Land use/land cover assessment as related to soil and irrigation water salinity over an oasis in arid environment

2.1 Study area

The study area is located in the eastern province of Saudi Arabia and lies between the latitude 25.30° N to 25.63° N and longitude 49.52° E to 49.77° E. It covers an area of 22,000 hectares (ha) approximately (Figure 1), with an average altitude of 149 m above the sea level [18]. The study area consists of Al-AhsaOasis, which is one of the main agricultural centers in Saudi Arabia and dominated by date palm plantation [19]. The study area lies about 300 km from the capital Riyadh and 70 km west of the Arabian Gulf. The climate of the area is characterized by six hot, dry months in the summer and relatively six months cold and wet in winter. The air temperature may exceed 45°C during the summer, while it might reach 5°C in winter. The rainfall occurs mainly during the winter, with an annual average of 50 mm [20]. The pedogenesis and characteristics of soils in the Al-Ahsa region determined by the extremely hot, dry climate, which causes soil water evaporation, result in accumulation of the dissolved salts in upper soil layers and create a crust on the soil surface. Therefore, most of the cultivated lands in Al-Ahsa Oasis are covered with saline soils [9]. The dominated soils of the study area are sandy to sandy loam soil that covers some taxonomic classes such as Gypsiorthids, Haplaquepts, Calciorthids, Torripsamments, Salorthids, and Torriorthents [11]. The primary water source of the oasis is the Neogene groundwater aquifer, and this groundwater is used mainly for irrigation and domestic purposes [21].

Figure 1

Location of the study area.

2.2 Satellite data

Earth observation satellite data are acquired from the Landsat series covering the study area during the last three decades. The data are retrieved from the United States Geological Survey website (https://earthexplorer.usgs.gov/). The used imagery was for the years 1986, 1998, 2007, and 2016 (Table 1). The data files downloaded in Tagged Image File (TIF) format and all obtained images had less than 10% cloud cover. The satellite images were acquired in the same season to minimize the influence of seasonal variations on the results.

The Landsat files were imported from TIF into Raster format using import tool of the ERDAS IMAGINE 9.2 software. The same software was used to perform a simple haze correction to give the images more contrast. All images had the same projection system; therefore, an image-to-image registration was made using the most recent image of the year 2016. The recertified images resampled to a 30 × 30 m ground resolution.

Due to an instrument malfunction onboard for Landsat-7 in May 2003, there were some gaps in images downloaded after this date. Thus, all images acquired after May 2003 needed further processing for gap filling. Gap filling processing was done for the image of 2007 using the IDL 7.0 software (https://glovis.usgs.gov/). For that three additional images were needed: (1) the image to be filled in (anchor), (2) one image acquired before May 2003, and (3) one image acquired after the anchor. For the gap filling of the 2007 image, the following images were used: (1) Landsat-7 SLC-off acquired on 11 November 2007 (Table 1) as the anchor, (2) Landsat-7 SLC acquired on 31 December 2002 as the filling scenes 1, and (3) L7-SLC-off acquired on 13 December 2007 as the filling scene 2.

2.3 Image classification and accuracy assessment

A classification scheme was developed during a field survey conducted between October and November 2016 to define the LULC classes in the study area (Table 2). A total of 220 training areas were selected based on image interpretation keys during field investigations. A supervised maximum likelihood classification (MLC) method was employed to classify the individual images independently. Based on the study objectives, the supervised classification applied in this study does not imply a comparison between different classifiers. Therefore, the MLC was adopted to be the only classification method for this study. MLC has been widely used for image classification [6,22,23].

The accuracy assessment for the image of the year 2016 was made by using 82 points out of the total training points collected during the ground truth (i.e., field reference data). On-screen visual evaluation supported by historical information obtained from the local inhabitant was used to measure the accuracy assessment of the years 1986, 1998, and 2007. Google Earth images were also used as an additional validation method for historical images. Consequently, error matrices based on the stratified random sampling of reference field data were used to assess the classification accuracy [24]. Hence, the overall accuracy, user’s and producer’s accuracies, and the Kappa coefficient were used to validate the LULC classification [25,26].

2.4 Soil and water sampling

A total of 60 and 63 random samples of soil and irrigation water were collected, respectively, from Al-Ahsa Oasis during December 2016.

Soil samples were collected from three different LULC types, namely, date palm, cropland, and bare land (Table 2). A composite soil sample was collected from 20 sites for each of the LULC types at a depth of 0–30 cm (Figure 2a). Soil salinity (EC_e) was determined in the laboratory by measuring the EC from the soil saturation extracts based on the study by Klute [27].

