A. Study area

The municipality of Nechí has an area of 935.135 km2, and is located in Antioquia's northwest corner in the Bajo Cauca region, limiting to the north with both Córdoba and Bolívar departments, to the east with the municipality of El Bagre, and to the west and south with the municipality of Caucasia (Fig. 1). The main network is formed by the watersheds of the rivers Cauca and Nechí, characterized by their high flows and considerable length. The most notable geological process in this area is the lateral erosion by scouring, which affects the margin of the rivers mentioned above in some reaches. Nechí has a low, undulating topography between 30 and 600 m.a.s.l., a tropical climate with temperatures between 28 and 30 °C, and an average annual rainfall between 2000 and 4000 mm. Additionally, Nechí has great biodiversity and natural resources, which are threatened by an accelerated deterioration mainly due to human activities ⁷.

Fig. 1 Geographical location of Nechí. 

The main economic activities in Nechí are mining, followed by farming, fishing, agriculture, forestry, and the trade of goods. The participation by economic activities in the Bajo Cauca region, mainly Nechi, El Bagre and Zegovia, to the Departmental GPD is mining with 36.3 %, and agriculture, farming, forestry, and fishing with 1.5 % ³⁵. Mining activities, particularly of gold and silver alluvium, are carried out without the adoption of clean development practices, which negatively impacts the environment ⁷.

B. Metallurgy and exploitation of alluvial gold

The alluvial gold in Nechí is extracted from the stream shores and riverbeds using dredges, bulldozers, hydraulic monitors, and wells or "apiques" ^14-15. In the benefit stage, the extracted material is deposited in gutters where mercury is added. Leakage of some of the mercury generates losses of pure gold, and produces pollution and sedimentation in soil and water ¹⁶.

The lack of knowledge about the geological characteristics of the deposits in the exploitation, along with the improperly planned operation produce a considerable loss of mineral, as well as serious impacts on the environment, mainly in soil cover and water bodies ¹⁵. Soil deterioration starts right after defining the interest sites through brief prospection, when forests are cleared, and the machinery, especially dredges and bulldozers, remove topsoil and sterile covering layers of economic importance. These materials are removed without control in large areas, without techniques to properly dispose them, and eventually become sediment loads into rivers and nearby effluents ¹⁵.

The impacts on soil and vegetation include deforestation by destruction and loss of healthy vegetation, soil contamination by fuel mismanagement, changes in the landscape, and soil losses by disposition of sterile equipment in the vicinity of the extraction sites and gutters ¹⁷. Although measures such as reforestation programs have been carried out to remediate the impacts on the ecosystems in the municipality of Nechí, these measures have not been sufficient, and the current situation shows a loss of forests and an increase in eroded soils.

C. Landsat satellite images

The primary information used in this research comes from three Landsat 7 ETM+ images of the study area. These images cover a period of 24 years (1986, 1996 and 2010), and are available at the Glovis website of the United States Geological Survey USGS. Each Landsat image is composed of seven bands, representing portions of the spectral range. The Landsat satellite has a cycle of 16 days, which means that Colombia is covered 22 times a year on average. The three images were chosen based on their quality, that is, with a mean percentage of cloud cover of zero, and with a time span that allows the analysis of land cover changes that may be related to historical, political, social, and economic conditions in Colombia.

D. Landsat imagery processing

The methodology used in this study is shown in Figure 2. The images were processed in ArcGIS10.1, using Arc Map, Arc Toolbox, and Arc Catalog modules. Initially, bands were combined to extract and emphasize the information contained on each satellite raster image, based on spectral responses of the Earth's surface objects. A visual analysis was performed combining three information bands of each satellite image (e.g., 4-3-2, 5-4-3, 3-21, and others) to form false and true color images with the Red Green Blue-RGB channels ¹⁸. The color combinations are used to discriminate aspects represented by the image, such as geology, land use, urban morphology, surface water, eroded areas, vigorous vegetation, and transition zones, among others ¹⁹. These combinations allowed the identification of soil cover types in the study area.

Fig. 2 Methodology used in the analysis of change in soil cover in Nechí. 

Subsequently, the Normalized Difference Vegetation Index (NDVI) was calculated on each image, in order to examine the density and intensity of green vegetation growth using the spectral reflectivity of solar radiation. This index uses the contrast of two bands of a satellite image: band 3 or visible red band, and band 4 or near infrared band. Healthy vegetation absorbs most of the visible light, and reflects a large part of the near infrared light. In contrast, unhealthy or sparse vegetation reflects more visible light and less near-infrared light ²⁰. The NDVI is calculated by Equation (1) ²⁰:

where R and IR represent bands 3 and 4, respectively. The NDVI is provided as a raster image with each pixel varying between -1 and +1. Values near zero indicate lack of green vegetation, values equal to zero mean absence of vegetation, and values close to one indicate the strongest vigorous vegetation density.

