Land cover and land use changes of native forests categories: the case of the Atencio District, Argentina, in the period from 1984 to 2013

Aim of study: The aim of this work was to assess land use and land cover change in Distrito Atencio, near Feliciano, in the province of Entre Ríos, Argentina, from 1984 to 2013 and to make a projection of possible changes in the native forests of Espinal Phytogeographic Region. Area of study: Espinal Ecorregion, Atencio (Argentina) Material and Methods: Ten LANDSAT 5-TM and LANDSAT 8-OLI satellite images were used and two classes were distinguished by vectorization: NF-Native Forests and OL-Other Lands. Besides, four categories were defined by supervised classification: CNF-Closed Native Forest, ONF-Open Native Forest, RF-Riverside Forest with Shrub Jungle, and ONG-Open Native Grassland or Savanna. Main results: It was estimated NF as being 76,619 ha and 59,994 ha for years 1984 and 2013 respectively, which represented a 21.69% reduction over a period of 30 years. In 1984, 32.93% of the district surface was occup∫ied by CNF; the same proportion was covered by ONF; and ONG followed with 16.361 ha (20.66%). In 2013, contrarily, CNF was reduced by 13.67%. Land use change was noted to be directed towards the central eastern region. By 2015, OL are expected to keep 88% of their surface, while ONG areas might suffer a 28.80% and 41.25% reduction for years 2025 and 2050 respectively. Research highlights: District’s areas with higher incidence on changes due to ecosystem fragmentation processes were identified. It is hoped that this study may contribute to settling native forests protection and recovery areas.


Introduction
Cultivated lands have globally experienced an increase of 12 million km 2 in the last 300 years at the expense of reducing not only forest areas but also grasslands and wetlands (Richards, 1990;Myers & Turner, 1991). The surface occupied by native forests currently reaches approximately 4,000 million hectares (FAO, 2007), representing 30% of the Earth's surface. By using satellite images at a spatial resolution of 30 m, recent studies by Hansen et al. (2013) quantified that forest loss was 2.3 millon km 2 and gain was 0.8 millon km 2 over the period 2000-2012. In such a world context, South America suffered the greatest net forest loss estimated in 4.3 million hectares between 2000(FAO, 2007. Argentina is not out of the global deforestation phenomenon. It has been estimated that it lost two thirds of its forest patrimony due to agricultural expansion. In 1914, the total surface of native forests was 106 million hectares, which represents 38% of the national territory (FAO, 2001); while by 1956 the forest surface scarcely reached 60 million hectares (Cozzo, 1979). Nevertheless, the native forest surface represented 12.1% of the total land surface of the country in 2005 (FAO, 2009). 2 to deforestation. The Province of Entre Ríos is strategically located in relation to various local and international markets such as Uruguay and Brazil. At the beginning of the 20 th century, the province of Entre Ríos had 2.5 million hectares of native shrublands. It was recently estimated that the forest surface in seven departments was 1,565,302 ha (Sabattini et al., 2009 a,b,c,d;2010a) and that the Feliciano Department contributes with 15.89% of the total surface of native forests in the province.
The object of this work is to evaluate land use and land cover change in Atencio District in the geographical Department of Feliciano, province of Entre Ríos, Argentina, from 1984 to 2013 and make a projection of possible changes in the native forests of Espinal Phytogeographic Region.

