is necessary.
Land-use changes from natural to agricultural systems in the Amazon may result in a number of socio-environmental problems, such as loss of biodiversity, climate change at mesoand microscale, among others related to the processes of appropriation of nature.
These aspects might be closely related to the changes of the Amazon landscape, altering the radiation balance and convective process associated with rainfall formation.
Thus, the main environmental problems related to land use are observed in areas of incompatibility between use and physical-natural characteristics, which correspond to areas of deforestation and burnings. These areas are normally used inadequately for the development of activities such as livestock, agriculture, and logging, where their physical-natural limitations are not respected.
Moreover, territorial expansion of urban areas towards the Amazon region in a disorderly and unequal way interferes with the land use and occupation and hence its microclimate. In this context, the aim of this study was to analyze the effect of land use changes on the air temperature behavior in Tefé, AM, Brazil.

Methodological procedures
Series of monthly and daily air temperature data in the region of Tefé were collected from the Meteorological Station of the Brazilian National Institute of Meteorology (INMET).
The INMET conventional meteorological station, which measures the maximum and minimum air temperature, started its operation on January 4, 1929. However, the historical data available in the Historical Database platform (BPMEP) begins from the 60/70's.
Daily, monthly, and annual data presented many gaps, being considered for the analysis of this study a 20-year series from 1993 to 2012, which is the most representative period and of better quality. Fault fills were performed in months with missing data.
Monthly data with gaps were treated by means of to fill in the missing data, using the data of the nearest meteorological station located in Coari, which also has a low altitude.
These data were statistically treated by statistical descriptive techniques such as mean, median, coefficient of variation, standard deviation, maximum and minimum values, amplitude, and quantiles.
The normalized difference vegetation index (NDVI) was also obtained by processing the data. This index consists of a parameter to establish reflectance levels of a healthy vegetation, allowing the analysis of areas of dense and dispersed vegetation, as well as over-time changes in the forest such as replacement with pasture areas, deforestation, agriculture, and urban expansion.
We selected Landsat 5 TM satellite images from 1991 to 2011, which encompass the vicinities of the urban area of Tefé.
In this sense, bands 4 and 5, which comprise the visible red (R) and near-infrared (NIR) channels of Landsat 5 TM images, were used to determine the vegetation indices, which were then processed by the SPRING software, using the following arithmetic operation:

C = Gain*((A+B)/(A−B)+Offset
Wherein: C is the NDVI, A is the near-infrared (NIR) channel, and B is the visible red (R) channel, as in Figure 1.
After applying the arithmetic operation, an image is generated by the SPRING application in the Geographic Information System environment. This image contains the vegetation index, whose variation of digital levels of pixels is from −1.0 to 1.0. Values near −1.0 represent areas with a low vegetation index such as bare soil, deforestation, grassy vegetation, alluvial deposits, urban area, and water bodies, while indices close to 1.0 represent areas with arboreal vegetation such as forest areas.
The NDVI of the processed images allowed a landscape analysis where positive values (> 0) stand for areas of green dense vegetation with arboreal size, whereas the null values (or close to zero) are surfaces with no vegetation and, finally, areas with negative values represent water bodies and clouds.