We calculated the NDVI differences between 1986 and 2010 to determine vegetation changes in the study area, and considered the change to be an improvement if vegetation was stronger or greener, or a depletion if vegetation was less vigorous or dry. The difference between NDVIs was calculated by Equation (2) ²¹:

where ^_NDVI19A6 and ^_NDVI2Q\Q correspond to the NDVI from 1986 and 2010, respectively. The raster was classified and organized into classes to describe the changes in vegetation. The values of each class were defined based on the spectral response of the different types of vegetation due to chlorophyll content. A conventional way of representing the differences in vegetation with the limits of spectral response was implemented in order to show the change in vegetation ^21-22.

After considering different band combinations, we identified four soil cover types: Forest, Pasture, Bare Soil, and Water Bodies (surface). We extracted these soil cover types as vector images from the difference of NDVIs, and used them to quantify the hectares and analyze soil cover maps and directions of vegetation changes in the 1986, 1996 and 2010 images.

E. Historical analysis

We reviewed the literature on the main historical facts related to the social and political economic conditions in Colombia during the period of study, 1986-2010, associated with the periods of the "gold rush", when the international price of this commodity rose, and whole type of mining exploitation become profitable.

A. Colombia's socioeconomic and political conditions

During the 1980s, Latin America experienced the crisis of "the lost decade", characterized by declines in per capita GDP, growing inequality in income distribution, and increased poverty levels, among others. Despite the crisis, Colombia performed better than other countries due to a prudent management of external debt in previous years, in addition to good macroeconomic management ²³. Between the mid-80s and mid-90s, the country adopted reforms toward an open trade liberalization of the economy to international markets, despite the limitations in existing infrastructure. The economic growth at the time was based mostly on primary goods from agriculture, manufacturing, and largely the exploitation of oil, coal and ferronickel. From then on, mining continued growing. In Nechí, gold exploitation increased between 1985 and 1990, becoming the main economic base of the municipality ⁷. A period of recovery in 1986 was driven by a phase of higher growth in mining exports in Colombian history.

Subsequently, the growth of the mining sector was starred in the Colombian economy. During the second half of the 1990s, the mining sector grew at an average annual rate of 4.8 %, becoming fundamental for the Colombian economy, contributing to the production of goods, services, and exports, as well as generating tax revenues at national and local levels. Even though mining has a small contribution to GDP, it maintained a positive trend since the 1980s. The mining GDP, including hydrocarbons, grew from 1.5 % in 1980 to 4.8 % in 1999, with a positive rate of incremental growth ²⁴.

Between 2009 and 2010, Colombia underwent changes in its economy due to the global economic crisis. The good performance of various economic sectors in the country allowed a fast economy recovery. The most outstanding sector was the mining industry, which had a significant growth of 11.1 %, based on the increased production of oil, natural gas, and minerals, such as gold, coal, silver, and nickel ^25-26, as a response to increasing international ore prices. Figure 3 shows the mining titles in the department of Antioquia, and the relationship of some of the largest gold mining projects in 2010 ²⁷.

Fig. 3 Map of gold mining titles and megaprojects of mining companies in antioquia, colombia. Adapted from ²⁷. 

B. Soil coverage and NDVI analysis

Table 1 summarizes the areas corresponding to the four types of coverage obtained after the spatial analysis of NDVI from each studied year. An atmospheric correction was carried out due to the presence of clouds in some parts of the images from 1996 and 2010. This correction consists in subtracting the cloudy areas from the total amount of the other types of coverage. Figures 4, 5 and 6 show the land cover maps of 1986, 1996 and 2010, respectively.

Table 1 Areas of coverage in 1986, 1996 and 2010 

Fig. 4 Soil coverages extracted from the NDVI in 1986. 

Fig. 5 Soil coverages extracted from the NDVI in 1996. 

Fig. 6 Soil coverages extracted from the NDVI in 2010. 

C. NDVI differences between 1986 and 2010

We classified the changes in vegetation cover over 24 years in the studied area based on the land cover maps of 1986 and 2010 under three categories proposed by ²¹ (Fig. 4 and 6): deteriorated condition of soil cover, maintained condition, and improved condition. Table 2 summarizes the areas where these categories were identified.

*The total does not consider the areas occupied by of vegetation cover remained stable without any water bodies on 1986 and 2010. disturbance or affectation, whereas 246 km2 (29.8 %)

deteriorated with vegetation losing vigor, and only According to the vegetation changes observed in 150 km² (18.1 %) of vegetation improved (Fig. 7). Nechí during the studied period, 431 km2 (51.2 %)

Fig. 7 Direction of vegetation change from NDVI differences between 1986 and 2010. 

