1. Introduction

Land subsidence is a remarkable phenomenon that seems to have attracted many experts and scholars. It is so important that UNESCO has introduced it as a designated topic for research [Gens, 1998]. According to the Geological Institute of America, this phenomenon is the collapse or subsidence of land surface downward, with little movement in horizontal orientation, that isn't limited in terms of the extent and severity. Important factors involved in this phenomenon, are melting of ices, movement of Earth's crust, earthquakes, volcanic activities and lava out of the ground or industrial and mining activities such as extraction of underground resources include water discharges from aquifers, extraction oil and so on.

One of the reasons for changes in volume of water aquifers is the water with drawls through the wells that excessive increase of water with drawls leads to increase effective stress in the aquifer, furthermore the water loss hydro due to reduction of water levels in aquifers, cause the subsidence phenomenon occurs [Quilin, 2004]. This phenomenon follows effects. Cracks in the land and buildings, upset the balance and channel slope and waterways, and eventually other material losses are among its effects. Land subsidence is a problem facing many countries, and Iran is no exception in this category. Kerman province is one of the provinces where this phenomenon occurred in and uneven decline in groundwater levels in this area, especially in Rafsanjan plain due to indiscriminate use of water for agriculture and irrigation pistachio has made groundwater level in the past 20 years more than 20 meters down and eventually occurs subsides between 5 to 15 centimeters each year [Luigi, 2005]. Critical activities has been done in order to build water storage facilities as well as water transfer systems in Iran which still is not enough and still reliance on groundwater resources and it remains very important factor [Berardino et.al, 2002]. Several studies have been conducted to examine this phenomenon in Rafsanjan plain which because of 3 reasons, its results is unacceptable for the research studies:

1- It did not conduct in accordance with the present technology and there is lack of appropriate data.

2- The antecedence data is less than the current data.

3- The main theme of those studies wasn’t land subsidence but they have discussed on examination of the amount of water in the wells.

In this paper land subsidence in Rafsanjan plain has been discussed. The main objectives of this paper are zoning of land subsidence and identify potential areas with high subsidence event. Thus, it has been used of different parameters of geology and hydrogeology affecting on the occurrence of this phenomenon. On the basis of influence degree of each, these data were combined in geospatial information systems (GIS). In the second part, in order to measure the accuracy of the extracted regions which have subsidence potential, Interferometric Synthetic Aperture Radar (InSAR) technique is used. InSAR technique is a method that uses the phase data of two radar images that are taken at different times from the same area and will estimate earth’s surface displacement during this period [Zebker, 1986]. This method will be able to map the surface displacement in a wide continuous coverage by high spatial resolution and with an accuracy of less than 1 cm. Since InSAR technique has been used to measure the earth’s surface displacement caused by subsidence [Dehghani, et. al. 2010, Crosetto, et. al., 2002 and Chang and Wang, 2004].

If the time interval between two radar images was so extent that changes the characteristics of the earth’s surface dispersion due to lack of time correlation, we would not be able to use this method [Zebker, 1992]. In this section, the InSAR displacement map of Rafsanjan plain which using short time intervals images to prevent temporal decorrelation is compared to zoning map. By comparison the results of two above methods can realize the accuracy of zoning land subsidence and attributed weight to each data layer.

2. Materials and Methods

In this part of paper, the zoning of land subsidence is examined in Rafsanjan plain area in terms of subsidence potential and talent. But first we should present data and information layers that are used. Several factors are involved in creation of this important phenomenon that independently has particular effects and has been analyzed by weight assigned to each layer. The vegetation of the area, which is divided into several categories are one of available information layers that its map was produced by the Forest and Rangeland and Watershed Organization and were used for zoning. Also in order to complete the mentioned layer this information were obtained of +ETM images of LANDSAT7 satellite. The final information layer was collected by combining the information layers of vegetation available maps in Forests and Rangeland Organization with vegetation maps derived from satellite imagery +ETM. Full output layer contains vegetation, including forests, farms and fields that are shown in the figure below.

Figure 1. Vegetation layer produced from LANDSAT 7 satellite data and maps of Forest and Rangeland and Watershed Org

Figure 2. Location of Pizometric water wells in the area

Pizometric wells in the area can be also a very good source of information to observe this phenomenon. The existence and necessity to use these wells is important as may find out the amount of the monthly or annual removal water by review fluctuations and changes of the existing water level. Due to the high compliance status of Pizometric wells with vegetation derived in the previous section, it was found that the maximum amount of water is used for agricultural purposes.