Materials and Methods
Atencio District is located in the northcentral region of the province of Entre Ríos and in the south area of Feliciano Departament (Figure 1). The studied area is 79,200 ha and is located at Latitude 30°30'-30°42' S and Longitude 58°36'-59°00' W.
According to Rojas & Saluso (1987), the northern region of Entre Ríos corresponds to the humid temperate climate of plains. The mean annual temperature is Land use and land cover changes are the most salient signature of human occupation of Earth's ecosystems (Weng & Wei, 2003). Land cover is one of the main biophysical attributes that affect ecosystem functioning (Brown et al., 2000); while land use is influenced by economic, cultural, political, and historical factors. In Argentina, the "agriculturation" phenomenon of extensive production systems is a characteristic and frequent case of land use change (Paruelo et al., 2004) that influences significantly on ecosystem goods and services supply.
Studies related to multitemporal analysis of land cover are of highest significance (Conese & Maselli, 1991) and they have been developed more successfully in detailed studies on land use (Jensen & Toll, 1982). On the other hand, Geographic Information Systems (GIS) allow evaluating land cover changes (Treviño, 2002) and also facilitate spatial data management (Qi & Wu, 1996;De Koning et al., 1999). Since the 1990's, researchers have been using Markov models to understand the dynamic characteristics of land use (Meyer & Turner, 1991;Mertens & Lambin, 2000;Flamenco-Sandoval et al., 2007).
Antecedents in the forested region of Argentina (Britos & Barchuk, 2008;Fernández et al., 2012;Navarro Rau et al., 2012) and in the province of Entre Rios (Sabattini & Sabattini, 2012;Brizuela et al., 2013) indicate severe changes in land cover and land use due  patterns for each class, land regularity and spectral signatures uniformity; and it was confirmed by analyzing geo-referenced censuses in field (Sabattini et al., 2009c;2013a). Photographs of fields and interviews with local producers were also taken into account to confirm the aforementioned situation. Subsequently, images of native forest surface taken in 1984 and 2013 were used to perform maximum likelihood supervised classification with bands 5, 4, and 3, training places being previously created for each class. Spectral bands TM3, TM4, and TM5 correspond to red, near-infrared, and mid-infrared wavelengths respectively, which are combined in a FCC-543. This is considered the most suitable combination for distinguishing vegetation and forest types of different forest regions (Horler & Ahern, 1986;Lencinas & Farías, 2005). Four Native Forest categories were distinguished: CNF-Closed Native Forest (presenting shrub and tree stratum covering more than 50% of land surface), ONF-Open native Forest (presenting tree and shrub stratum covering less than 50% of land surface), RF-Riverside Forest with Shrub Jungle (100% covered by trees and multi-stratified system), and ONG-Open Native Grassland -Savanna (ecotonal areas between grassland and forest biomas). These Native Forest categories have been described by Sabattini et al. (1999;2009 a,b,c,d;2010b), considering height of tree stratum (high forest: more than 6 m high; low forest: less than 6m high) and covering of shrub stratum (open and closed) as structural features. These characteristics correspond to Categories Forests and Other Forrest Lands areas according to FRA, 2010. A contingency matrix (Chuvieco, 1996), the global reliability index (0-100%) and Kappa index were assessed in each classification.
Native Forests classes and categories were identified by gathering 297 instances of information in surveyed works by Sabattini et al. (2009cSabattini et al. ( , 2013a, referenced with GPS and in which sample points were performed with preferential method (Matteuci & Colma, 1982). In each sample point, a form was filled for native forest characterization, specifying the most conspicuous tree (native and exotic) and shrub species characterizing plant succession state, native forest type, and degradation as follows: overgrazing, thinning, naked soil, fire and land clearing (Sabattini et al., 2010c).
Changes occured in land cover classes from 1984 and 2013 ware analyzed by using the Change Analysis module. Maps were obtained to show the changes ocurred in every class and to depict losses, gains, and remains for each class. Predictive trend about the possible land cover changes among classes and categories of native forests was carried out by using the Change Prediction module for years 2025 and 2050 (Eastman, 2009). 18.9°C, while average maximum temperature is 24.8 °C. In winter seasons, average minimum temperature is 12.0 °C (Plan Mapa de Suelos [Soil Mapping Plan], 1986). Mean annual rainfall is 1,300 mm mainly concentrated between the months of October and March.
Native forests of Atencio District are quite heterogeneous in appearance and in the arboreal stratum structure and shrub stratum develoment, depending on soil diversity and the profuse hydrographic network running across the region. In general, the current state of the native forest is characterized by the arrangement of plant succession where it is possible to identify: virgin or pristine (stable) forests at the final stage of succession, successional forests that correspond to intermediate stages in which diversity is improved but the native forest is not stabilized, and renewable shrublands at the beginning of the succession after land clearing where Acacia caven usually predominates (Sabattini et al., 2013b).
The area under study (Cabrera, 1976) corresponds to the Espinal Phytogeographic Province, which is characterized by xerophilous forests dominated by Prosopis nigra, Prosopis affinis, and Acacia caven. The so-called "wetlands with shrublands" and "shrublands" are also found, which correspond to the Paranaense Phytogeographic Province. According to Sabattini et al. (1999), the latter denomination corresponds to those native forests that, judging by their conjunction and transition, are located in proximity to rivers and streams and include elements of both Espinal and Paranaense Phytogeographic Provinces.
Atmospheric calibration was performed by using the COST model (Chavez, 1996). Georeferencing was based on a pattern satellite image by applying 50 to 75 sample points to obtain an average quadratic error that is smaller than pixel size (Chuvieco, 1996).
A manual vectorization was carried out, distinguishing classes NF-Native Forests (natural environments defined by their characteristic arboreal structure) and OL-Other Lands (farming and ranching areas, irrigation dams, urban zones, roads and routes, and rural settlements), thus creating a thematic layer with the particular attributes for each of the 10 selected images. Vectorization was based on the interpretation of visual 4 (Table 1). Open Natural Grassland-Savanna category follows with 16,361 ha (20.66%) and is mostly located in the northcentral region of the district (Figure 3A). Riverside Forests occupied 11.53% of the department's surface (9,134 ha); and Other Lands covered the smaller proportion (3.28%). On the contrary, when compared to 1984, Closed Native Forests had a 13.67% reduction in 2013, representing 22,514 ha (28.43% of the district's surface). Open Native Forests represent 25.36% (20,080 ha), recording a 19.75% loss in relation to the initial cover (Table 1). It is important to mention the 22.22% increase of Riverside Forests in 2013 when compared to year 1984, representing 11,164 ha (14.10% of the total district's surface in 2013).

