Saline lakes are widespread throughout the arid and semi-arid regions of the world and have considerable ecological importance. They are also very vulnerable to climatic changes or changes in their hydrological regimes. Most saline lakes of Turkey are close to the verge of extinction due to natural and anthropogenic impacts. This study analyzes the spatial and temporal changes at a relatively pristine saline lake (Tuzla (Palas) Lake) in Kayseri, Turkey, from 1987 to 2011 using satellite imagery techniques. Landsat Thematic Mapper images acquired in 1987, 2000, 2003, and 2011 were used in the analysis. The images were geometrically corrected by registering them to ground control points. The study area on each image was classified into seven information classes, i.e., water, watery ground, dry lake, mud/vegetated flats, salt flats, shrubs/sedges, and agriculture. The accuracies of the classifications were evaluated using a standard error matrix and kappa statistics. The analysis showed that the surface area of Tuzla Lake was highly variable during the 1987–2011 period. Lake surface area was the largest in 1987 and the smallest in 2003. Analysis of the climatic conditions for 4 years showed that the surface area of the lake is highly vulnerable to changes in precipitation and air temperatures.

INTRODUCTION

Saline lakes occur in arid to semi-arid environments whenever internally drained basins are formed due to tectonic activity or dissolution processes and the floors of such basins intersect the water-table (Duffy & Al-Hassan 1988; Yechieli & Wood 2002). These lakes are generally small and shallow water bodies, and therefore are very sensitive to changes in climatic and hydrologic conditions (Yechieli & Wood 2002). Fluctuation in air temperatures, precipitation, and wind activity can cause changes in their depths, extents, and salinities by affecting water inflows and outflows. Under adverse climatic conditions, lakes may disappear entirely (Poff et al. 2002). These lakes provide critical habitats for endemic species, and breeding and migratory birds, and also they are economically important due to their utilization as salt reserves.

Monitoring saline lakes is important for evaluating the health of these fragile systems or determining the extent of degradation. There are some studies in the literature that have examined the changes in saline lakes in different regions of the world (Kasishke & Bourgeau-Chavez 1997; Simmons & Narayan 1997; Dwivedi et al. 1999; Rao et al. 1999; Kashima 2002; Jellison et al. 2004; Castenada et al. 2005; Kohfahl et al. 2008; Rodríguez et al. 2010). Most of these studies showed that saline lakes have been altered considerably. They also showed that one of the major threats for sustainability of saline lakes is deterioration of their natural hydrologic regimes either due to climatic changes or water/land use activities in their catchments (i.e., intensification of irrigated agriculture).

The studies conducted regarding saline lakes in Turkey are rather limited. Ormeci & Ekercin (2007) investigated water reserve changes at Salt Lake, Turkey, and showed that water reserves in the lake decreased significantly between 1990 and 2005 due to drought and uncontrolled water use. Ekercin & Ormeci (2010) investigated the climate change impacts on Salt Lake and its vicinity using satellite imagery techniques, and suggested that the use of surface and groundwater supplies around Salt Lake should be controlled for an effective management of water and salt resources in the region. Dadaser-Celik et al. (2008) examined the changes at the Sultan Marshes ecosystem in Turkey, which hosts two saline lakes (Yay Lake and Çöl Lake), and found that lake surface areas decreased by 93% from 1980 to 2003 and in some years (e.g., 2003) the lakes went completely dry. The causes of changes in Turkish saline lakes were similar to the causes of changes at those around the world.

The scarcity of hydrologic data about saline lakes often prevents the efficient monitoring of these ecosystems. Satellite imagery has been used in recent decades to cope with data limitations. Satellite imagery provides an efficient method of conducting large-scale studies of the Earth's surface. Much information about the Earth's surface, related to water, vegetation, rocks, minerals, and soils can be extracted from multispectral data provided by satellite images (Kienast & Boettinger 2007). Satellite observations are particularly useful to modelers because they provide repetitive broad regional views of good spatial resolution in digital format (UWPCC 2007). There has been increasing interest in characterizing and monitoring lakes and wetlands by satellite images in recent decades (Singh 1989; Haack 1996; Nakayama et al. 1997; Munyati 2000; Piwowar & Ledrew 2002; Castenada et al. 2005; Jensen 2005; Ormeci & Ekercin 2007; Dadaser-Celik et al. 2008). Some of these studies were conducted at saline lakes (Drake & Bryant 1994; Bryant 1999; Williams 2002; French et al. 2006; Rezaei & Saghafi 2006; Ormeci & Ekercin 2007; Reis & Yılmaz 2008; Thakur et al. 2012). In most of these studies, Landsat imagery has been preferred due to the repetitive coverage since the 1980s, and radiometric and spectral resolution sufficient for change detection studies (Singh 1989; Haack 1996; Nakayama et al. 1997; Munyati 2000; Piwowar & Ledrew 2002; Castenada et al. 2005; Jensen 2005; Ormeci & Ekercin 2007; Dadaser-Celik et al. 2008).