For the study and measurement fluctuations and changes of Pizometric wells water level and processing of data obtained from the time series, the programming was done in MATLAB environment. The aims to consider MATLAB programming environment are include:

1- Organizing wells data by distinct observed during for structuring

2- Create a time series for depth of water level and its charting; this issue is important because the drop in water level in wells reached 2 m in 2 years which shows high decline of aquifer water levels. Time series of the depth of the water level in some wells, is shown in the following figure. It should be noted that the rise in depth of the water in a wells corresponding to drop water level in that well.

The point to be noted is that despite, being fed aquifer in rainy season, the overall water level in wells has been trend downward. The result of this event is that the extraction of water from underground aquifers is more than feed them in seasonal rainfall and if this trend continues, there will be adverse and irreparable effects to the environment, and ultimately will lead to the phenomenon of subsidence and its effects.

3- The other objectives of the MATLAB program was the regression line fitted to the time series of wells to determine water falling levels in those wells. The water wells in the subsidence area, based on the slope of the regression line was divided into 6 categories and then any of these categories with different weights entered into the modeling.

Figure 3. Time series of the depth of the water level in some wells located in the area

Table 1. Classification of wells Wells layer Water drop level Well 1 Water rise level Well 2 Constant water level Well 3 Low water drop levels Well 4 Medium water drop levels Well 5 high water drop levels Well 6 Too high water drop levels

The regional soil type and gender is also considered one of the main factors causing land subsidence. Soil itself can be divided into several categories which each will play a significant role due to their capabilities to cause this phenomenon. Conducted studies on various sources of lands lead to make four different groups based on having talent and potential for causing land subsidence. These groups are: particularly prone areas, relatively prone areas, prone areas and non-prone areas.

Concentration and density of wells is among useful information that can help to identify areas affected by land subsidence. Since the accumulation and large numbers of this occurrence (natural or artificial) in the area will drop ground water levels, increasing the probability of land subsidence by more concentrated wells. In other words it can be stated that a single or small number of wells cannot lead to land subsidence. The number of wells in the area is increased; the probability and subsidence rates will be increase. In this research to provide the wells density, they were counted in certain levels (Figure 4).

Figure 4. Map of assembly wells in the region

Up to this step, required factors for the zoning of land subsidence was prepared in different information layers by presented techniques which have different contribution in occurrence subsidence phenomenon and is assigned specific weight for any of them. The most important steps in land subsidence modeling are to determine the weight of each layer separated that to avoid the bias and also to decrease the correlation between layers, one of the methods of multi-criteria decision models (MCDM) is used. Since the contribution of each criterion to create the subsidence is different, for every of them, a weight according to influence should be chosen. In this step, for measuring the related weights to each factor, the AHP technique that is one of the multiple attribute decisions is used. The reason of using this method is the hierarchical structure indices, the experts considered relevant to this study while it is easy to implement. In this point, the imposed comments in AHP are determined according to the obtained experiences by experts in different subsidence like: Tehran flat, Mashhad, Nishabour, Qazvin and Semnan in Iran. Now these layers of information had to combine so carefully and logical. In this research, have been used overlay weighted sum method and multiply each data layer to its attributed weight and then their results sum together. This method was chosen because it is simple to implement and yet the results are reasonable. Associated weights with each data layer are shown in Table 2. Generally can’t establish a linear relationship between the drop in water levels and subsidence rates, because many factors are involved in this phenomenon that one of them is existence of compressible inter layers of silt and clay in the aquifer. Small hydraulic conductivity in these inter layers in vertical direction may cause the balance of water level occur with a time lag which is due to irregular water harvesting in underground aquifers and may take several years to re-balance according to characteristics of the aquifer. So all the wells in operation which their water level has little change should be considered in the calculations too. In any of the six layers of the wells, neighborhood buffer analysis (Buffering) was performed and was created 5 buffers around each well by distances of 100, 200, 300, 400 and 500 m during the analysis. And intervals of buffers are smaller, more vulnerable parts of its range were exposed to subsidence. Thus, more weight is assigned to itself and going farther beyond the wells provide regular increase the buffer distance and less effective weight in the analysis.

Table 2. Buffer distances from the center of the wells, and Effective Weight Analysis Buffer amount of Wells Centre (meter) weight 100 5 200 6 300 7 400 8 500 9

In the phase data layers preparation to start modeling and using in overlay, were used raster data formats. The layers of vector data type were converted to raster by choosing 50 m grid size. Layer of density and accumulation of the well itself was raster and produced in 50-meter grid sizes.

After preparation layers and assignment of weights associated to each layer, by using the Weighted Sum method data were combined. Its output was integration of raster layers in which the pixel values are assigned 3 to 33.1 and indicates that larger amounts has higher risk degree for land subsidence and pixels with less value have less risk degree for land subsidence.