Changes of Native Forests and Other Lands
According to the analysis of LANDSAT images, 76,619 ha of Native Forests were estimated for Atencio District in 1984, which represents 96.74% of the district's total surface. The remaining areas corresponded to Other Lands. There is no record of areas with dams from 1984 to 1993 but they are detected since 1996. Figure 2 shows the surface occupied by Native Forests and Other Lands in Atencio District for each date of the study. In 2013, the proportion of Native Forest represented 75.75% (59,994 ha) of the total surface. It is to be noted that in the last 30 years native forest areas decreased to 16,624 ha, which represents a 21.69% reduction in relation to this class. The relative weight of Other Lands is due to the increase in the areas devoted to dams for irrigation, where there was a significant increase between 1996 and 1999: from 92 ha to 2,038 ha respectively. Later on, the increase was significant with values close to 2,709 ha in 2013, which represents 14% of Other Lands surface.

Occupation of Native Forest categories
Native Forest categories were classified in the 1984 and 2013 images, identifying their spatial location ( Figure 3A and 3B) and surface estimation (Table 1). In 1984, 32.93% of the district's surface was occupied by Closed Native Forests, representing 26,081 ha. The same proportion was covered by Open Native Forests Other Lands category increased considerably, occupying 19,192 ha (24.23% of the district surface) in 2013 ( Figure 3B). This is equivalent to a 6.39 times increase in relation to 1984. In contrast, Open Natural Grassland-Savanna category suffered a significant decrease in occupation, reaching 6,428 ha in 2013 (8.12% of the district's surface). Figure 3B shows the increase of Other Lands areas located in the eastern zone of the district.
Satellite images classifications had good global reliability indexes for both dates: 76.4% for 1984 and 82.7% for 2013. The Kappa index is 0.654 and 0.762 for years 1984 and 2013 respectively. Both indexes were qualified as good according to Chuvieco (1996). According to the confusion matrix, confusions were observed between Open Native Forest and Open Natural Grassland-Savanna categories in 1984. In 2013, in contrast, better results were obtained in distinguishing each class (Table 2) while the aforementioned confusion remains.

5
Land cover changes of native forest categories (Atencio, Argentina)

Dynamics of Changes in Native Forest Categories
First of all, it is important to mention that over the period 1984 and 2013, approximately 31,650 ha remained unchanged, which would indicate that 40% of the district's surface remained without any changes in the ecosystem structure. The district surface conversion from Other Lands to Native Forests was 0.68% of the district surface, representing 540.12 ha. The significance conversion rate was towards Closed Native Forest category (256 ha) while 60% of the district's surface (47,520 ha) suffered changes in system composition. During the analyzed period, 35% of the surface that had changes corresponds to Other Lands class and would indicate a total change in vegetation structure and composition of natural systems (Figure 4).
The main changes occurred in Open Natural Grassland-Savanna category, which constituted a negative conversion of 5,903 ha to Other Lands. However, other Native Forest categories were suppressed as Closed Native Forest, losing 5,497 ha and 5,007 ha of Open Native Forest (Figure 4).
At the same time, there was a conversion of 5,375 ha from Open Native Forest category to Closed Native Forest, which would indicate a degradation process of these ecosystems as a result of shrub expansion in a competition for light and nutrients against herbaceous species. This process also took place in Open Natural Grassland-Savanna category but in a lower proportion (2,236 ha).
As regards Riverside Forest with Shrub Jungle category, it has suffered changes because local producers use wire fences to avoid problems of cattle handling in pasturelands near water courses. This process, thus, was evidenced in the 5,774 ha increase of this category, belonging 53.56% to Closed Native Forest followed by Open Native Forest (36.99%), and the remaing proportion to Open Natural Grassland-Savanna.

