Study area

The study was carried out at the boundary of urban influence of the city of Villahermosa (92° 55’ O and 17° 59’ N), which has an area of 20,655 ha and is located in the lower part of the drainage zone of the Grijalva river. The river of the Sierra drains by the east and the carrizal River borders the north. The city has an average height of 10 m and a minimum relief dominated by low floodplain areas and some hills in the East (^{CONAGUA, 2012}).

Database development

Thematic layers of land use were created in vector format in 1984, 2000 and 2008 (1: 75,000, 1: 20,000, 1: 10,000), digitized from black and white aerial photographs using ArcGis® 10.2.2 software (^{Environmental Systems Research Institute [ESRI], 2016}). In order to correct inconsistencies in the pixel size, a transformation was made by means of cartographic restitution (RMS < 0.5, WGS 84), based on the image of 2008 and using the PCI Geomatics® V9.1 software (PCI Geomatics Enterprises, 2003). A total of seven categories of land use were established: (1) arboreal vegetation, (2) wetlands, (3) grassland, (4) wasteland, (5) industrial land, (6) roads and (7) urban land. Along with the digitization, a field study was carried out to verify the defined classes.

Land Use Change Analysis

The multitemporal analysis was performed using the Land Change Modeler for Ecological Sustainability and the CrossTab module of the IDRISI Selva® software, generating a cross tabulation matrix (^{Eastman, 2012}). The periods considered covered from 1984 to 2000 and from 2000 to 2008. The matrices obtained were validated with the Kappa statistic (K) = 0.8963 (1984- 2000) and Kappa (K) = 0.9033 (2000-2008), close to 1.0000, generating a reliable analysis of spatial dynamics. The results include the summary of the matrices showing the surface of each category in comparison with others, in terms of gains, losses and contributions among categories (Eastman, 2012).

Change rates

Land use change rates were calculated using the formula of ^{Palacio-Prieto et al.(2004)}:

Td = [(S2 / S1) (1 / n) -1] * 100

Projection of land use change (2030)

The probabilities and spatial scenarios of land use change were projected using the combined techniques of transition models: Markov chains and cellular automata. Markov chains were used to calculate the probability of change from one pixel to another and to generate a transition probability matrix and a transition area matrix (^{Eastman, 2012}; ^{Reynoso-Santos et al., 2016}). For this calculation, the land use maps of 1984 (Figure 1) and of 2008 (Figure 2) were used, and the MARKOV module of IDRISI Selva® software was used, regarding a 22 year interval (2030). The result was a transition probability matrix and a matrix of transition areas, with a collection of maps representing transition areas for the seven land use categories in 2030.

Figure 1 Land use map of the city of Villahermosa, Tabasco (1984). Scale 1:10,000. 

Figure 2 Land use map of the city of Villahermosa, Tabasco (2008). Scale 1:10,000. 

Subsequently, the map of 2008 (Figure 2), the transition matrix and the collection of maps of transition areas of 2030, generated with MARKOV, were used as variables to run the cellular automata module (CA-MARKOV) of the software IDRISI Selva® (^{Eastman, 2012}; ^{Reynoso- Santos et al., 2016}), to generate the map of 2030 (Figure 3). The suitability of the model was evaluated through a comparison of similarity between the image of 2008 and the projected map of 2030, using the VALIDATE module. The Kappa statistic indicated that Standard K = 0.9081, Kno = 0.9488 and Klocalition = 0.9630 were close to 1.0000; showing precision for the scenarios building. Both images were intersected in the Land Change Modeler and CrossTab modules to obtain the change matrices; thus, gain, loss and contribution between coverages for the map of 2008 and the scenario of 2030 were calculated.

Figure 3 Land use map of the city of Villahermosa (projection 2030). Scale 1:10,000. 

Change analysis of 1984-2000-2008

Table 1 reports the land use areas and change rates by period analyzed. Land use in 1984 was distributed as follows: 44.5 % of the land was occupied by grassland, while 28.6 % and 12.3 % corresponded to arboreal vegetation and wetlands, respectively. By contrast, the urban area occupied 10.6 % of the territory. Figure 1 shows the land use map in Villahermosa of that year, in which it is observed that most of the arboreal vegetation was close to aquatic ecosystems in non-urbanized areas.

In the first period of analysis (1984-2000), arboreal vegetation lost 2,384 ha with a high rate of change of -3.18 %; wetlands lost 76 ha with a change rate of -0.19 %. In contrast, grassland and urban area increased by 1,507 and 942 ha, with change rates of 0.95 and 2.27 %, respectively. ^{Sánchez-Munguía (2005)} found that in Tabasco, from 1950 to 2000, about 83,518 ha of wetlands had been lost at a rate of 3,341 ha∙year-1 and that in Villahermosa, the urban advance of 2,296 ha between 1990 and 2000 invaded lagoons and marshes, and freshwater wetlands and cattatil vegetation that functioned as regulating vessels, were removed.

