Description of the study area

The study area is located up north in the state of Tamaulipas (Figure 1). It is worth noting the economic importance of this region, where sorghum is the main agricultural activity, and according to the Agrifood and Fisheries Information Service (SIAP, 2014), concentrated 40.3% of domestic production of sorghum and 38.2% of the generated value in 2014 (∼$ 7.5 million pesos).

Figure 1 Dynamism from salinity index (SI) in the study area by image date;A:April 17, 1995; B: May 19, 1995; C: June 6, 2000; D: November 22, 2005; and E: November 2, 2015; F image corresponds to band composition of true color. 

The origin of salinity indices are satellite imagery, Landsat 5, Thematic Mapper (TM), the oldest and Landsat 8, Operational Land Imager (OLI), the latest of Landsat missions. For its collection and time recording, the time series of data from the Landsat series is the longest of the observation missions of the land. The study period included 20 years from 1995 to 2015. It was decided to use an image taken at a regular interval every 5 years. For its optical origin, Landsat images are influenced by the presence of clouds, therefore selected images with less than 5% of cloudiness. The dates of the images were: April 17, 1995, May 19, 1995, June 6, 2000, November 22, 2005 and November 2, 2015; the first four dates are from Landsat 5 TM and the last from Landsat 8 OLI.

Landsat images have a spatial resolution of 30m, temporal resolution 16-day and coverage area of 185 km. Each image represents a set of data of ∼38 M pixels. Before calculating salinity indices, all images were treated to correct atmospheric, radiometric, topographic effects and for illuminated surface. Salinity indices were calculated based on estimated reflectance ratio. Spectral radiance is the radiant f lux emitted, reflected, transmitted or received by a surface; the spectral reflectance is the portion of incident energy that is reflected from the contact surface. It is measured as a function of wavelength (Eq. 1 and Eq. 3, respectively). Therefore reflectance is obtained in two stages: 1) conversion of DN values (digital number) to spectral radiance values; and; 2) conversion of spectral radiance values to spectral reflectance on the top of atmosphere (Top of Atmosphere).

1)

2)

3)

4)

5)

From Eq 1, gain is the slope and bias is the intercept of the equation of spectral radiance conversion/DN; DN is the digital number. From Eq 2, the estimated ref lectance at top of the atmosphere is obtained; π is Sr (white Lambertino hypothesis; Lλ is the product from Eq. 1; d² is the distance from Earth to the Sun in astronomical units; ESUNλ, is solar spectral irradiance at top of atmosphere; θ_S is the solar zenith angle. This equation is for Landsat 5 TM images. From Eq. 4, the reflectance value at top of atmosphere is obtained; Mp is the rescaling multiplicative factor and is specific for each band; Q_cal is the DN value; A_p, is the additive rescaling factor and it specific for each band. Equation 5 is a correction equation of ref lectance by the solar local angle. ρλ’ is the product from Eq. 4; θ_SE is the local elevation angle of the sun; is located at the center of the scene.

Although the result from the equations used is an estimate of the reflectance value, it must perform the calculations separately as each source sensor has its own setting parameters. Eq. 1 and 2 correspond to data structure of 8-bit from Landsat 5 TM images, while Eq. 3, 4 and 5 are suitable for data processing at 16 bits from Landsat 8 OLI.

Salinity index was obtained according to equation 6.

6)

From Eq 6, SI is the index value; blue is the blue band (0.45.0.51μm) and red (0.64 to 0.67 μm) is the red region from the electromagnetic spectrum.

As alternative to class intervals proposed by ^{Shannon (1997)}, to categorize soil salinity, the following intervals are proposed: <16 dS m-1, moderately saline; <16-20> dSm^-1 saline; > 20 dSm^-1, highly saline.

Data analysis

For statistical analysis of the data a representative sample of 1 800 ha of land, equivalent to 20 000 coordinate pairs, which are the centroid of an equal number of pixels of Landsat image were obtained. As mask analysis, was used the land use serial vector VI from the National Institute of Statistics and Geography (INEGI). Due to the amplitude in class number, the number of strata was reduced to ten: irrigated (riego), rainfed (temporal), grassland (pastizal), bare soil (s_desnudo), scrub (matorral), mesquite (mezquite), dune (duna), body water (cpo_agua), human settlements (ah) and halophyte vegetation (v_halofita).

Statistical analysis was performed using ANOVA (analysis of variance) single factor; to explore differences between groups Tukey 0.05 was used. This is a test that can be used when required to compare each group with the rest and the number of groups is high (more than 6).

Mapping soil salinity

Figure 1 shows the sequence of thematic maps showing spatial and temporal variation in the study area according to salinity index.

