RADAR INTERFEROMETRY FOR MONITORING LAND SUBSIDENCE DUE TO OVER-PUMPING GROUND WATER IN CRETE, GREECE

Transcription

1 RADAR INTERFEROMETRY FOR MONITORING LAND SUBSIDENCE DUE TO OVER-PUMPING GROUND WATER IN CRETE, GREECE S. P. Mertikas (1), E.S. Papadaki (1) (1) The Technical University of Crete, GR , Chania, Crete, Greece, ABSTRACT The level of water in several wells at the Messara valley, Crete, Greece has dropped 40 meters over the last 20 years. Anecdotal reports describe cracks in the concrete foundations of some residential structures around the valley. These also suggest that ground subsidence may exceed one centimeter, at least, over a period of recharging and withdrawal ground water. Conventional differential SAR interferometry using ERS-SAR and ALOS-PALSAR images, as well as the stacking technique have been applied to monitor this ground deformation. The used images covered the period from 1992 to 2000 and , respectively. A total of 29 ERS-1&2 SAR and 7 ALOS images have been used for forming interferograms. Image pairs with Doppler difference less than 0.20 pulse repetition frequency, and perpendicular baseline smaller than 100m have been used for processing. A Digital Elevation Model with 20-m pixel size and ±7 m height accuracy has also been incorporated in interferometric processing. Atmospheric artifacts have been compensated by using image stacking. The valley to be monitored is densely cultivated and irrigated. Thus, loss of coherence in images has been observed in the C-Band and could not be overcome. Consequently, interferometric results with the ERS-SAR images have been limited. On the other hand, processing of L-band data has brought up a ground deformation that amounted to a subsidence of at least 4 cm/yr. The correlation of the observed ground deformation with respect to water pumping and other geological parameters has also been investigated. 1. INTRODUCTION Land subsidence could be induced by pumping excessive ground water in a region. Such deformation has already been observed in wide areas all over the world. The Po Valley in Italy, the Antelope and the San Joaquin Valleys in the USA, as well as the Bangkok in Thailand are some example cases where this type of deformation is taking place [7]. The phenomenon occurs especially in non-homogeneous porous media environments. Lowering of the level of ground water leaves underground caves which are unable to support the land overlying them. Hence, land subsidence occurs. In urban areas, subsidence could be exaggerated because of the additional weight imposed by the constructions above. There, buildings may crack and sometimes collapse, while roads could twist and brake. Measuring subsidence is commonly carried out by using conventional ground-based techniques, such as precise levelling and geodetic surveys. Nevertheless, new remote-sensing techniques for monitoring ground deformation have recently emerged. Such a powerful tool is the differential interferometry (DInSAR) using Synthetic Aperture Radar (SAR) satellite images on repeat orbits. Several studies on monitoring ground subsidence by DInSAR have already been conducted, especially in the arid, south-western parts of the United States, where agricultural irrigation greatly shrinks aquifers; perhaps irrecoverably ([1], [2], [3] and [6]). A detailed literature review on the application of DInSAR techniques for monitoring subsidence could be found in [9]. The aim of this work is to detect and map ground deformation at the Messara valley in Greece by using differential SAR interferometry. The area under investigation is situated in the south of the Heraklion prefecture of Crete in Greece. During the past 20 years, the volume of ground water in storage is decreasing. Over-exploitation of water aquifers, for irrigation and water supply of cities, has led to a dramatic drop of 40 meters in the level of groundwater. Anecdotal reports of cracks in concrete foundations, in residential structures around the Messara valley, suggest that ground subsidence induced by such intense pumping, exceeds a centimeter over a one winter-summer rechargewithdrawal cycle. Experience however suggests that little information is available about the exact volume of water withdrawn or consumed in the area. Only in 1990, the authorities in Heraklion prefecture of Crete issued some 500 licenses for drilling and pumping water wells [8]. On the other hand, the local population increases. These issues raise great concern over the sustainability of the available water resources in Crete with a possible desertification in the horizon. Therefore, monitoring ground subsidence using precise data can help us in evaluating, and consequently managing properly the balance between supply (inflow) and demand (recharge) in water aquifers. Proc. Fringe 2009 Workshop, Frascati, Italy, 30 November 4 December 2009 (ESA SP-677, March 2010)