In the period of 2000-2008, the loss of areas of arboreal vegetation (1,624 ha) and wetlands (213 ha) increased compared to the previous period, with wide change rates (-7.45 and -1.13 %, respectively), while grassland (540 ha) and urban areas (1,334 ha) continued to occupy more area, with change rates of 0.62 and 1.46 %, respectively (Table 1).

During the 24 years (1984-2008), arboreal vegetation and wetlands lost 4,008 and 289 ha, respectively. In contrast, grassland and urban areas increased by 2,047 and 2,276 ha (Table 1), respectively, being the greatest impacts of the last three decades reflected in their spatial-temporal dynamics. In the period 1984-2008, the rate of land use change of the arboreal vegetation was of -4.63 %, being greater than that recorded in the Grijalva-Usumacinta basin and country. ^{Kolb and Galicia (2012)} observed that the rate of deforestation in the Grijalva-Usumacinta basin was 0.90 % during 1993 and 2007, and reports from the FAO (^{Food and Agriculture Organization of the United Nations, 2015}) reported deforestation rates of -0.3 % in Mexico during 1990 and 2015.

The analysis of land use change shows that Villahermosa is expanding uncontrollably in the face of poor sustainable development, from being a compact city it transformed into a sectorial perimeter and then into a fragmented city, which is common in Latin American cities (^{Bähr & Borsdorf, 2005}). The city shows a growth pattern associated with industrialization, land use regulations, regional economy, population movements, demand for agricultural products and political environment providing total control to the real estate sector and socio-cultural processes (^{Kolb et al., 2013 Linard, Tatem, & Gilbert, 2013}). The territory also has the influence of the relief, the low slope and the roads that facilitate the establishment of new population centers that demand urban infrastructure (^{Gutiérrez, Condeço-Melhorado, & Martín, 2010}; ^{Kolb et al., 2013}).

Table 2 reports the losses and gains of surface of the seven categories analyzed in the periods of 1984-2000 and 2000-2008. In the period of 1984-2000, arboreal vegetation gained 0.08 % of its surface area, but it was the category that provided more areas (3.65 %) to other land uses. Grasslands and the urban area were the categories that gained more areas (3.31 and 1.44 %, respectively), although the former also lost 1.05 %, while the latter showed no loss. Of the total wetland area, 0.18 % changed to other uses, while roads gained 0.03 % of their area. Between 2000 and 2008, the pressure on wetlands intensified, as their area decreased by 0.74 %. Similarly, grassland loss (3.31 %) was higher, although 2.26 % of its area was restored. The arboreal vegetation lost 2.61 %, while other categories such as industrial land, roads and urban land recorded an overall growth equivalent to 2.27 % of the surface.

Table 3 shows the surface contributions among the seven categories in the two periods evaluated. In the first period (1984-2000), 38 ha of arboreal vegetation were transformed into wetlands and 2,011 ha into grassland. In addition, another 306 ha of arboreal vegetation contributed to the growth of urban areas, industrial areas and roads, and only 29 ha remained as wasteland. Although wetlands gained 38 ha of arboreal vegetation, they gave 113 ha to the grasslands. This last category added another 2,011 ha coming from the arboreal vegetation, but lost 528 ha because of the growth of urban areas, industrial areas and roads. The urban area required 288 ha, 528 ha and 123 ha of arboreal vegetation, grassland and wasteland, respectively. In the second period (2000-2008), arboreal vegetation continued to provide areas to grassland (1,252 ha) and urban area (424 ha). On the other hand, due to the reforestation, arboreal vegetation recovered 77 ha coming from wasteland. The negative trend of the wetlands was maintained by contributing another 212 ha to grassland. This last category expanded with 1,252 ha of arboreal vegetation, although it reduced 102 ha due to the increase of the industrial area. In these eight years, 424 ha of arboreal vegetation, 762 ha of grassland and 141 ha of wasteland were urbanized.

In this study, deforestation and drastic reduction of wetlands were primarily caused by the transition to grassland for agricultural use and secondly by urbanization. These results agree with ^{Zavala et al. (2009)}, who found that agricultural activities (60 %), especially grassland for livestock cattle, and urban areas (9.1 %) were dominant in Villahermosa. In the same study, the authors pointed out that 74 % of the landscape was transformed in the years 1984 and 2005 and that areas with arboreal vegetation and the wetlands occupied 25.3 and 5.2 % of the urban territory, respectively.