To highlight from Figure 1, images A and B correspond to the beginning of the study (1995) and as shown in Figure 2 (zooming), salt content in the soil decreases considerably from one date to another in agriculture land; it shows that natural vegetation keeps its salinity level. The difference between one image and another is one month and it seems that the crops have been removed from ground cover and it is evident the decline of soil salinity value; this can be distinguished by the predominance of orange tones in the image. The presence of extreme values in the image (blue tones) can be considered aberrant (implausible data) and can be attributed to the presence of clouds.

Figure 2 Close up to SI images representing the initial condition of the study area. 

Figures E and F from Figure 1 correspond to the last date from the data series (November 2015). A better definition in land use is observed, which facilitates photo interpretation analysis of the image. Table 1 shows the descriptive data of the sample data is displayed.

Table 1 Descriptive statistics of sample data. 

From Table 1, the column headings are: n is the number of coordinate pairsperstratum of land use; A 01 is the image date, April 17, 1995; A02, May 19, 1995; X02, June 6, 2000; C00, November 22, 2005; and OLI, November 02, 2015. The value presented corresponds to the estimated average value of salt content in the soil (dSm^-1). For convenience, hereinafter, the image date will be referred with these acronyms.

An important aspect to observe from the data series was aimed at identifying land use substrate where changes in salinity condition were more evident (Figure 2).

The most notable aspect from Figure 3 is the disappearance of the "moderate" and "saline" class, virtually since 1995, starting date when gradually decreased till disappearance in 2015. It is noted that the substrate is higher dynamism is agricultural independently if it is under irrigation or rainfed. Another aspect that stands out is the consistency in SI values of one image to another till the most recent (OLI). From this perspective, it is identified that in substrates without vegetation (dunas, s_desnudo, and ah), and pastizal, where variations were not so noticeable.

Figure 3 Number of pixels of the sample with changes in location of land use substrate according to the SI index value 

The trend that data suggests is an increase of salt content in the soil. This is a warning point especially for agricultural areas under irrigation and rainfed. According to ^{Fernández-Mata et al. (2014)} the increase of salts in the soil limits agricultural activity in large land extension causing a decrease in soil capacity and low crop yields; in this regard producers, scientists and decision makers should join efforts and develop research proposals projecting a program for mitigation measures to reverse this trend. The biggest concern is that the more saline the soil plants are less capable to absorb water. The reason is simple, as salinity levels in the soil are high, the water present in the roots returns to the ground and the plant does not have the capacity to absorb it; grown plants will eventually wither and die, regardless of the amount of water being supplied.

Data analysis

ANOVA test was significant (p< 0.01) for land use. Figure 4 shows the distribution of means by sampling date per land use stratum.

Figure 4 Soil salinity by land use during the study period. 

In Figure 4, can be highlighted the explosive increase in soil salinity from 2000 to 2015 in all strata. Three points to highlight from the graph are: 1) agricultural substrate (irrigation and rainfed) shows the growth rate in higher salinity; this can be interpreted as own practices applied by producers, inherent to the production system; and 2) the stability of some land uses in those despite the increase in soil salinity, remain stable; these are: scrub, grassland and mesquite. Table 2 shows the results from the homogeneous groups test Tukey (0.05).

Table 2 Tukey test for homogeneous groups on sample data. 

El valor mostrado es el promedio. Los caracteres en negritas corresponden con el valor mínimo y los subrayados en itálicas al valor máximo de SI (dSm-1). El número de grupos significativamente diferentes es el que se reporta en la fila debajo del valor de SI, deben leerse en sentido vertical, por fecha de la imagen.

From Table 2, specifically strata where higher salt content was observed are those lacking vegetation. The lowest values were found for strata "riego" (in April and May 1995) and "mesquite" (in April 1995). As mentioned above, water use and management and technological packages could be the cause explaining these values in "irrigation" class. For mesquite, it is remarkable the parallelism with the substrate "matorral" in salt content in the soil. ^{Jarrell and Virginia (1984)} reported that mesquite root systems are well adapted to grow in areas where water source is mostly underground. Also mention that mesquite is able to grow and flourish in saline soils higher than 20 dSm^-1 and its roots can take saline water of up to 28 dSm^-1. For scrub community, numerous reports generally highlight that this biotic community is tolerant to high salinity in water and saline soils. Our results seem to support these findings and suggest that these biotic communities can survive to more saline conditions than those reported by these authors. However, our study is not conclusive and demands more attention with the approach of new research questions that include georeferenced of specific areas and spatially continuous of sites with these biotic communities.