Landscape and air temperature changes in Tefé, Amazonas state, Brazil
Most of the deforestation concerns in the Brazilian Amazon are focused on the agricultural frontier known as the "Arc of Deforestation", among the states of Rondônia, Mato Grosso, Pará, and southern Amazonas. However, deforestation process is less intense in the interior of the Amazon when considering only the deforested area. This process is equally harmful to local populations and biodiversity, both from an environmental and social point of view.
The replacement of green/forest areas is associated with changes in local microclimate, leading to alterations in quality of life of the population living in the surrounding cities.
Tefé is located in the Middle Solimões region, in the Amazonas state ( Figure 2), about 520 km far from the capital Manaus in a straight line. This city presents an equatorial climate with high temperatures and wide rainfall volumes (ALEIXO and SILVA NETO, 2015a; CORDEIRO and FREITAS, 2016). The highest total monthly rainfall is concentrated between January and May, a period known as the flood season in the Central Amazon, and the lowest total monthly rainfall occurs from June to December, characterizing the dry season (ebbing) (ALEIXO and SILVA NETO, 2015b).
The analysis in a local perspective reveals significant changes in the landscape near the urban area and in road margins passing through Tefé.
A temporal analysis using images from different periods, within about two decades, allowed observing that, by means of NDVI, the analyzed cities of the Middle Solimões region presented a representative loss of areas of dense and healthy vegetation. In this sense, in the vicinities of urban areas of the analyzed cities, forested areas have been replaced with areas of deforestation, burning for agricultural purposes, logging, and livestock (Figures 3 and 4).
The main change in NDVI was an increase, in 2011, of areas with values between 0.6 and 0.4. In 1991, on the other hand, these areas had values between 0.8 and 0.6. These areas of healthy vegetation in 1991 were replaced with areas of secondary vegetation, deforestation, urban sprawl, and agriculture.
The replacement of forest areas in the last 20 years is associated with a harmful process of deforestation, where the burning process is carried out as it is a method that requires little financial or technical resources but generates unrecoverable impacts on biodiversity and natural attributes of these areas.   (2016) In addition, the expansion of urban areas resulting from the increased population that has come to live in the city over the years contributes to this landscape change. According to IBGE (2010), urban population surpassed that living in the rural areas of Tefé from 1980 (Graphic 1). Graphic 1. Urban and rural population of the municipality .
Source: Pessoa (2004), IBGE (2010) These land use changes may have direct effects on the microclimate of Tefé, such as an increase in temperature associated with changes in soil use and hence an arboreal vegetation reduction. Thus, regional urban sprawl has also demonstrated wide changes in physical-natural attributes, such as vegetation, watercourses, and soil, and even in the local climate. Table 1 shows that the average minimum temperature over the 20-year period was 22.7 °C whereas the average maximum temperature was   (2016) The analysis of the average monthly temperature showed that over the years there was a tendency of increasing the number of warm and very warm months, especially from October to April. From 1993 to 2002, we observed 12 months with values tending to moderately warm and 2 months with values tending to warm.
However, in the period from 2003 to 2013, we observed 40 months with the highest minimum temperature (moderately warm) and 11 months with a temperature tending to warm (Table 2). This increased average minimum temperature over the months, especially in the rainiest months, is in accordance with Aleixo and Silva Neto (2014) (2016) The average monthly maximum temperature also showed a tendency to increase the values considered as moderately warm and warm, mainly from August to November. Thus, the less rainy months presented a tendency to increase the maximum temperatures, with a predominance of warm months from 2009 (Table   3).
From 1993 to 2002, 21 months presented a tendency of being moderately warm and 1 month warm, while from 2003 to 2012, 34 months tended to be moderately warm and 7 months warm.
According to Aleixo and Silva Neto  (2014, p. 8), "the increased temperature in the city may be related to land use changes and occupation around the meteorological station, which has in the surroundings a greater soil waterproofing, with the constructions of villages of military dwellings, altering the surface-atmosphere energy balance." In addition, the months in which minimum and maximum temperatures increased over the years are different. Minimum temperatures increased in the rainy months (flooding) and maximum temperatures in the less rainy months (ebbing).

Analysis of the average minimum temperature in Tefé, Amazonas state, Brazil
The statistical analysis of the minimum temperature data showed, in general, a small variation, i.e. the data were less dispersed from the mean of the series and had values of coefficient of variation (CV) between 0.01 (1995,1998,2009) and 0.06 (2012) ( Table 4).
The average minimum temperature presented the lowest value (22 °C) in 2000 and the highest value (23.8 °C) in 2010.
The standard deviation of minimum temperatures showed that February, March, June, July, August, and September had negative deviations and April, May, October, November, and December presented deviations above the average. The analysis of annual deviations associated with the average minimum temperature showed that the limit of positive deviation was exceeded only in 2010 whereas the limit value of negative deviation was exceeded in 1994 and 1995.
In general, the average minimum temperature presented annual values with significant increments from 2000 when compared to the previous decadal period, except for 2011 and 2012 (Graph 3).

Analysis of the average maximum temperature in Tefé, Amazonas state, Brazil
The statistical analysis of the maximum temperature data showed, in general, a small variation, i.e. the data were less dispersed from the mean of the series and had values of coefficient of variation (CV) between 0.00 (1995) and 0.04 (2001) ( The standard deviation of maximum temperatures showed negative deviations from January to July whereas deviations above the average were observed from August to November. In Graph 4, the analysis of the annual deviations associated with the average maximum temperature showed an oscillation between positive and negative annual deviations. In general, the average maximum temperature presented annual values with significant increments from 2000 when compared to the previous decadal period.  In spite of these problems in the data, maximum and minimum temperature values were analyzed by quantiles.  (Table 7).
Daily values of minimum temperature also increased and values equal to or lower than 19 °C decreased, which is an indication of episodes associated with cold spell days in Tefé.
In the decade of 1993-2002, values equal to or lower than 19 °C occurred in 32 days and, in the following decade, only in 5 days.
Values from 19 to 20 °C were observed in 31 days in the decade of 1993-2002 and in 38 days in the following decade. Moreover, daily values from 20 to 21 °C occurred in 316 days in the decade of 1993-2002 and in 143 days in the following decade, indicating a weakening and absence of cold spell events in the area in the last analyzed decade and an increase of values of minimum daily air temperature.

Final considerations
Land use change in Tefé showed that the balance of surface-atmosphere radiation has been altered by the process of appropriation of nature at a local scale.
The increase in the number of days with increased maximum and minimum temperatures were also observed in monthly and annual averages, demonstrating the importance of a territorial planning that takes into account the physical-natural conditions of the Amazonian landscape and incorporates the climatic dimension as support for planning and managing actions from a socio-environmental perspective.
This study also provides subsidies regarding climate changes at a meso-and macroscale over the years. However, the scarcity of observational data did not allow the analysis of a broader temporal and spatial data series.