2 2. GENERAL SETTING OF THE AREA The Messara valley covers an area of more than 300 km 2 in south east Crete (Fig. 1). North of it, there exists the highest mountain of Crete, the Idi mountain range, with a peak altitude of 2456 m. The Asterousia mountain chain, south of the Messara basin, rises 600 m in the west to 1200 m in the east and constitutes the southernmost mountain range of Europe (Fig. 2). The climate is dry and sub humid. The Asterousia Mt Figure 2. The main crops cultivated at the Messara basin are olive groves and vineyards. The Asterousia Montains are shown in the background. From the geographical and hydrogeological point of view, Messara is distinguished into the eastern part, represented by the Praitoria basin, and the western part which consists of the basins of Asimi, Vagionia, Moires, Pompia and Tympaki. All these areas have been named after the neighbouring small towns (Fig.5). The Messara basin is covered mainly by Quarternary alluvial clays, silts, sands and gravels with thickness from a few metres to 100 m or more. The inhomogeneities of the Plain deposits give rise to great variations in the hydrogeologic conditions even over small distances. The northern slopes are mainly siltymarly Neogene formations while the southern slopes are mainly schists and limestone Mesozoic formations. Figure 1. The Figure above shows the location of the area under investigation in the island of Crete, Greece. The lower image is a LANDSAT TM false colour composite image showing the broader area of Messara Valley, Crete, Greece (near infrared=red, visible red=green and visible blue=blue). Around 80% of the area in Messara is used for agricultural irrigation (Fig. 2). The valley is the largest and most productive region of the island of Crete. 3. METHODOLOGY FOR DEFORMATION MONITORING Conventional differential SAR interferometry using ERS-SAR and ALOS-PALSAR images, as well as the stacking technique have been applied to monitor this ground deformation. The used images covered the period from 1992 to 2000 and from 2007 to 2009, respectively. At the first phase, the conventional DInSAR technique has been applied by processing a total of 29 SAR images from the European Remote Sensing satellites 1&2. Images were acquired in descending mode and were available in the Single Look Complex (SLC- CEOS) format. The frame title was Track 422/Frame These SAR images cover the time interval They have been used to form 406 potential master-slave image pairs for monitoring land subsidence. The number of available interferograms has been reduced using the following sorting criteria:

3 1. Image pairs having perpendicular baseline smaller than 50 m have been selected for processing. 2. Pairs exhibiting a Doppler difference greater than 0.2 Pulse Repetition Frequency (PRF), have not been taken into account. 3. To avoid loss of image coherence, interferograms spanning more than 3 years have been excluded for processing. 4. Pair-wise logic has been performed for eliminating atmospheric and orbital fringes [5]. Interferometric processing has been performed using the DIAPASON software (originally developed at CNES, France). The precise determination of satellite orbits have been based on the Delft orbit archive, available at A Digital Elevation Model (DEM) has been used to eliminate topographic artifacts. The available DEM had a 20-m pixel resolution and a vertical accuracy of ±7m. It was produced using SPOT-4 stereopair images. The stacking technique, i.e., summing or averaging of multiple differential interferograms into a single interferogram, has also been carried out. This technique has the advantage of overcoming the two main shortcomings of conventional DInSAR: 1) the loss of coherence occurring when long temporal baselines are used, and 2) the atmospheric interference mainly due to the wet component of the troposphere. Stacking is justified because reasonable coherence levels can only be achieved over short time intervals. Therefore, averaging as many short-time intreferograms as possible may finally form a single pseudointerferogram corresponding to a longer period. Moreover, the atmosphere is a temporally and spatially uncorrelated propagation medium. Consequently, the more SAR image pairs are applied in stacking, the higher the possibility for the elimination of the atmospheric artifacts. The presence of atmospheric artifacts in a particular date does not necessarily preclude the use of that image in a stack; as long as the master image is being used the same number of times as the slave does. The result of adding deformation interferograms with common images can