Between 1940-1996, 95 % of Tabasco forests were lost, due to the increase of areas for agricultural and livestock activities (^{Zavala & Castillo, 2007}). ^{Sánchez-Munguía (2005)} analyzed the land use in Tabasco for the period of 1950-2000 and reported that the forest area distributed in ejidos and private property was 538,861 ha in 1950; 10 years later, the area of natural vegetation decreased to 453,411 ha; and in 1970 an accelerated deforestation initiated, leaving 146,485 ha of forest, which were reduced to 71,387 ha in 1980 and 41,079 in 1990. This decrease meant that, in 40 years, forests changed from representing 21.7 % to 1.6 % of the state area. ^{Sánchez- Munguía (2005)} linked deforestation between 1950 and 1991 with the increase in the number of heads of cattle, which was the economic activity that replaced the export of banana and other agricultural products before the oil boom (^{Allub & Michel, 1979}).

The trend of land-use change agrees with other studies (^{Kolb & Galicia, 2013}; ^{Perezechchikova & Lezama, 2010}; ^{Sánchez-Munguía, 2005}; ^{Zavala & Castillo, 2007}; ^{Zavala et al., 2009}) regarding the replacement of arboreal vegetation and wetlands cover by grassland and urban use; however, different surfaces and rates of change are detected due to the periods of analysis, evaluation methods and study scales. ^{Velázquez et al. (2002)} pointed out that the approaches used for the analysis of land use change are not homogeneous and, therefore, the results of different studies are varied in mapping categories and study scales. Thus, in order to compare with greater accuracy and reliability the dynamics of the different ecosystems in Mexico, it is necessary to systematize the mechanisms of evaluation, prediction and monitoring with compatible databases in categories and study scales.

Although the land use change modeler detected the accelerated growth of grassland for agricultural use against the declining of arboreal vegetation and wetlands, ^{Sánchez-Munguía (2005)} mentioned that in the last decades, there was a notable abandonment of agricultural activity in the territory, reflected in the collapse of the slaughter of cattle in the Frigorífico of Villahermosa. In spite of the decrease of this activity, the transition from grassland to arboreal vegetation by natural regeneration was not recorded; in contrast, only small reforested areas were located. Soil erosion in the Grijalva-Usumacinta basin may explain the lack of natural restoration, as 47.64 % of the area has slopes of more than 8 degrees (^{Sánchez-Hernández, Mendoza- Palacios, De la Cruz-Reyes, Mendoza-Martínez, & Ramos-Reyes, 2013}).

In Tabasco, wetlands have deteriorated due to increased grassland, urban infrastructure and the construction of industrial areas and roads (^{Estrada, Barba, & Ramos, 2013}), despite the fact that such coverage is valued as environmental regulators, flood damping, ecosystems hosting high biodiversity and habitat for resident and migratory species (^{Henny & Meutia, 2014}; ^{Hettiarachchi et al., 2014}). With regard to the contamination of wetlands, the discharge of chemical products and the contribution of sediments derived from urban developments have led to hypertrophic conditions since the beginning of the 1990s (^{Goñi-Arévalo, Hernández- Pérez, Toledo-Gómez, & Pérez, 1991}). These conditions have increased in Villahermosa (^{Hansen, Van Afferden, & Torres-Bejarano, 2007}; ^{Sánchez et al., 2012}) and in other urban and coastal areas of the basin (^{Salcedo, Sánchez, De la Lanza, Kamplicher, & Florido, 2012}).

Projection of the probabilities of land use change (2030)

By 2030, the replacement of arboreal vegetation by grasslands recorded 0.52 probability, which was followed by 0.13 related to urban growth (Table 4). In this sense, wetlands will disappear to become grasslands with a probability of 0.11; however, grassland will in turn be replaced by the urban area with a probability index of 0.11 (Table 4). Wasteland have a probability of 0.18 of transforming into arboreal vegetation, because these are places suitable for reforestation; however, wasteland will decrease by the growth of the urban area (probability of 0.45).

It was detected that the arboreal vegetation and the wetlands will decrease 1,187 and 254 ha, respectively with the model of cellular automata. In contrast, grassland will apparently increase 1,114 ha, as it will lose 1,431 ha. The urban area will continue to accumulate area and will add 6,137 ha in 2030 (Table 5). The contributions between coverages from 2008 to 2030 (Table 6) predicted that the arboreal vegetation will lose 860 ha when replaced by grassland; moreover, 254 ha of wetlands will be transformed into grassland. Likewise, for its imminent expansion, the urban area will invade 316 ha of arboreal vegetation and 1,252 ha of wetlands.

The land-use change model, Markov chains and cellular automata accurately detected the distribution of natural coverages and artificial uses, probabilities and spatial projection of change for year 2030, providing useful information for environmental planning for the city of Villahermosa. ^{Jiménez-Moreno, González- Guillén, Escalona-Maurice, Valdez-Lazalde and Aguirre- Salado (2011)} mentioned that it is essential to use one or several models of land use change so that soil authorities and planners can understand the scope of the changes recorded and the risks involved. In addition, the changes recorded allow to identify the factors that are causing them and, therefore, are useful to follow the territorial order.