The objective of this study is to analyze the spatial and temporal changes at Tuzla (Palas) Lake in Kayseri, Turkey (Figure 1), during the 1987–2011 period. Tuzla Lake is one of the saline lakes in the semi-arid Central Anatolia region of Turkey. It is a relatively pristine system compared to other saline lakes (e.g., Salt Lake and Yay Lake) in the same region. Therefore, the response of the lake to long-term variations in climate can better be observed. Until now, only a few studies have been conducted on the lake to understand its physical and hydrological characteristics and to determine how they evolved during recent decades. A recent study by Cengiz & Dadaser-Celik (2012) examined the changes in water levels at Tuzla Lake from 1998 to 2005 and found that water levels went down at a rate of 6.9 cm yr−1 during this period. This analysis revealed that the decreases in water levels could be related to changes in hydrologic and climatic conditions. Unfortunately, the period included in the analysis was very short due to the unavailability of water level data prior to 1998 and after 2005 and it was, therefore, not possible to make a comprehensive evaluation of the causes of water level decreases.
Figure 1

Physical setting of Tuzla Lake and its location in Turkey.

Figure 1

Physical setting of Tuzla Lake and its location in Turkey.

In this study, we aim to fill this gap by analyzing four satellite images acquired in years with different climatic conditions from the 1987–2011 period. With this analysis we aim to understand how lake surface area changed over the recent decades and determine how these changes were related to changes in climatic conditions (i.e., precipitation, air temperature, evaporation). This kind of analysis can reveal the susceptibility of the Tuzla Lake ecosystem to climatic changes and also provide insights for saline lakes located in similar climatic settings.

MATERIAL AND METHODS

In this study, we used satellite imagery techniques to analyze the changes at Tuzla Lake from 1987 to 2011. Below we first provide brief information about the study area, Tuzla Lake. Then, we explain the image analysis procedures.

Study area: Tuzla (Palas) Lake

Tuzla Lake is one of the important saline lake ecosystems in the Central Anatolia region of Turkey. It is a closed-basin lake and located at 38 °02′N and 35 °49′E, about 40 km away from Kayseri city center (Figure 1).

The lake is located at the Palas Basin, which has a drainage area of about 456 km2 (DSI 1970). The Palas Basin is a very isolated, hot, and arid landscape, where climate is characterized by hot and dry summers and wet and cold winters. There is no meteorological station within the borders of the Palas Basin. The closest meteorological station is located in the city of Kayseri (38 °43′N and 35 °29′E) about 50 km away from the Palas Basin. Figure 2 shows annual average air temperature, annual precipitation, and annual pan evaporation measured at Kayseri station from 1987 to 2011. These data were obtained from the State Meteorology Service. The average annual air temperature during this period changed between 8 and 13 °C, with an average of 10.1 °C (±1.01 °C). The annual precipitation changed between 258 and 614 mm, with an average of 412 mm (±91 mm). As can be seen, precipitation was particularly variable in the region and went very low in some years. Annual pan evaporation during the 1987–2011 period changed between 885 and 1,506 mm with an average of 1,067 mm (±132 mm). Compared to precipitation, evaporation is very high. The large difference between precipitation and evaporation caused the lake to be highly susceptible to climatic variations.
Figure 2

Annual average air temperature and annual precipitation in Kayseri for the 1975–2011 period.

Figure 2

Annual average air temperature and annual precipitation in Kayseri for the 1975–2011 period.

Tuzla Lake is fed by precipitation, flows from some small streams (e.g., Degirmen stream) and small springs (Yertaspınar, Körpınar, Baspınar, and Sogukpınar), located at the south and west of the lake, overland flow, and groundwater flow (DSI 1970). As water flows from streams and springs and overland flow are very low, precipitation and groundwater flows constitute the major water inputs (Cengiz & Dadaser-Celik 2012). The lake is a closed system and water is lost only through evaporation. Since the Palas Basin is topographically low pitched, the lake surface area is highly dependent on precipitation and varies throughout the year (Somuncu 1999). The average surface area of Tuzla Lake changes from 25 to 35 km2 and its average depth is 2 m (ENCON 1999). In the summer season, due to low precipitation and high evaporation, the surface area of Tuzla Lake becomes much smaller, exposing an average of 15 cm thick salt layer in August.

Around the south and southeast parts of the lake there are fresh water and salt water wetlands and wet meadows showing high habitat diversity. It is, therefore, a unique feature in this landscape and was declared a protected natural site area of first degree in 1993. Tuzla Lake is a critical habitat for a variety of plant and animal species. The lake is located on the bird migration routes from Europe to Asia. The lake ecosystem supports about 28 species of reptiles, 232 species of plants, and 93 species of birds (ENCON 1999). Owing to its rich biodiversity, Tuzla Lake and its immediate surroundings, covering about 48 km2, has been declared a protected area by the Ministry of Forestry and Waterworks of Turkey.

Image analysis

In this study, five steps were followed to analyze changes at Tuzla Lake: (1) image acquisition, (2) preprocessing, (3) image classification, (4) accuracy assessment, and (5) change detection. All analyses were performed by ERDAS Imagine 9.2. Below, these steps are explained in detail.

Image acquisition

Acquiring images from the same sensor is important for obtaining consistent spatial, spectral, and radiometric resolution in change detection studies (Swain & Davis 1978; Jensen 2005). In this study, we used Landsat Thematic Mapper (TM) images. Landsat TM images are available from 1984 to date and can therefore be used for long-term monitoring studies. Landsat TM images also have a spatial resolution of 30 × 30 m and provide data in seven multispectral bands, which are sufficient for monitoring changes at Tuzla Lake. Many previous studies used Landsat images for estimating changes in lakes of similar size to Tuzla Lake (Bouvet et al. 2003; Li et al. 2013; Pope & Rees 2014).

Another important issue in change detection studies is the comparison of images from the same time of year. In this study, we selected imagery from the late spring/summer season, as high-quality (i.e., cloud-free) images were available for this time of year. Our final set included four images that were acquired on 4 July 1987, 21 June 2000, 16 July 2003, and 6 July 2011 (Figure 3). The study area is entirely contained within Landsat path 175, row 33.
Figure 3

Landsat TM images of the study area used in the analysis.