be compared with one interferogram produced using the first and last image. Atmospheric influence is thus limited to the atmospheric interference present in these two final images [4]. Also, when adding interferograms, the unwrapping of the individual interferometric pairs is not necessary. A large number of master-slave interferometric pairs and several stacking combinations have been thus produced. Finally, it has been concluded that the C-band frequency has not been efficient in producing reliable results for the land subsidence of the area because of the dense vegetation and irrigation applied in Messara Valley. Consistent loss of coherence has been readily noticed for most of the interferograms in the C-band. Small temporal baselines have not helped in mapping the land subsidence expected. Temporal baselines have been either characterized by atmospheric perturbations at the acquisition time of one or both of the images used to form an interferometric pair, or have been too small, e.g., with 1-month temporal separation. On the other hand, long temporal baselines have led to temporal decorrelation and total loss of coherence in intereferograms within the valley. In the next phase, differential SAR intereferometry has been applied using the ALOS (Advanced Land Observing Satellite) PALSAR (Phased Array type L- band Synthetic Aperture Radar) images, acquired by Japan's latest Earth observation satellite. The idea of using ALOS imagery has been based on the fact that the interferometric signal of L-band (λ=23.5 cm) is expected to be more coherent on vegetated areas than the C-band (λ = 5.6 cm). In this preliminary analysis, three ALOS data sets have been used for processing (Tab. 1). For computational purposes, processing has been carried out using only the area of interest and not the full frame images. The potential disadvantages of L-band versus C-band are: First, the lower fringe rate, measured in cm instead of mm, may result in less precise monitoring for the land subsidence; and second, the ionospheric refraction could be 16.5 times worse in L-band than in C-band frequency. The path delays in radar signals, caused by water vapour in the troposphere, are independent of wavelength. So, this distortion should affect both bands, equally [4]. The ALOS images (Tab. 1) had been acquired in ascending mode (~9:00 p.m.). This should overcome both the ionospheric and the tropospheric contamination to a quite satisfactory level. In this work, the ionospheric path delay has been taken into consideration. It is proportional to the number of electrons per unit area in a column extending from the earth surface to the satellite. This columnar electron density count is referred to as the Total Electron Content (TEC). It is measured in TEC units, or TECU. Each TECU equals electrons/m 2. The TEC units as determined in this work are presented in Table 1. Results have been based on the International Reference Ionosphere model ( Table 1: ALOS PALSAR images used in this investigation. Orbit Date Polarization mode TECU /1/2007 Fine Beam Single /1/2009 Fine Beam Single /10/2008 Fine Beam Double 1.6 The zenith ionospheric correction, d ion, is inversely proportional to the square of the radar frequency:

4 d ion k ( TEC) = (1) 2 f where k is a constant and equal to k= m GHz 2 /TECU. The negative sign of k is justified by the fact that a decrease in TEC results in an increase of phase delay [4]. In the case of DInSAR, an interferogram is the difference between two SAR acquisitions acquired at different days, but at the same local time. Therefore, identical TEC values will cancel out the effect of ionosphere in the interferogram. At the L-band frequency of ALOS, with approximately 1.27 GHz, the ionospheric delay in the radar signals amounts to about 25cm for each TEC unit variation in the ionosphere. In other words, a 1 TECU increase between the master and slave images yields a 25cm phase advance, i.e., 25cm should be added to the interferometric values determined. 4. RESULTS The valley to be monitored is densely cultivated and irrigated (wet ground surface). Thus, loss of coherence in images has been observed in the C-Band and could not be overcome. Consequently, interferometric results with the ERS-SAR images have been limited. Decorrelation has been observed for most of the image pairs, spanning time intervals greater than 6 months. In addition, the atmosphere has significantly deteriorated the radar signal for even smaller time intervals. 