Figure 2

Location of sampling sites: (a) soil samples and (b) irrigation water samples.

The groundwater samples were collected from bore wells located in agricultural farms within the oasis (Figure 2b). The irrigation water salinity (Ec_iw) was measured in situ using a portable EC-meter (470 Jenway Manufacturer) [28].

The coordinates of the soil and water samples were recorded using a high accuracy differential GPS instrument (Trimble ProXR GPS). The Inverse Distance Weighted (IDW) tool of Geostatistical Analyst in the ArcGIS 10.2 software was used to perform data interpolation for EC_e and EC_iw. The IDW interpolation explicitly implements the assumption that things that are close to one another are more alike than those that are farther apart. To predict a value for any unmeasured location, IDW uses the measured values surrounding the prediction location. The measured values closest to the prediction location have more influence on the predicted value than those farther away. IDW assumes that each measured point has a local influence that diminishes with distance [29]. The method properties and weights assigned to EC_e data points are shown in Figure 3. The interpolation output of the ECe was classified based on Food and Agriculture Organization (FAO) [30]. However, the spatial distribution of the EC_iw is classified according to the FAO guidelines for irrigation water quality [30].

Figure 3

The weight window contains the list of weights assigned to each data point of the soil samples that are used to generate a predicted value at the location marked by the crosshair.

2.5 Statistical analysis

The change in LULC throughout the study period was quantified using the descriptive statistics of the Microsoft Excel 2010 software [31]. The second-order polynomial regression was used to describe the trends of LULC changes over time. Using the Statistica (StatSoft) software package [32], the Whisker Plot Box of the factor analysis was used to measure the ECe for different LULC types.

3.1 LULC mapping and accuracy assessment

Based on the produced LULC maps (Figure 4), it was found that date palm and bare land were the dominant LULC types during 1986–2016. Computed percentages of LULC classes showed that in the year 1986, date palm, cropland, and bare land had occupied 56%, 10.5%, and 32.5% of the study area, respectively, whereas urban land occupied only 0.7% of the study area (Figure 5). However, the significant spatial expansion in bare land and the large decrease in date palm were observed in the LULC map of the year 1998. Consequently, the area of bare land increased to 53%, whereas the date palm and cropland area was reduced to 39.5% and 6.1%, respectively. A slight increase of 1.2% in the spatial extent of urban land was also observed. Spatial analysis of the year 2007 LULC map revealed that there is a slight increment of urban land, which occupied 1.5% of the study area, whereas cultivated area decreased to 5% and no change was observed for date palm and bare land. Nevertheless, a considerable amount of spatial expansion of the urban land and a slight decrease in the date palm were observed in the year 2016 LULC map. The area of urban land and cropland increased to 4.1% and 6.4%, respectively, whereas the date palm decreased to 36.8%.

Figure 4

LULC maps of the study area during 1986–2016.

Figure 5

Changes in LULC areas (%) for the period 1986–2016.

The general patterns of the LULC maps identified for the study area showed a variable rate of change among different LULC classes. The significant decrease in date palm areas indicated the expansion of the bare land due to desertification. Aldakheel and Al-Hussaini [33] reported that desertification widely documented as one of the major drivers of LULC changes in Al-Ahsa Oasis over the period 1987–1998. Also, in the eastern province of Saudi Arabia, the size of the dune fields was observed to be doubled in 15 months [34]. Although date palm trees are the most dominated vegetation cover in the study area, the rate of its change indicated more capability of declining during the coming years. Based on the FAOSTAT data [35], the date palm areas in Saudi Arabia were reduced from 1,56,901 ha in 2013 to 1,08,133 ha in 2017, with an annual decline rate of 14%. The profits of date palm production become low due to urbanization and development in the study area. As a result, quite a number of date palm farms shifted to residential areas. Also, the cost of constructing and operating new wells at the date palm farms increased due to the high price of electricity and labor wigs.

The decline of the cropland shown during the first and second parts of the study period was mainly attributed to the aridity conditions and the high cost of agricultural inputs due to the increase of salinity in the soil [9]. However, the increase of cropland during the end of the study period is most likely due to the expansion of rice and vegetable farms in the southern part of the study area. These farms occurred as a result of urban agricultural activities, which are used for providing the local markets in the city centers with some products.

The expansion trend of the urban land was at the expense of date palm. Urban lands were increased significantly during the study period as a result of population increase and the extensive development of the study area. In Dammam city located in the same region of the study area, built-up lands found to be increased by 29% between 1990 and 2014 [36]. The rapid expansion of the urban lands was mainly due to the socioeconomic processes, which were closely associated with the agricultural activities in the study area. Urban agriculture can be broadly defined as the production, processing, and distribution of foodstuffs from crop and animal production within and around urban areas [37].