Trends of change in land cover and use
Taking into account transition matrixes for years 2025 and 2050, Other Lands areas are expected to keep 88% of their surface by 2025. Meanwhile, Open Natural Grassland-Savanna cover is expected to have a 28.80% conversion to Other Lands. Likewise, 12.69% of Closed Native Forests and 9.91% of Open Native Forests will become Other Lands. Probability that Closed and Open Native Forests will remain is 56.61% and 54.47% respectively (Table 3A). The probability that Riverside Forests will be modified is low or null and they are expected to keep 69.57% of their surface.
Permanence probability for Native Forest categories is expected to be notably reduced by year 2050 (Table  3B). It is expected that 35% of the current surface of Closed Native Forests (7,880 ha) to remain unchanged, which means a significant reduction when compared to year 2025. In addition, 25.75% of the surface in this category will be converted to Other Lands, while 15.11% would turn into Riverside Forests with Shrub Jungle due to weed expansion by smaller shrub and tree species. On the contrary, Other Lands class will make new contributions to Open Natural Grassland-Savanna category, since 41.25% of the surface in this category is likely to be converted to areas for agricultural production or forage taking into account that they correspond to areas with high agroecological potential.
A significant increase in Other Lands areas (29,070 ha) would be observed by year 2050 as a result of a decrease in Open Natural Grassland-Savanna (74%), By carrying out isolated analyses, it was shown observed that Other Lands class lost 539 ha in the period between 1984 and 2013, which were converted to other areas while 2,052 ha remained in the category. Gain areas of this class represent 17,130 ha ( Figure 5A). On the contrary, by analyzing the Closed Native Forest category ( Figure 5C), a 14,558 ha loss is noted, while gains were 10,962 ha. It is also important to mention that 11,540 ha of this category remained unchanged. The most relevant losses are identified in the central eastern sector and are associated to conversion to Other Lands. Changes behaviour for Open Native Forest category is similar to that for Closed Native Forest category as regards losses and persistence of this cover (14,922 ha and 10,093 ha respectively). Contrarily, gains of this ecosystem were lower, recording 9,986 ha over the period under study. Figure 5B shows that the area in which these ecosystems remained is the southeastern region, which borders Provincial Route No. 28.
Open Natural Grassland-Savanna areas presented the highest losses over the analyzed period: approximately 13,243 ha in the central western region of the district ( Figure 5D). Threfore, only 3,114 ha remained and 3,307 ha were gained in the central northern region of the district. In Riverside Forest with Shrub Jungle category, a 4,280 ha reduction was observed in the borders of Feliciano Stream, while 6,156 ha were gained. In this case, it was due to fluvial action of the stream, which refers to greater energy absorbed by Band 3 and is interpreted as a swelling or downspout condition ( Figure 5E).     Open Native Forest (42%), and Closed Native Forest (22%) categories. An increase in Riverside Native Forest with Shrub Jungle category will be also observed ( Figure 6).