Figure 3

Landsat TM images of the study area used in the analysis.

Image preprocessing

All images were rectified to the Universal Transverse Mercator (UTM), Zone North 36 projection (WGS84 spheroid) using at least 43 well-distributed ground control points that can be easily identified on the images. A nearest neighbor algorithm was used for resampling of all images to retain original pixel values. The root mean square errors for all images were less than 0.25 pixel or 7.5 m. Radiometric correction was not applied as post-classification change detection techniques were used for identification of changes between different dates (Jensen 2005). The study area, which covers the Tuzla Lake protection zone determined by the Ministry of Forestry and Waterworks of Turkey, was extracted from each image (Figure 3).

Image classification

An unsupervised classification algorithm (ISODATA) was used to classify the selected Landsat images. The ISODATA (iterative self-organizing data analysis technique) algorithm is an iterative method that uses minimum distance to cluster data elements into different classes (Tou & Gonzales 1974). Unsupervised classification is useful when there are no historical data about the study area. In this classification method, the number of classes assigned by the analyst is important as it affects whether spectral-radiometric variability is captured to obtain homogeneous classes (Yang & Lo 2002).

In this study, the images were classified initially into 75 classes, and then these classes were grouped in seven information classes explained in Table 1 (water, watery ground, dry lake, mud/vegetated flats, salt flats, shrubs/sedges, and agriculture). During class assignment, we examined the signature characteristics using signature mean plots (which are plots that show the mean reflectance values of each class in all bands of the image to be classified; Figure 4) and also used our knowledge of the study area. The classes included in each information class had a common spectral behavior, which is easily separable from the spectral behavior of other classes. Mean reflectance values of the information classes are displayed in Figure 4.
Table 1

Information classes and their descriptions

Information class Description 
Water Open water such as lakes or ponds 
Watery ground Water-imbedded surface 
Dry lake Dry lake bottom with no vegetation cover 
Shrubs/Sedges Sparsely vegetated soils with Carex spp. and Tamarix spp. 
Salt flats Saline bare soils 
Mud/Vegetated flats Saline-hydromorphic soils on the lake shore and saline soils with dense hydro-halophytic vegetation, such as Salicornia 
Agriculture Irrigated or non-irrigated crop fields. Crops include sugar beets, sunflower, wheat, and barley 
Information class Description 
Water Open water such as lakes or ponds 
Watery ground Water-imbedded surface 
Dry lake Dry lake bottom with no vegetation cover 
Shrubs/Sedges Sparsely vegetated soils with Carex spp. and Tamarix spp. 
Salt flats Saline bare soils 
Mud/Vegetated flats Saline-hydromorphic soils on the lake shore and saline soils with dense hydro-halophytic vegetation, such as Salicornia 
Agriculture Irrigated or non-irrigated crop fields. Crops include sugar beets, sunflower, wheat, and barley 
Figure 4

Mean reflectance values of the seven information classes (water, watery ground, dry lake, mud/vegetated flats, salt flats, shrubs/sedges, and agriculture).

Figure 4

Mean reflectance values of the seven information classes (water, watery ground, dry lake, mud/vegetated flats, salt flats, shrubs/sedges, and agriculture).

Accuracy assessment

We analyzed the accuracies of the classifications using 150 random points. Points were selected using stratified random sampling by taking at least 15 points from each class. Reference data were extracted from the images and from the land cover maps (1/25,000) provided by the Ministry of Forestry and Waterworks’ Land Monitoring System.

To determine the accuracy of the classifications, a standard error matrix (confusion matrix), which reported producer's, user's, and overall classification accuracies, and kappa statistics were used (Congalton & Green 1999). An error matrix is a standard matrix, where each row represents the image classification and each column represents the reference classification. Table 2 provides an example error matrix, where n samples are distributed into k2 cells, and k refers to the number of classes. Using the error matrix, the producer's, user's, and overall accuracy can be calculated by Equations (1), (2), and (3), respectively. In these equations, nij denotes the number of sampling points classified into class i (i = 1,2,3,…k) on the image and class j (j = 1,2,3,…k) in the reference data set. The producer's accuracy shows the proportion of sampling points that are correctly classified, and can be found as the ratio of correctly classified points to the total points assigned to that class in the reference data set (Equation (1)). The user's accuracy shows the proportion of classified pixels that are correctly classified and can be found as the ratio of correctly classified pixels to the total pixels assigned to that class in the classified map (Equation (2)). Overall accuracy is the ratio of correctly classified points to the total points in an image (Equation (3)). 
formula
1
 
formula
2
 
formula
3
The Kappa statistic shows the chance-corrected agreement between classified and actual classes. It can be calculated as in Equation (4). In this equation, expected accuracy can be calculated by first creating a matrix of the products of rows and columns in the error matrix and then dividing the sum of the diagonal cell values by the sum of all cell values (Verbyla 1995). 
formula
4
Table 2

An example error matrix

  Reference 
Classification Class 
n11 n11 n1k 
n21 n22 n2k 
nk1 nk2 nkk 
  Reference 
Classification Class 
n11 n11 n1k 
n21 n22 n2k 
nk1 nk2 nkk 

Change detection

After image classification and accuracy assessment steps, areal coverages for all classes were determined on each image and then compared with each other. The changes between two successive images (e.g., 1987 and 2000, 2000, and 2003) and between the images of the wettest (1987) and driest (2003) years were identified using a post-classification change detection algorithm. The post-classification approach requires independent classifications of images, followed by pixel-by-pixel comparison. We selected this approach because we wanted to obtain ‘from-to’ change information for each time interval considered.