During the first stage of the study period, bare lands showed a significant increase in taking parts of the date palm and cropland areas. However, during the second and last stages of the study period, bare lands keep no change due to the difficulty associated with the utilization of lands resulted from desertification.

The overall LULC classification accuracy levels for four dates of Landsat satellite images ranged from 84% to 91%, with Kappa indices of agreement ranging from 78% to 87% (Table 3), which is satisfactory for the study area based on the validation measures applied to the classified LULC maps.

The achieved classification accuracy is satisfactory for the study area considering the multitemporal analysis of Landsat data and the visual interpretation adapted to image classification.

Also, the data in Table 3 showed that the producer’s and user’s accuracies for crop and urban lands were lower compared with the other LULC types. This is mainly due to the misclassification of some cropland pixels that have been classified as date palm, while the urban land ones were classified as bare lands.

3.2 Soil salinity measurements

The spatial distribution of soil salinity along Al-Ahsa Oasis is shown in Figure 6. Based on the FAO soil salinity classification, it can be seen that the EC_e values ranged from nonsaline (0–2 dS/m) to very strong saline (>16 dS/m). However, the strong saline (8–16 dS/m) is the most dominated class along the oasis. EC_e was higher in date palm with an average value of 18.97 dS/m compared to cropland and bare land that show mean values of 10.25 and 5.20 dS/m, respectively (Figure 7).

Figure 6

Spatial distribution of soil salinity (EC_e) over Al-Ahsa Oasis.

Figure 7

Soil salinity (EC_e) for the different LULC types in Al-Ahsa Oasis. DP = date palm; CL = crop land; BL = bare land.

Soil salinity is likely to be higher in cultivated lands (i.e., date palm and croplands); this might be due to the accumulation of salts as a result of saline water flow from the water table and also due to using saline irrigation water. Over-irrigation in the cultivated areas can result in the rising of the water table and hence accumulation of salinization at the root zone. Therefore, salinization problems can be more severe when the salinity of groundwater is high, as is usually the case in arid regions [30]. High level of salinity might restrict the growth of some crops in the area. However, data palm considered as a salt tolerance crop [38], and this indicates its domination within the study area compared with the other field crops.

Moderate soil salinity level is found to be associated with the bare land since most of the bare lands are covered by sandy soils and sand dunes, which are in dynamics due to desertification. Nevertheless, some parts of the bare lands showed extreme salinity level and these types of lands covered by inland saline flat areas are known locally as “Sabkha”. Rain and the rising groundwater can flood the Sabkha lands during the winter season and then turns dry during the summertime, these lands are converted to salt soils that are covered with thin crust surfaces of salts [10].

3.3 Irrigation water salinity measurements

The classified range of water salinity across Al-Ahsa Oasis is displayed in Figure 8. The EC_iw have values of nonsaline (<0.7 dS/m), slight to moderate saline (0.7–3.0 dS/m), and severe saline (>3.0 dS/m) according to the FAO guidelines for irrigation water. The spatial distribution of the EC_iw over the oasis indicated that 94% of irrigation water ranged between moderate and severe saline, while only 6% was under none to moderate saline conditions (Figure 9).

Figure 8

Spatial distribution of irrigation water salinity (EC_iw) over Al-Ahsa Oasis.

Figure 9

Irrigation water salinity (EC_iw) frequency across Al-Ahsa Oasis.

Ayers and Westcot [39] reported that the critical EC_iw level that causes severe salinity problems for most vegetable and field crops is 3 dS/m. Therefore, most of the date palm and cropland areas are facing a water salinity problem, which might lead to a decrease in the quality and quantity of agricultural productions [40]. However, some farmers use desalination plants to overcome the irrigation water salinity [41]; nevertheless, this process is of high cost and not sustainable for long-run crop production.

Moreover, Minhas [42] indicated that crop type, soil characteristics, irrigation methods, cultural practices, and climatic conditions should be studied when the saline groundwater is considered for irrigation purposes. The spatial variability in irrigation water salinity levels was observed in Al-Ahsa Oasis, which subjected the irrigation groundwater in the oasis to a process of quality degradation [43]. Along with different aquifers, in Saudi Arabia, the water salinity was found in range 1.56 and 8.24 dS/m, and accordingly, a leaching requirement of 15% to 20% above crop water requirements is suggested for safe use of this water [44]. However, Zaidi et al. [45] argued that the large variation in groundwater salinity (i.e., 1.2–3.4 dS/m) in the northwestern part of Saudi Arabia is probably due to the geochemical evolution of water in an evaporation-dominant environment.