Discussion
Land use change phenomenon became evident in Atencio District during the period from 1984 to 2013 due to an increase in Other Lands class and consequent decrease in all Native Forest categories. The studied area is located in a region affected by intense agriculturation and sudden land use change, moving from extensive cattle ranching in natural shrublands to rice and soybean cultivation (Díaz et al., 2009). It is important to mention that over the first 10 years (1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996), the surface devoted to Other Lands kept a 206 ha/year trend, while a positve linear trend with a higher change rate is noted in the subsequent period (1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013). A global average increase of 629.8 ha/year was estimated for the period 1984-2013 ( Figure 2). In turn, the increase of damming up areas for irrigation since 1996 is a result of two necessary conditions for rice cultivation: water availability in quantity and quality together with the impervious subsuperficial horizon of soils (Muzzachiodi, 2007).
This phenomenon was also observed in other regions of Argentina Espinal Phytogeographic Region. The situation is clearly serious in the Province of Córdoba where a very strong decline was detected of Chaco Agricultural frontier expansion in the district occurred in the central eastern region, associated to proximity and accessibility to paved routes, leads to agricultural frontier expansion due to higher availability of production factors (Briassoulis, 1999); and, at the same time, is due to an edaphological aptitude with agricultural characteristics (Plan Mapa de Suelos [Soil Mapping Plan], 1986). However, in the south western and western region, nearby Provincial Route No. 20, soils with shallow superficial horizon are found with high water retention on its surface producing conditions of radicle anoxia. Few crops may adapt or tolerate such condition, but plant composition of native forest is very well adjusted, and there are species of Carex, Cyperus or Eleocharis, which have medium forage potential.
Over the period between 1999 and 2004, moreover, there was a decrease in Other Lands areas in the western region. Other Lands areas conversion to Native Forest during the period from 1999 to 2004 may be due 9 Land cover changes of native forest categories (Atencio, Argentina) tation or patch deforestation reducing forest areas in smaller fragments from the original, and finally the complete and total deforestation of the native forest. This phenomenon was observed in the Northern area of Atencio District due to the increasing rice cropping and to the construction of dams for irrigation (Diaz et al., 2009).
On the other hand, landscape fragmentation observed as an increase of Other Lands causes changes in physical flows across the landscape, such as radiation, wind, water and can have important effects on the remaining native vegetation (Saunders et al., 1991). Removing native vegetation and substituting it with cultivable species with different morphology and phenology, as it was the case in the studied area, may alter radiation balance and affect the nutrient cycling process (Parker, 1989) and it may alter the local water regime (Kapos, 1989). Besides, in fragmented landscapes it increases wind exposure leading to damages on vegetation (Lovejoy et al., 1986).
Closed Native Forests conversion to Open Native Forests may be due to the incorporation of technology used for recovering these ecosystems. Many authors, then, posed strategies for handling and restoring forests by means of mechanical and chemical practices (Cottani & Sabattini, 2006); and considering phosphated fertilization, intercropping, cattle load handling, grazing time and pressure (Casermeiro & Spahn, 1999) would allow taking these factors as grounds for analyzing the occurred changes.
Changes observed in Open Native Forest and Open Natural Grassland-Savanna categories would be due to issues related to classification processing, taking into account the extent of confusion between both categories (Table 2). Ecotonal proximity between grassland and forest biomas creates a mosaic view of areas with Open Low Native Forest, Savannas, and natutalized Grasspland, as pointed out by some authors (Sabattini et al., 2009c).

Conclusions
District's areas with higher incidence on changes due to ecosystem fragmentation processes were identified. By using this information, trends were projected for 2025 and 2050. In this way, the study would allow settling native forest protection and recovery areas so as to improve the layout and planning of strategies for conservation and use of forest areas in Atencio District. The testing of this remote sensing technique together with field support in identifying vegetation units shows possible answers to solve environmental problems at local and regional scales.
to farm neglect and, consequently, it gives rise to the secondary succession of the forest. This process is also associated to weather variability (droughts and floods) and to variations in farm products prices, among other causes. In addition, this region presents edafic limitations for farming activity, the latter being quite marginal and scarsely stable over time (Plan Mapa de Suelos [Soil Mapping Plan], 1986).
More recent studies carried out in Feliciano Department would indicate not only native forest surface by 2008 but also its conservation level (Sabattini et al., 2009c). Comparing this data with Muzzachiodi (2007) work, the trend of native forests can be outlined for that area of Entre Ríos. Therefore, native forest represented 79.27 % of the total department's surface. According to its conservation state, 79.33% of the forest surface corresponded to the Red Zone (high level of conservation) and Yellow Zone (medium level of conservation), while 20.17 % are forests with low level of conservation (Green Zone). This enables the possibility of changing land cover -land clearing-and land use -agricultural use.
Other Lands conversion to Native Forests is due to farm neglect for various reasons (biological, productive, technical, and economic). These lands start to convert to initial forest sucessions that are called 'renewable' and are associated to loss of soil productivity as a result of erosion or lack of planning e.g. crop rotation. It could be also related to a sample error (Sabattini & Sabattini, 2012).
It can be also highlighted that the area under study is located in a region where the impact of natural system conversion is lower when compared to other areas with marked 'agriculturation'. In another study carried out in Departamento La Paz (Entre Ríos), in Alcaraz 2° District, which is an area highly impacted by land clearing, native forest was estimated to represent 31.97% of the district's total surface by 2011 (Sabattini & Sabattini, 2012).
Land use change in Atencio District confirms that "the substitution of agriculture for native forests in the province had a considerable growth in the last years, thus increasing hydric erosion and loss of streams flow volume" (Casermeiro et al., 2001;Muñoz et al., 2005). In turn, the pattern of change, in areas where deforestation of Atencio District predominates, is geometric in the form of patches, corridors and islands (Mertens & Lambin, 2000), creating a fragmented environment that results in loss of biodiversity (Bustamante & Grez 1995) and soil degradation (Farina, 2006).
The resulting deforestation was produced in three stages, as it is expressed by Reboratti (2010). Degradation and loss of native forests by selective logging of valuable species was the starting point; then fragmen-