RESULTS AND DISCUSSION

Overall accuracies of the 1987, 2000, 2003, and 2011 classifications (using reference data from the images) were 86%, 88%, 90%, and 95%, respectively (Table 3). Producer's and user's accuracies of individual classes ranged from 70 to 100%. Kappa statistic values for the same classifications were 0.83, 0.86, 0.87, and 0.94, respectively.

Table 3

Producer's, user's, and overall accuracies (%) and kappa statistics for the 1987, 2000, 2003, and 2011 classifications

  1987
 
2000
 
2003
 
2011
 
Class Producer's User's Producer's User's Producer's User's Producer's User's 
Water 100 89 100 98   100 100 
Watery area 78 90 95 86   100 100 
Dry lake     100 96   
Mud/Vegetated flats 86 80 93 87 88 100 94 100 
Salt flats 100 95 91 91 100 82 100 95 
Shrubs/Sedges 87 77 79 77 78 89 86.67 93 
Agriculture 68 85 77 92 86 86 92 85 
Overall accuracy 86 88 90 95 
Overall Kappa statistics 0.83 0.86 0.87 0.94 
  1987
 
2000
 
2003
 
2011
 
Class Producer's User's Producer's User's Producer's User's Producer's User's 
Water 100 89 100 98   100 100 
Watery area 78 90 95 86   100 100 
Dry lake     100 96   
Mud/Vegetated flats 86 80 93 87 88 100 94 100 
Salt flats 100 95 91 91 100 82 100 95 
Shrubs/Sedges 87 77 79 77 78 89 86.67 93 
Agriculture 68 85 77 92 86 86 92 85 
Overall accuracy 86 88 90 95 
Overall Kappa statistics 0.83 0.86 0.87 0.94 

The accuracies of classifications were highest for water, watery ground, dry lake, mud/vegetated flats, and salt flats classes and lowest for agriculture and shrubs/sedges classes. The low accuracy for agriculture and shrubs/sedges classes originates from the similarity of their spectral signatures. As can be seen from the mean reflectance of information classes provided in Figure 4, the best separation was possible between the classes representing water, watery ground and the other classes representing waterless surfaces (shrubs/sedges, agriculture, and dry lake). In 2003, which was the driest year, mean reflectance values for dry lake and mud flats classes were higher than those detected in other years. However, the general trends were almost the same. This difference is most probably caused by the layer of salt at the bottom of the lake. As the lake was completely dry in 2003, the response of the dry salt layer, which has a higher reflecting value, was more apparent.

Areal coverage of individual land cover classes in 1987, 2000, 2003, and 2011 are shown graphically in Figure 5 and numerically in Table 4. To further evaluate the results of land cover conversions, matrices of land cover changes from 1987 to 2000, 2000 to 2003, 2003 to 2011 were created (Table 5). The land cover change matrices cross-tabulate land cover at two different years. For example, in the first matrix in Table 4, the land cover changes from 1987 to 2000 are presented. In the matrices, unchanged pixels are located on the diagonal (e.g., 15.1 km2 area was covered with water both in 1987 and 2000). The changes from one class to the other are represented in other cells (e.g., 2.7 km2 area covered by water in 1987 was converted to watery ground in 2000).
Table 4

Areas of individual land cover classes in 1987, 2000, 2003, and 2011

  1987
 
2000
 
2003
 
2011
 
Class Area (km2Area (%) Area (km2Area (%) Area (km2Area (%) Area (km2Area (%) 
Water 17.8 37.3 15.6 32.7 0.0 0.0 17.6 36.9 
Watery ground 3.6 7.5 5.6 11.8 0.0 0.0 4.5 9.5 
Dry lake 0.0 0.0 0.0 0.0 17.3 36.3 – – 
Mud/Vegetated flats 1.1 2.3 1.9 4.0 1.3 2.8 1.2 2.6 
Salt flats 5.4 11.4 5.4 11.4 8.6 18.1 5.0 10.5 
Shrubs/Sedges 14.9 31.5 11.1 23.4 12.2 25.7 9.7 20.4 
Agriculture 4.8 10.1 7.9 16.6 8.2 17.2 9.6 20.1 
Total 47.6 100 47.6 100 47.6 100 47.6 100 
  1987
 
2000
 
2003
 
2011
 
Class Area (km2Area (%) Area (km2Area (%) Area (km2Area (%) Area (km2Area (%) 
Water 17.8 37.3 15.6 32.7 0.0 0.0 17.6 36.9 
Watery ground 3.6 7.5 5.6 11.8 0.0 0.0 4.5 9.5 
Dry lake 0.0 0.0 0.0 0.0 17.3 36.3 – – 
Mud/Vegetated flats 1.1 2.3 1.9 4.0 1.3 2.8 1.2 2.6 
Salt flats 5.4 11.4 5.4 11.4 8.6 18.1 5.0 10.5 
Shrubs/Sedges 14.9 31.5 11.1 23.4 12.2 25.7 9.7 20.4 
Agriculture 4.8 10.1 7.9 16.6 8.2 17.2 9.6 20.1 
Total 47.6 100 47.6 100 47.6 100 47.6 100 
Table 5

