Introduction

Agriculture is one of the most vulnerable sectors to the impacts of climate change (^{FAO, 2007}). The rainfed and irrigated agriculture yet are sensitive to changes in regional weather patterns, including interannual scale. Climate predictions of twenty-first century (^{IPCC, 2007}) are not promising and rather indicate that climate change will be emphasized during the course of this century and established most extreme conditions for crops, both in terms oftemperature and humidity, so it will be necessary to generate measures of adaptation to new climatic conditions. One of the most important steps that are displayed, is the generation ofnew varieties, which are expected to be adapted to warmer and probably less humid environment (^{IPCC, 2007}; ^{Zarazúa, 2011}).

Against this background, it is likely that the formation of new gene to feed the generation of genotypes with features such populations required. In this process, the use of wild relatives of crops will be critical, since it is known that associated genes found in these plant genetic resources for higher hardiness in relation to several factors related to crop production, such as drought resistance (^{Ruiz et al., 2008}), resistance to high and low temperatures (^{López et al., 2005}; ^{Ruiz et al., 2013}), and resistance to pests and diseases (^{De la Paz et al. , 2010}), among other.

The presence and distribution of wild species is closely related to their agroclimatic requirements and how those requirements are satisfied in various environments (^{FAO, 2007}) . The presence of climate change may threaten the satisfaction of these requirements and thus for the presence and distribution of wild species (^{Jarvis et al., 2008}) .

Mexico is considered as one of the countries with the highest biodiversity (^{Mittermeier and Goettsch, 1992}), among other factors due to its great climatic diversity (García, 2004). It is currently estimated that 12% of the total world's biodiversity is found in Mexico; this biodiversity, 12% of genera and 50-60% of the plant species are endemic (CONABIO, 2006).

With the presence of climate change, climate diversity of the country may be altered, so that may change the composition ofecosystems (^{Lasco et al., 2008}; ^{Schneider et al., 2009}) and the geographical distribution of plant genetic resources. Gossypium hirsutum is considered one of these resources, as Mexico with Central America is considered the centre of origin of this species (^{Jenkins, 2003}). The importance of this resource is that currently 90% of world cotton production is derived from G. hirsutum. Probably the native inhabitants of the Yucatán Peninsula were the first to domesticate G. hirsutum (^{Brubaker and Wendel, 1994}). These early forms of semi-domesticated cotton dispersed to other Middle and into the Caribbean (^{Iqbal et al., 2001}). From here, there was a selection to a reduced seed dormancy, annual growth habit and photoperiod independent flowering, creating very similar genotypes to current cultivars (^{OGTR, 2008}).

The aim ofthis investigation was to determine the potential impact of climate change on the geographic distribution of niches G. hirsutum L. under baseline and climate change with three GCMs (MCG). ECHAM5, MIROC (Medres) and UKMO_HADCM3 under the scenario of greenhouse gasA2 greenhouse. For this, algorithms and models are used distribution niches.

One ofthese models is Maxent which is a method ofartificial intelligence that applies the principle ofmaximum entropy to calculate the most probable geographic distribution for a species. The result of the model expresses the most appropriate value ofthe potential habitat for the species as a function of environmental variables used. A high value of the distribution function at each pixel indicates that it has favourable conditions to find the species (^{Phillips et al., 2008}; ^{Morales, 2012}). This model has recently been used to determine both the distribution of species of plants, animals and insects (^{Contreras et al, 2010}; ^{González et al, 2010}; ^{Ibarra et al, 2010}; ^{Sork et al., 2010}; ^{Torres and Khayat, 2010}; ^{Paredes et al., 2011}; ^{Sobek et al, 2012}).

The comparison between the results ofthree different MCG give an idea of the trend of the distribution of the species under study.

Materials and methods

Modelling the geographic distribution

In order to determine the potential geographic distribution of G. hirsutum we used the model Maxent (Phillips et al, 2006) version 3.3.3 k, which has been tested and proven as one of the best modellers niche distribution (^{Elith et al, 2006}). This software is based on the statistical principle of maximum entropy, which calculates the probability distribution of maximum entropy (closest to the uniform distribution), from presence data (collection) and absence, adjusting species distribution to all pixels in the study area and on the principle the idea that the best explanation for an unknown phenomenon maximize uncertainty probability distribution depending on those pixels where the species has been detected (Phillips et al., 2004, 2006). The end result is an expression ofthe type:

Where= C1, C2 ... Cn, are constant; F1, F2 ... Fn are climate variables and Z is a constant which ensures that the sum of all values of P for all cells is equal to 1.

The result can be given by three types of outputs: raw, cumulative and logistics, forthis case the output oflogistic type is used, which provides results in a map of potential distribution where the maximum ambient fitness is between 0 to 1.

Results and discussion

The Table 1 shows some basic statistics of environmental variables sites of the accession that describe G. hirsutum. The environmental ranges of this species are large. This vs la points to the hardiness of this species, a feature that could be beneficial in environments where climate change will manifest more aggressively (^{FAO, 2007}).

Table 1 Basic statistics of environmental variables and ecological descriptors of G. hirsutum L. 

The maximum value, minimum value and average value of each variable expressed in Table 1 also represent descriptors ecological G. hirsutum. According to the accessions tested, the range of annual precipitation varies between 198 and 1945 mm, contrasting values (especially the minimum value) with those reported by Aragon (1995) who sets a range from 750 to 1 500 mm annually.Also, the analysis result altitudinal limits ofthe accessions studied, reported that the distribution range of G. hirsutum goes from 0 to 2 344 m, values that differ from those reported by Baradas (1994), who notes that the altitudinal range is from 0 to 600 m.