23/8/1995; 2) 23/8/ /8/1995; 3) 22/8/1995-7/8/1996; 4) 7/8/ /10/1996; 5) 16/10/1996-1/10/1997; and 6) 1/10/ /11/1999. Summation of these interferograms at consecutive periods has aimed at eliminating the atmospheric interference and at providing a land deformation signal covering the entire time interval The use of the tandem image pair taken on 22/8/ /8/1995 and applied in reversed chronological order has been made only for eliminating the atmospheric noise. No land subsidence is expected to be masked using this reversed one-day interval in the stacking chain. As has been mentioned earlier, the atmosphere in that particular linear combination of images should be limited to the first and last images only, specifically the images acquired on 20/6/1993 and 10/11/1999. Note that all individual pairs chosen for processing have been characterized with high values for the altitude of ambiguity (h a ) and small differences in Doppler values (Δdop). Nonetheless, consistent loss of coherence has been evident in the broader area of interest. The signal has been enhanced in a couple of regions (see the areas indicated by the white arrows in Fig.3). However, for these features, an atmospheric perturbation at the acquisition time of the first image acquired on 20/6/1993 has been the most probable explanation for it. On the other hand, the longer wavelength of ALOS L- band has overcome the problem of incoherence caused by wet ground surface because of irrigation and vegetation at the valley. Even for long time intervals, the land subsidence has been very well mapped on the final interferograms with ALOS imagery (Fig. 4). The noticeable fringes shown in Fig.4 cannot be attributed to topographic artefacts. Because they appear in both interfograms, regardless of the different values for the height of ambiguity (h a = -585m and -97m, respectively). In ALOS PALSAR data, one full colour cycle corresponds to λ/ cm of surface displacement and along the line of sight (LOS) between the ground target and the satellite. Figure 3. The ERS-SAR pseudo-interferogram corresponding to the period 20/6/1993 and 10/11/1999. Fig.3 presents the linear combination (in this case the wrapped sum in complex domain) of six ERS SAR images. Those cover the following periods 1) 20/6/1993-

5 MASTER SLAVE h a (m) ΔdopPRF 14/1/ /10/ Figure 5.Unwrapped interferogram for the two year time span, overlain by the drainage system and the main small towns of the area. MASTER SLAVE h a (m) ΔdopPRF 14/1/ /1/ Figure 4. ALOS PALSAR processing has mapped quite well the land subsidence occurring at the Messara valley even for long temporal baselines. Unwrapping of the 2-year interferogram ( ) (as shown in Fig.5) has resulted in a land subsidence of ~8cm/yr along the direction of the line of sight. Taking into account the 0.1 decrease in TEC units between the two image acquisitions (see Table 1), a path delay for ionosphere equals to 2.5cm. Consequently, this value should be subtracted from the unwrapped values presented in Fig.5. The highest rate of subsidence is located close to Praitoria basin area and is of the order of ~5cm/yr. It is clear that the broader area of Messara undergoes strong land deformation. At the last phase, correlation of the DInSAR results using ALOS data, with respect to water pumping has been examined. More particularly, in the framework of INTERISK project, an interregional telematic network for the management of Risks in the Balkan and South Meditteranean area, and the BEWARE-CRINNO project (Crete Innovative region), the Region of Crete has initialized a telematic system consisting of five meteorological stations and 30 more stations for measuring both underground and surface water. These stations are placed in the main water wagons of the island. Fig. 6 shows the groundwater level fluctuations recorded below the station placed at Praitoria area. The station is established at an altitude of 44.91m above sea level surface. The general decline of ground water level is easily observed. Figure 6. Groundwater level fluctuations (shown in meters) recorded at the Praitoria station and for the period It appears (Fig. 6) that the level of water table keeps dropping during the last 4 years to a considerable amount. In particular, a worrying decrease of about 18 m has been marked during the period October 2006