‘From-to’ changes (km2) at Tuzla Lake from 1987 to 2011

1987–2000
 
  2000
 
1987 Water Watery ground Dry lake Mud flats Salt flats Shrubs/Sedges Agriculture Total 
Water 15.1 2.7 0.0 0.0 0.0 0.0 0.0 17.8 
Watery ground 0.5 2.4 0.0 0.4 0.2 0.0 0.1 3.6 
Mud flats 0.0 0.3 0.0 0.6 0.1 0.0 0.1 1.1 
Salt flats 0.0 0.1 0.0 0.5 3.5 1.0 0.2 5.4 
Shrubs/Sedges 0.0 0.0 0.0 0.3 1.5 8.2 4.9 14.9 
Agriculture 00 0.0 0.0 0.1 0.2 1.8 2.7 4.8 
Total 15.6 5.6 0.0 1.9 5.4 11.1 7.9 47.6 
2000–2003 
  2003
 
2000 Water Watery ground Dry lake Mud flats Salt flats Shrubs/Sedges Agriculture Total 
Water 0.0 0.0 14.9 0.4 0.2 0.0 0.0 15.6 
Watery ground 0.0 0.0 2.3 0.9 2.2 0.1 0.0 5.6 
Mud flats 0.0 0.0 0.0 0.0 1.1 0.7 0.1 1.9 
Salt flats 0.0 0.0 0.0 0.0 3.4 1.6 0.4 5.4 
Shrubs/Sedges 0.0 0.0 0.0 0.0 1.2 6.6 3.4 11.1 
Agriculture 0.0 0.0 0.0 0.0 0.4 3.2 4.3 7.9 
Total 0.0 0.0 17.3 1.3 8.6 12.2 8.2 47.6 
2003–2011 
  2011
 
2003 Water Watery ground Dry lake Mud flats Salt flats Shrubs/Sedges Agriculture Total 
Dry lake 16.9 0.4 0.0 0.0 0.0 0.0 0.0 17.3 
Mud flats 0.5 0.9 0.0 0.0 0.0 0.0 0.0 1.3 
Salt flats 0.2 3.3 0.0 0.8 2.9 1.0 0.4 8.6 
Shrubs/Sedges 0.0 0.0 0.0 0.4 1.7 5.9 4.2 12.2 
Agriculture 0.0 0.0 0.0 0.1 0.4 2.7 4.9 8.2 
Total 17.6 4.5 0.0 1.2 5.0 9.7 9.6 47.6 
1987–2000
 
  2000
 
1987 Water Watery ground Dry lake Mud flats Salt flats Shrubs/Sedges Agriculture Total 
Water 15.1 2.7 0.0 0.0 0.0 0.0 0.0 17.8 
Watery ground 0.5 2.4 0.0 0.4 0.2 0.0 0.1 3.6 
Mud flats 0.0 0.3 0.0 0.6 0.1 0.0 0.1 1.1 
Salt flats 0.0 0.1 0.0 0.5 3.5 1.0 0.2 5.4 
Shrubs/Sedges 0.0 0.0 0.0 0.3 1.5 8.2 4.9 14.9 
Agriculture 00 0.0 0.0 0.1 0.2 1.8 2.7 4.8 
Total 15.6 5.6 0.0 1.9 5.4 11.1 7.9 47.6 
2000–2003 
  2003
 
2000 Water Watery ground Dry lake Mud flats Salt flats Shrubs/Sedges Agriculture Total 
Water 0.0 0.0 14.9 0.4 0.2 0.0 0.0 15.6 
Watery ground 0.0 0.0 2.3 0.9 2.2 0.1 0.0 5.6 
Mud flats 0.0 0.0 0.0 0.0 1.1 0.7 0.1 1.9 
Salt flats 0.0 0.0 0.0 0.0 3.4 1.6 0.4 5.4 
Shrubs/Sedges 0.0 0.0 0.0 0.0 1.2 6.6 3.4 11.1 
Agriculture 0.0 0.0 0.0 0.0 0.4 3.2 4.3 7.9 
Total 0.0 0.0 17.3 1.3 8.6 12.2 8.2 47.6 
2003–2011 
  2011
 
2003 Water Watery ground Dry lake Mud flats Salt flats Shrubs/Sedges Agriculture Total 
Dry lake 16.9 0.4 0.0 0.0 0.0 0.0 0.0 17.3 
Mud flats 0.5 0.9 0.0 0.0 0.0 0.0 0.0 1.3 
Salt flats 0.2 3.3 0.0 0.8 2.9 1.0 0.4 8.6 
Shrubs/Sedges 0.0 0.0 0.0 0.4 1.7 5.9 4.2 12.2 
Agriculture 0.0 0.0 0.0 0.1 0.4 2.7 4.9 8.2 
Total 17.6 4.5 0.0 1.2 5.0 9.7 9.6 47.6 
Figure 5

Land cover classifications in 1987, 2000, 2003, and 2011.

Figure 5

Land cover classifications in 1987, 2000, 2003, and 2011.

From 1987 to 2011, major changes were detected in agriculture and shrubs/sedges classes. Agriculture expanded by 4.7 km2 in the study area from 1987 to 2011. This has almost doubled the area covered by agricultural areas in 2011, compared to 1987. Expansion was mostly on the western and eastern sides of Tuzla Lake. The areal coverage of shrubs/sedges decreased from 1987 to 2011. The area covered by the shrubs/sedges class decreased by 4.4 km2. Almost all areas lost from shrubs/sedges class were converted to agriculture. The areas covered by classes, water, watery ground, salt flats, and mud/vegetated flats were very similar for the years 1987 and 2011. As the climatic conditions in these 2 years (i.e., 1987 and 2011) were very similar (Table 6), we can say that the lake area remained almost the same under similar climatic conditions. However, agricultural areas showed a gradual expansion and shrubs/sedges areas showed a gradual shrinkage from 1987 to 2011.