For more clear results, the output layer based on the pixel values and amount are divided into 5 categories and new values were expressed in vector format (Figure 6).

Figure 5. Result of combining layers in Weighted Sum Method

Figure 6. The result of re-classification of the output layer in Weighted Sum Method

After obtaining the results of modeling subsidence is better to compare it with subsidence observations and then assess it. Among the space and ground approaches only InSAR technique is capable to measure earth's surface movement with spatial resolution and high accuracy in a wide land cover. So in this study above method is used to evaluate modeling process. The method is based on measuring the movements of the earth's surface using duplicate images prepared from radar sensors [5]. The method is integrating the image was taken from an area within a specified time and the image is made from the same district in another time by a radar sensor and after production interferogram and related processing; we attempt to determine the displacement and subsidence rate. Outlines the principles and basis and also performance of InSAR techniques will discuss later and this amount is sufficient for this paper.

3. Assessment and Results

The GIS system analyzes existing data layers, to determine the level and range of subsidence rates by InSAR technique. The proposed radar data that prepared in a certain time interval were purchased from the Europe Space Agency (ESA). These data were provided from 3 radar images of Track 205 by useable format in InSAR processing called SLC format. Processing the above 3 video data by assistance of software Gamma were produced 3 interferograms.

Table 3. Features of processed interferograms Time baseline (Day) Space baseline (meter) Date of picture (Slave) Date of picture (Master) No. of interferogram 70 185 2005/05/17 2005/03/08 1 140 173 2005/10/04 2005/05/17 2 210 12 2005/10/04 2005/03/08 3

Proposed region is an area with vegetation and mainly are farm landsand for this issue, by passing of time, many changes occur in the earth’ssurface. Using pictures with such a long interval of time makes problems associated with this feature such as temporal decorrelation and finally noise in interferogram that can eliminates the recovery phase (Phase Unwrapping). 3 region thatsubsidence phenomenon in different extent are occurring can be seen in yellow to red spots in interferograms. In generated interferograms the orbital error is visible asmutations phase in the background of interferogram and excluded through fitted to a page outside of the deformation zone and remove the page from the interferograms. Because the data obtained from this area covers less than a year, so the land subsidence rate cannot be converted to the rate of deformation.

The pattern of land subsidence in Rafsanjan plain area is "v" type which shows a gradual change in land subsidencerate of the marginal zone. Using the more radar data with longer time intervals can be more extensive and more detailed studies to determine the rate of land subsidence and deformation. In this paper, the results of the zoning of land subsidence in the area by GIS system were compared with the results of InSAR. For a closer look, layers information on subsidence areas enlarged and assessed their compliance with each other. In the figure below, the two-layers of interferogram 1 and the model results in GIS system have been overlapped.

Figure above shows where have been extracted as land subsidence through interferograms have full compliance with the areas obtained from GIS system modeling and are most likely to show subsidence. In general, radar data and the results of interferometry which have ability to measure earth’s surface and shell displacement with precision less than 1cm can be used to evaluate the results of other methods, such as the zoning of land subsidence in the GIS system. Finally, according to full compliance of the results of measurements and determine land subsidence rate using the two proposed methods, we can concluded that the subsidence phenomenon created in Rafsanjan plain is due to excessive extraction of water from aquifers for agricultural purposes. Certainly, land subsidence was not only due to extraction of water from aquifers, but there are other conditions such as type of vegetation, gender and type of the soil, soil layer compressibility and which these circumstances are provided in Kerman province and especially in Rafsanjan plain that led to high subsidence rate. According to obtained results from this study can be suggested that presented model with specified ranges with high probability of land subsidence should be consider in the issues related to urban development, and by researchers.

Figure 7. Interferogram 1 that has been removed its orbital error

Figure 8. Interferogram 1 and the vector layer in GIS modeling environment

4. Conclusions

Zoning and land subsidence rates in this study, were studied and measured through two different methods. Way in which the available data layers of important factors affected in land subsidence with associated weights were used in determining the range of subsidence by GIS software and also another approach using InSAR techniques to measure the range and rate of land subsidence. Finally, after reviewing the plans of each of the methods we determined that high correspondence of the maps indicate the correct way of combining layers of information and assign a weight to each of them. However, InSAR techniques are the exact method for accurately monitoring the displacement, deformation and land subsidence that can be measured less than 1 CM.

5. Appreciation and Thanks

At the end, we appreciate and thank of the European Space Agency (ESA) to give ENVISAT ASAR radar images and orbital data and sensor calibration requirements for the study and also we thank and appreciated of management of Geological Survey of Iran (GSI).