Also, the species is adapted to a wide climatic range, as might find them in places where the temperature ranges from -0.1 up to 41.2 °C.

In the Figure 1, are shown the accessions used for the modelling of G. hirsutum using the software Maxent where we can clearly see that, the species is distributed in the warm sub-humid region ofthe country, presenting some accessions in the arid region.

Figure 1 Climatic types and accesions of G. hirsutum L. 

This algorithm performs some statistical tests based on 25% of data allocated in the random test. One of the statistics is performed regarding sensitivity vs. specificity of data for model generation, receiver operating characteristic (ROC); i.e., the ratio oftrue positives against the rate offalse positives, corresponding to the probability of selecting at random a pair ofvalues, one positive (pixel with presence of the species) and negative (pixel with no species), the model awarded to the positive value higher value (^{Hanley, 1982}). This test shows that the AUC (area under the ROC curve) for the training data is 0.937 (data used for model generation), this means that the model fits almost perfectly.

It is noteworthy that, the species is adapted to arid and semi-arid regions with warm climates and semi-warm (Aragón 1995; ^{Ruiz et al., 1999}), although in the map can be seen that the accessions are mostly in the warm humid region of the country.

In the Figure 2, the resulting map between environmental analysis classified by ranks reference to the weather is displayed. The classification was made in 5 slots fitness: very high, high, medium, low and very low, corresponding to 0.20 fitness ranges from logistics performance thrown by Maxent (0-1), with areas of very low environmental aptitude whatever is between 0 to 0.20 and the high ability of those found in ranges of 0.80 to 1.

Figure 2 Environmental fitness of G. hirsutum: a) climatology reference 1960-1990 period; b) 2040-2069 climatology ECHAM5 GCM; c) 2040-2069 climatology MCG UKMO_HADCM3; d) 2040-2069 climatology MCG MIROC (Medres). 

Moreover, the mapping b, c and d of Figure 2, showing the potential geographic distribution of G. hirsutum under the 2040-2069 climate change scenario for the different MCG used in this study, which have very contrasting results with the current situation and very similar to each other (Table 2).

Table 2 Percentage of territorial occupation of G. hirsutumL., baseline for territorial occupation and potential for climate change 2040-2069 scenario with three MCGs . 

It is noteworthy that, the percentages listed in the Table 2 were subtracted in surfaces of water bodies, urban sprawl and all those types of vegetation where G. hirsutum cannot be set as coniferous forests and riparian vegetation, to mention a few. This layer ofinformation comes from the digital chart using soil and vegetation (^{INEGI, 2005}).

In general it can be seen that compared to the baseline, all models have the same trend, i.e., decreasing of the lowest category and increase in the categories oflow, medium, high and very high environmental suitability.

The model that showed a higher decrease in the surface ofthe very low category was UKMO_HADCM3 with a decrease of 21.6%, with model MIROC which showed less decline from baseline climatology in this category (16.4%). As for the other categories MIROC has more gain surface in the high category and ECHAM5 features higher gain in the middle and high categories.

Models that have lower profit percentage area occupied by category are ECHAM5 in the low category, UKMO_ HADCM3 in middle and high class and MIROC (Medres) in the high category.

In general it can be observed that all models show a trend in which the species expanded their distribution to climate change conditions, extending the States of Sonora, Sinaloa, Tamaulipas and Nuevo León.

It is quite evident that, the impact caused by climate change will benefit the distribution of G. hirsutum L., unlike many other species whose geographical distribution is affected by changes in weather patterns (^{Jarvis, 2008}; ^{Villers, 2009}; ^{Contreras et al., 2010}; ^{Sork et al, 2010}). Within the benefit climate change will bring, is the expansion ofareas with high and very high environmental suitability, which will cause the appearance ofnew areas with this level ofenvironmental fitness, particularly in the region consists of the southern and Sonora and northern Sinaloa, the region formed by the boundary area of Tamaulipas and Nuevo León and the Southern Gulf State of Tabasco (Figure 2).

Studies in the southwestern United States show that, the fields for cotton cultivation (Gossipyum sp.) will increase up to 26-36% of its current condition, if the levels of carbon dioxide continue to increase (^{Doherty, 2003}).

The plant processes that are directly affected by changes in the atmosphere (CO2 concentration) are mainly photosynthesis, photorespiration, night respiration and transpiration (^{Fitter, 1987}). The temperature increase under future climate change and elevated CO2 levels that will be beneficial to the productivity of cotton to a certain threshold, where the rate of photosynthesis is less than optimal or environmental stress is high (^{Reddy et al., 2000}).

Conclusions

For future climate scenarios, it is expected thatthe distribution ofthe species will be favoured, because climate change will increase the areas of highest environmental suitability for G. hirsutum. On this conclusion there is agreement on the three MCGs used, being the model MIROC (Medres) which reported higher environmental surface high fitness.

Therefore, it is important to identify those areas that will be enhanced or be affected and begin to implement the necessary measures to ensure the conservation ofthe species as genetic resources strategies.

As for the cotton production from G. hirsutum it is very likely that no additional time is required for the varieties already available, but it is a medium term that will have to develop strategies focused on developing varieties that efficiently go though photosynthesis at high temperatures and under low moisture availability on the soil, in order to adapt to any climate scenario.