Table 6

Annual average air temperatures, annual precipitation, and annual pan evaporation measured at Kayseri Meteorology Station for the years 1987, 2000, 2003, and 2011 (data obtained from Turkish State Meteorology Service (2012))

Years Annual average air temperature (°C) Annual precipitation (mm) Annual pan evaporation (mm) 
1987 10.1 532 884 
2000 10.2 356 1,038 
2003 11.3 286 1,116 
2011 10.1 490 913 
Years Annual average air temperature (°C) Annual precipitation (mm) Annual pan evaporation (mm) 
1987 10.1 532 884 
2000 10.2 356 1,038 
2003 11.3 286 1,116 
2011 10.1 490 913 

In 2000, we see some expansion in the area covered by watery ground and mud/vegetated flats, and shrinkage in the area covered by water. 2000 is the year with lower precipitation and higher temperatures, compared to 1987 and 2011. With the classification scheme in 2000, we can see the response of the lake to climatic variations. Under dry conditions, water areas show a reversible conversion to watery ground and mud/vegetated flats. Decreased direct precipitation as well as decreases in groundwater levels due to extensive pumping in dry years (Cengiz & Dadaser-Celik 2012) can be responsible for these changes. The decrease in water levels for the same year can also be seen in Figure 6, developed using the data reported by Cengiz & Dadaser-Celik (2012).
Figure 6

Changes in water levels at Tuzla Lake from 1998 to 2005 (adapted from Cengiz & Dadaser-Celik (2012)).

Figure 6

Changes in water levels at Tuzla Lake from 1998 to 2005 (adapted from Cengiz & Dadaser-Celik (2012)).

The classes identified in the 2003 image were completely different. In this year we see a new class, dry lake, which covers an area of 17.3 km2 (36.3%). This area is almost equal to the sum of water and watery ground classes detected in other years. This shows that the lake was completely dry in 2003. As can be seen from Table 6, 2003 was the driest and hottest year among the years analyzed. The precipitation amounts were almost half of the precipitation measured in 1987 and 2011 and average air temperature in 2003 was about 3 °C higher than air temperatures in 1987 and 2011. High temperatures and low precipitation, as well as decreases in groundwater flows due to extensive pumping, caused the loss of all water from the lake. 2003 was detected as the year with lowest water levels during the 1998–2005 period based on data reported by Cengiz & Dadaser-Celik (2012) (Figure 6).

To be able to compare the conditions in different climatic settings, we calculated the percent changes between the wettest and driest years (i.e., 1987 and 2003) (Table 7). Between the wettest (1987) and driest (2003) years, significant changes were detected in the area covered by water and watery ground classes. Almost all the area covered by water and watery ground classes in 1987 was converted to dry lake in 2003. Salt flats were also expanded a little from 1987 to 2003. We see an expansion in agricultural areas and shrinkage in shrubs/sedges areas.

Table 7

‘From-to’ changes (km2) at Tuzla Lake between the wettest (1987) and the driest (2003) years

  2003
 
1987 Water Watery ground Dry lake Mud flats Salt flats Shrubs/Sedges Agriculture Total 
Water 0.0 0.0 17.0 0.5 0.2 0.0 0.0 17.8 
Watery ground 0.0 0.0 0.2 0.8 2.2 0.3 0.1 3.6 
Mud flats 0.0 0.0 0.0 0.0 0.7 0.3 0.1 1.1 
Salt flats 0.0 0.0 0.0 0.0 3.7 1.3 0.4 5.4 
Shrubs/Sedges 0.0 0.0 0.0 0.0 1.5 8.6 4.9 14.9 
Agriculture 0.0 0.0 0.0 0.0 0.2 1.7 2.8 4.8 
Total 0.0 0.0 17.3 1.3 8.5 12.2 8.2 47.6 
  2003
 
1987 Water Watery ground Dry lake Mud flats Salt flats Shrubs/Sedges Agriculture Total 
Water 0.0 0.0 17.0 0.5 0.2 0.0 0.0 17.8 
Watery ground 0.0 0.0 0.2 0.8 2.2 0.3 0.1 3.6 
Mud flats 0.0 0.0 0.0 0.0 0.7 0.3 0.1 1.1 
Salt flats 0.0 0.0 0.0 0.0 3.7 1.3 0.4 5.4 
Shrubs/Sedges 0.0 0.0 0.0 0.0 1.5 8.6 4.9 14.9 
Agriculture 0.0 0.0 0.0 0.0 0.2 1.7 2.8 4.8 
Total 0.0 0.0 17.3 1.3 8.5 12.2 8.2 47.6 

The changes in water and land cover in Tuzla Lake can have implications related to the ecological structure of Tuzla Lake and human activities in the Palas Basin. Although no formal studies are available on how bird diversity has changed at the Tuzla Lake through the years, severe decreases in the number and diversity of birds were observed in dry years (e.g., 2000 and 2003) at other wetlands in the Central Anatolia region (Reis & Yilmaz 2008). Shrinkage of the lake and conversion of shrubs/sedges areas are likely to decrease bird habitats and cause negative impacts on bird diversity. The decrease in lake area, in contrast, favors salt formation and salt extraction, which is one of the major economic activities in the region.