6 October The rate of the decrease is considerably high during the hydrological cycle May-October, i.e., during the summer seasons. As a result of this, the Region of Crete has decided upon the rejection of all the applications requesting new water wells in the area. Similar dramatic data can be found in annual reports given by the Region of Crete. It suffices to say that: 1) in the case of Pompia, a drop in the level of water table by 11m has been observed over the period October October 2007 and by 19m over the period October October 2007; 2) at the Moires station, a decrease of about 4 m has been recorded during the period October 2006-October 2007 and of 13m over the period October 2005-October 2007; 3) a lower decrease of 1.5m has been marked at the Tympaki station during the period October 2005-October 2007; 4) at Asimi, a 11m decline of the groundwater s level has been observed over the period October 2004-October 2005; and 5) on October 2005, the sensor s readings at the Vagionia station indicated that the groundwater level has reached the drilling floor. Undoubtedly, the estimated drop to such low levels raises serious concern over the sustainability of water resources in the area. To invert the situation detected, special actions and measures should be taken immediately. 5. CONCLUSIONS The continuity of ERS missions for almost 15 years provides a unique data archive for monitoring slow land deformations caused by over-pumping water resources. However, the short wavelength (~ 5 cm) of ERS signals brings about a temporal decorrelation in the images over long time intervals. Also, these radar signals cannot penetrate extreme atmospheric conditions. Preliminary results of classical DInSAR and stacking ERS SAR images has indicated that the C-band has not been capable of mapping the land deformation at the Messara valley. Atmospheric corrections using in situ measurements of temperature, pressure and humidity have not been carried out. This is because the coherence levels have been extremely low, even for interferograms spanning from a few months to less than a year. On the other hand, the ALOS PALSAR images and its unique capability to penetrate vegetation canopy seems to have mapped quite well land subsidence. An improvement in temporal correlation has been far more than obvious. However, atmosphere is still an issue that should be further investigated. First, the TEC values determined here account for a zenith ionospheric delay. These values should be recomputed in the radar s line of sight. Tropospheric interference should also be examined. Lastly, geological data should be investigated to justify the differences between the relatively small land deformation revealed by DInSAR in the case of Asimi and Pompia basin areas and the respectively high groundwater level decrease recorded over the period Plans for future research include processing of more ALOS PALSAR images, at variant time intervals and polarization modes. Two Global Navigation Satellite Systems reference stations will be established in the broader area of interest to monitor independently land subsidence with millimeter accuracy. The first GNSS station will be placed close to the region exhibiting the highest subsidence rate, i.e., the Praitoria or the Moires basin, while the second one will be placed in an area of general equilibrium, ~20 km northern of the valley. Acknowledgements. This work has been supported by the FP7-REGPOT , Project No (SOFIA), sponsored by the European Commission. The ERS SAR and ALOS PALSAR images have been provided by the European Space Agency- ESA in the framework of Category-1 Scientific Research Project C1P REFERENCES 1. Amelung, F., Galloway, D.L., Bell, J.W., Zebker, H.A. & Laczniak, R.J. (1999). Sensing the ups and downs of Las Vegas: InSAR reveals structural control of land subsidence and aquifer-system deformation. Geology 27, Bawden, G.W., Thatcher, W., Stein, R.S., Hudnut, K.W. & Peltzer, G. (2001). Tectonic contraction across Los Angeles after removal of groundwater pumping effects. Nature 412, Galloway, D.L., Hudnut, K.W., Ingebritsen, S.E., Phillips, S.P., Peltzer, G., Rogez, F. & Rosen, P.A. (1998). Detection of aquifer system compaction and land subsidence using interferometric synthetic aperture radar, Antelope Valley, Mojave Desert, California. Water Resour. Res. 34, Hanssen, R.F. (2001). Radar Interferometry: Data Interpretation and Error Analysis, Springer, pp , 187, Massonnet, D. & Feigl, K.L. (1995). Discrimination of geophysical phenomena in satellite radar interferograms. Geophys. Res. Lett. 22(12), Schmidt, D.A. & Bόrgmann, R. (2003). Timedependent land uplift and subsidence in the Santa Clara valley, California, from a large interferometric synthetic aperture radar data set. J. Geophys. Res. 108(B9), 2416, doi: /2002jb Tomás, R., Márquez, Y., Lopez-Sanchez, J.M., Delgado, J., Blanco, P., Mallorquí, J.J., Martínez,

7 M., Herrera, G. & Mulas, J. (2005). Mapping ground subsidence induced by aquifer overexploitation using advanced Differential SAR Interferometry: Vega Media of the Segura River (SE Spain) case study. Remote Sensing of Environment 98, Xepapadeas, A. (1996). Quantity and quality management of groundwater: An application to irrigated agriculture in Iraklion, Crete. Environmental modeling and Assessment 1, Zhou, X., Chang, N. & Li, S. (2009). Application of SAR Interferometry in Earth and Environmental Science Research. Sensors 9,