D. Analysis of soil cover changes

Soil cover maps show that the east side of Nechí has preserved the forest cover, since this part is within the area of jurisdiction of the Serranía de San Lucas, a forest reserve that has been recommended for protection due to high levels of endemism ^28-29. According to Figure 7, most of this area has remained unaltered, and even a small extension has improved. In contrast, deterioration is observed in the area surrounding the Nechí River due to the development of alluvial mining in the Complejo Cenagoso El Sapo ⁷. Marked changes in land cover are perceived at southwestern Nechí, characterized by a sudden increase of bare soil and pastures between 1986 and 1996. A significant increase in pastures is also noticeable by 2010. Most areas considered deteriorated are also located in this part of the municipality (Fig. 7). An increase in water bodies occurs in areas bordering the Nechí River, a zone with permanent presence of alluvial gold mining.

Over the 24-year period, there have been significant changes in land cover (Table 3). Forests lost nearly 22 % of coverage, while pasture advanced by 68 %. Surprisingly, the amount of bare soil resulted from erosion processes, loss of vegetation cover, and possibly gold mining was reduced by 30 %. According to the results, it is possible to establish a relationship between livestock and the effects of mining in Nechí. The increase in livestock exploitation in Nechí has become an important factor, not only for the economy of the municipality but also for the Bajo Cauca region; this growth can be associated with the increased pasture cover observed from imagery analysis, since this type of soil cover has an average load capacity of 1.3 animals per hectare, fostering the production, commercialization, and distribution of bovine products in areas where native pastures predominate ⁷. In the case of water bodies, there was an increase of 66 %, which may be related to the type of exploitation of gold, which consists of dredging of the ground by hydraulic monitors and the flow toward a gravity separation structure or pits made on land to remove metals. Furthermore, in the amalgamation process, mercury is added directly to the excavation pit, and the gold is recovered by the movement and transportation of the load. Water quality is affected by discharges of toxic substances, such as cyanide and heavy metals (copper, silver, nickel, lead, and zinc, among others), which are accumulated and then deposited in bottom sediments ¹⁶. In the Bajo Cauca region, reforestation projects have been developed over the studied period. As a result, they may have influenced an apparent recovery of the vegetation ³⁰. The abandonment of old gold mining quarries may also explain part of this recovery. Moreover, it is necessary to consider the climatic conditions during the studied years; "El Niño" and "La Niña" conditions occurred over the 24 years of study, with "La Niña" influencing the weather in 1996 and 2010 with conditions of weak and moderate intensity, respectively ³¹. This is a cool phase characterized by strong rain in the region. Therefore, the climate may have influenced the presence of ponds and flooded areas in the soil cover changes.

An average annual production of more than 10,000 kg of gold was registered in Antioquia between 2007 and 2011, which represents approximately 49.96 % of the average annual national production. For 2011, the production percentage of alluvial and "filon" gold in Antioquia was 73 % and 27 %, respectively ¹⁶. Table 4 shows the amounts of gold produced by different departments between 2002 and 2010.

Gold mining activities lead the mining sector in Antioquia, with their participation threefold in the department's economy increasing from 0.5 % of the departmental GDP in 2000 to 1.4 % by the end of 2012. Furthermore, this growth is also reflected in the national economy, since gold production in Antioquia increased from 0.07 % of nation's GDP in 2000 to 0.18 % in 2012 ³².

In 2011, the gold production in Antioquia increased in 258 kg compared to the previous year; the total gold production of the department was 19,156 kg, which came from the municipalities of Taraza (4,812 kg), El Bagre (2,997 kg), Cañasgordas (1,771 kg), Nechí (1,619 kg), Remedios (1,329 kg), Zaragoza (1,143 kg), Segovia (1,409 kg), Caucasia (1,232 kg), and Angostura (1,214 kg) ³³.

Table 5 shows the volume of annual gold production in kg in Nechí and Antioquia between 2001 and 2010. An increase in gold extraction is observed in 2004, 2009 and 2010. Figure 8 shows the gold annual production in Nechí, and the percentage contribution of production in Nechí to the total production of the department of Antioquia. In 2010, Nechí made the greatest gold contribution to the total gold production in Antioquia, with 12 %.

Fig. 8 Annual gold production in Nechí and its contribution to the total production of Antioquia. Adapted from ³⁴. 

Several studies ^11,35-40 have confirmed the usefulness of classifying and extracting features from remote sensing products related to soil cover. The spatial and temporal information contained on NDVI about vegetation growth represent helpful complementary data for land cover classification. Furthermore, the process of land use and soil cover change in mining regions can be evaluated and quantified by the use of remote sensing information in a time series trend analysis. Finally, historical baselines of forest cover based on management and planning territory studies are needed to analyze the changes and to assess the effectiveness of land-use policies.