The conversion of the shrubs/sedges class to agriculture can constitute a threat to the sustainability of Tuzla Lake. Agricultural expansion and associated irrigation activities increase the pressure on groundwater resources and can affect the natural water regime of the lake. Agricultural expansion can also increase the transport of pollutants to the lake and cause degradation in water quality.

CONCLUSIONS

We analyzed land cover changes at Tuzla Lake using Landsat TM images acquired for July 1987, 2000, 2003, and June 2011. We used an unsupervised ISODATA classification method that classified images into seven information classes: water, watery ground, dry lake, mud/vegetated flats, salt flats, shrubs/sedges, and agriculture. The analysis showed that the surface area of Tuzla Lake was significantly smaller in 2000 and especially in 2003 compared to 1987 and 2011. Both decreases in precipitation and increases in air temperatures can be contributing factors to the changes detected for the years 2000 and 2003. An increase in air temperatures would have increased evaporation rates from the surface of the lake and increased the water demand (e.g., irrigation and domestic) in the basin. The scarcity of precipitation would have affected the lake directly by decreasing the amount of precipitation over the lake surface and indirectly through decreases in surface runoff, surface flows from streams and springs, and groundwater flows. The analysis showed that Tuzla Lake responds quickly to variations in climatic conditions. The most important changes from 1987 to 2011 were detected in agriculture and shrubs/sedges classes around the lake. Agriculture expanded by 4.7 km2 in the study area from 1987 to 2011, and a large proportion of this expansion was due to the conversion of areas previously covered by shrubs/sedges to agriculture. Expansion of agriculture in the basin may threaten the sustainability of Tuzla Lake in the future by increasing the water demand for irrigation and causing transport of pollutants to Tuzla Lake.

Overall, this paper shows that saline lakes in semi-arid landscapes are vulnerable to changes in climatic conditions that affect water inflows and outflows. The expansion of irrigated agriculture in semi-arid regions will make conservation of salt lakes extremely difficult, especially under the expected climatic changes. In these critical areas, in order to preserve the ecological balance, the decisions about infrastructure works (dams, diversions, and irrigation) should be made more carefully. Continuous monitoring (at least annually) of such critical environmental regions is also essential.

ACKNOWLEDGEMENTS

This study was supported by the Research Fund of Erciyes University (Project No: FBA-12-3953 and FBD-12-4131).

REFERENCES

REFERENCES
Bryant
R. G.
1999
Application of AVHRR to monitoring a climatically sensitive playa
.
Case study: Chott El Djerid, Southern Tunisia. Earth Surf. Process. Landf.
24
,
283
302
.
Castenada
C.
Herrero
J.
Auxiliadora Casterad
M.
2005
Landsat monitoring of playa-lakes in the Spanish Monegros desert
.
J. Arid Environ.
63
(
2
),
497
516
.
Cengiz
E.
Dadaser-Celik
F.
2012
Hydrologic changes at Tuzla (Palas) Lake in Turkey
.
BALWOIS 2012-International Conference on Water, Climate and Environment
,
Ohrid
,
Republic of Macedonia
,
28 May–2 June
.
Congalton
R.
Green
K.
1999
Assessing the Accuracy of Remotely Sensed Data: Principles and Practices
.
CRC/Lewis Press
,
Boca Raton, FL
,
USA
.
Dadaser-Celik
F.
Bauer
M. E.
Patrick
L. B.
Stefan
H. G.
2008
Changes in the Sultan Marshes ecosystem (Turkey) in satellite images 1980–2003
.
Wetlands
28
(
3
),
852
865
.
Drake
N. A.
Bryant
R. G.
1994
Monitoring the flood ratio of Tunisian playas using Advanced Very High Resolution Radiometer (AVHRR) imagery
. In:
Environmental Change in Drylands: Biogeographical and Geomorphological Perspectives
(
Millington
A. C.
Pye
K.
, eds).
Wiley
,
New York
, pp.
347
364
.
DSI
1970
Kayseri-Sarıoglan-Palas Ovası Planlama Kademesi Hidrojeolojik Etüt Raporu
[
Kayseri-Sarıoglan-Palas Plain Hydrogeological Survey Report
].
General Directorate of State Hydraulic Works
,
Kayseri
,
Turkey
.
Duffy
C. J.
Al-Hassan
S.
1988
Ground-water circulation in closed desert basins: topographic scaling and climatic forcing
.
Water Resour. Res.
24
(
10
),
1675
1688
.
Dwivedi
R. S.
Rao
B. R. M.
Bhattacharya
S.
1999
Mapping wetlands of the Sundaban Delta and its environs using ERS-1 SAR data
.
Int. J. Remote Sens.
20
(
11
),
2235
2247
.
ENCON, Cevre Danısmanlık Ltd Sti
1999
‘Tuzla Gölü’, Uluslararası önemi olan sulakalanların biyolojik ve ekolojik yönden arastırılması projesi nihai raporu
[
ENCON Environmental Consulting Ltd Sti, Tuzla Lake Project Report – Investigation of biological and ecological aspects of wetlands of international importance
].
Ankara
,
Turkey
.
French
R. H.
Miller
J. J.
Dettling
C.
Carr
J. R.
2006
Use of remotely sensed data to estimate the flow of water to a playa lake
.
J. Hydrol.
325
(
1–4
),
67
81
.
Jellison
R.
Zadereev
R. S.
DasSarma
P. A.
Melack
J. M.
Michael
R.
Andrei
G.
DasSarma
S.
Zambrana
G.
2004
Conservation and Management Challenges of Saline Lakes: A Review of Five Experience Briefs. Lake Basin Management Initiative Thematic Paper. http://www.worldlakes.org/uploads/Salt%20Lakes%20Thematic%20Paper%2022Jun04.pdf (accessed 11 February 2014).
Jensen
J. R.
2005
Introductory Digital Image Processing: A Remote Sensing Perspective
.
Pearson Prentice Hall
,
Upper Saddle River, NJ
,
USA
.
Kasishke
E. S.
Bourgeau-Chavez
L. L.
1997
Monitoring South Florida wetlands using ERS-1 SAR imagery
.
Photogramm. Eng. Remote Sens.
63
(
3
),
281
291
.
Kienast-Brown
S.
Boettinger
J. L.
2007
Land-cover classification from Landsat imagery for mapping dynamic wet and saline soils
.
Dev. Soil. Sci.
18
(
31
),
235
244
.
Kohfahl
C.
Rodriguez
M.
Fenk
C.
Menz
C.
Benavente
J.
Hubberten
H.
Meyer
H.
Paul
L.
Knappe
A.
Lopez-Getae
J. A.
Pekdeger
A.
2008
Characterizing flow regime and interrelation between surface-water and ground-water in the Fuente de Piedra salt lake basin by means of stable isotopes, hydrogeochemical and hydraulic data
.
J. Hydrol.
351
(
1–2
),
170
187
.
Li
M. S.
Mao
L. J.
Shen
W. J.
Liu
S. Q.
Wei
A. S.
2013
Change and fragmentation trends of Zhanjiang mangrove forests in southern China using multi-temporal Landsat imagery (1977–2010)
.
Estuarine Coast. Shelf Sci.
130
,
111
120
.
Nakayama
Y.
Tanaka
S.
Sugimura
T.
Endo
K.
1997
Analysis of hydrological changes in lakes of Asian arid zone by satellite data
. In:
Earth Surface Remote Sensing, Proceedings of SPIE. The International Society for Optics and Photonics
,
vol. 3222
(
Cecchi
G.
Engman
E. T.
Zilioli
E.
, eds).
Bellingham, Washington
, pp.
201
210
.
Piwowar
J. M.
Ledrew
E. F.
2002
ARMA time series modeling of remote sensing imagery: a new approach for climate change studies
.
Int. J. Remote Sens.
23
(
24
),
5225
5248
.
Poff
N.
Brinson
M.
Day
J.
2002
Aquatic Ecosystems and Global Climate Change: Potential Impacts on Inland Freshwater and Coastal Wetland Ecosystems in the United States. Pew Center on Global Climate Change
,
USA
.
Pope
A.
Rees
G.
2014
Using in situ spectra to explore Landsat classification of glacier surfaces
.
Int. J. Appl. Earth Observ. Geoinform.
27
(
Part A
),
42
52
.
Rao
B. R. M.
Dwivedi
R. S.
Kushwaha
P. S.
Bhattacharia
S. N.
Anand
J. B.
Dasgupta
S.
1999
Monitoring the spatial extent of coastal wetlands using ERS-1 SAR data
.
Int. J. Remote Sens.
20
(
13
),
2509
2517
.
Reis
S.
Yılmaz
H.
2008
Temporal monitoring of water level changes in Seyfe Lake using remote sensing
.
Hydrol. Process.
22
(
22
),
4448
4454
.
Rezaei Moghaddam
M. H.
Saghafi
M.
2006
A change-detection application on the evolution of Kahak playa (South Khorasan province, Iran)
.
Environ. Geol.
51
(
4
),
565
579
.
Simmons
C. T.
Narayan
K. A.
1997
Mixed convection processes below a saline disposal basin
.
J. Hydrol.
194
(
1–4
),
263
285
.
Singh
A.
1989
Digital change detection techniques using remotely-sensed data
.
Int. J. Remote Sens.
10
(
6
),
989
1003
.
Somuncu
M.
1999
Palas Ovası’nda cografya gözlemleri
[
Geographical observations at the Palas Basin
].
Ankara University – Journal of Turkish Geography Research and Application Center
5
,
183
211
.
Swain
P. H.
Davis
S. M.
1978
Remote Sensing: The Quantitative Approach
.
McGraw-Hill Book Company
,
New York, NY
,
USA
.
Thakur
J. K.
Srivastava
P. K.
Singh
S. K.
Vekerdy
Z.
2012
Ecological monitoring of wetlands in semi-arid region of Konya closed Basin, Turkey
.
Reg. Environ. Change
12
,
133
144
.
Tou
J. T.
Gonzales
R. C.
1974
Pattern Recognition Principles
.
Addison-Wesley
,
Reading, MA
,
USA
.
Turkish State Meteorological Service
2012
Yıllık ortalama sıcaklık ve yağış Analiz Raporu
[
The average annual temperature and precipitation Analysis Report]. http://www.mgm.gov.tr/veridegerlendirme/yillik-toplam-yagis-verileri.aspx?m=kayseri (accessed 19 November 2012).
UWPCC
2007
University of Washington, Program on Climate Change. http://www.uwpcc.washington.edu (accessed 5 May 2012
).
Verbyla
D. L.
1995
Satellite Remote Sensing of Natural Resources
.
CRS Lewis Publishers
,
New York
,
USA
.
Yechieli
Y.
Wood
W. W.
2002
Hydrogeologic processes in saline systems: playas, sabkhas, and saline lakes
.
Earth-Sci. Rev.
58
(
3–4
),
343
365
.