P. K. Champati ray1, Kenny Foreste2 and P. S. Roy1
1Indian Institute of Remote Sensing
(NRSA, Dept. of Space), Dehradun, India
Email: [email protected] On the 26th January 2001 Kachchh region of Gujarat State was struck by a devastating earthquake of magnitude 7.7 (Ms). The epicentre was located by USGS towards north east of Bachau (23o36/N, 70o34/E) approximately along the Kachchh Mainland fault. This earthquake is one of the most deadliest to strike India in its entire history, the death toll is over 20,000 and the number of injured is up to 166,000, more than 600,000 people were left homeless, with 348,000 houses destroyed and 844,000 damaged. This event has attracted attention from all over world and many scientists, engineers and experts have visited the region and studied various aspects of the earthquake. The IIRS team along with students from Korea and France visited the region to understand the role of remote sensing and GIS in hazard assessment and mitigation. An important lesson was learnt- although prediction of earthquake remains an elusive goal, the affected areas and destruction can be known within minutes of knowing the epicenter and magnitude using predictive models. Just after the event it is very critical to know what are the areas must have been affected and at this point of time any information towards this direction is vital. Therefore, while assessing damage at the preliminary stage emphasis is laid on how fast this information can be generated and accuracy issue remains secondary due to the fact that many critical parameters for the model cannot be known immediately. In the second step results of such models can be improved and it be applied for microzonation purposes.
Remotely sensed data analysis
In the present study the pre and post event data sets of IRS PAN and LISS III were analysed in order to detect surface changes induced by the Bhuj earthquake. These four images were geometrically corrected with reference to 1:250,000 SoI toposheets. In order to highlight the changes, following image processing methods have been used.
- Image difference- the subtraction of the post event data by the pre event data shows the changes after the earthquake. This processing was used for LISS III data in order to highlight changes at regional scale (Figure 1 and 2).
- Special band combination- it allowed to display satellite image in False Colour Composite (FCC) using different band combination giving another possibility to see surface changes. This processing was used for both PAN and LISS III data.
- Combining PAN data with multispectral LISS III data- By using Intensity-Hue-Saturation (IHS) transformation technique, high resolution (5.8 metres) PAN data could be combined with the multispectral (4 spectral bands) LISS III data. This processing was used to see surface changes at locale scale (construction, infrastructure and surface deformation).
- Supervised classification- It was used to classify and highlight the saline liquid intrusion in the Great Rann of Kachchh due to liquefaction (Figure 3).
- Principal Component Analysis (PCA)- PCA was carried out on LISS III data (single date and multidate) to see the liquefaction areas and increase in the moisture content of the soil after the earthquake (Figure 4).
Using image processing on pre- and post- LISS III data sets it was possible to delineate precisely earthquake induced liquefaction areas indicated by liquid emanation. Principal component analysis has demonstrated that increase of soil moisture can also be mapped, one such area is observed around the USGS epicentre zone (Figure 4). Further using image difference, increase in turbidity level due to the earthquake in the Gulf of Kachchh (Figure 5) could also be detected. This increase in turbidity level is mainly due to lateral spreading in sea floor due to liquefaction in the saturated soil. Damage to buildings and surface deformation were also detected using image processing on PAN data sets. For preliminary damage detection a special band combination was used where pre-earthquake data was kept in red channel and post earthquake in green and blue channel, damaged buildings appeared in red colour. Further this preliminary detection was verified using PAN data merged with LISS III data by IHS transformation. It was also possible, using PAN data, to detect an important surface deformation around Budharmora, which appears like a linear feature running for about 1 km. On ground it was verified as a flexure with a height of about 0.5 to 1 m, probably caused by a blind fault as it happened during 1819 earthquake along the Allha Bund fault. This linear feature is very significant as it runs along the Kachchh Mainland fault and is situated very close to the epicentre.
In the field, different types of surface features were observed mostly associated with liquefaction such as lateral spreading, slumping, sand boils (mud fountain), loss of bearing strength and uplifting. Many cracks on the ground indicate normal faulting with northern block going down compared to the southern block. Near Manfara village a large right strike slip fault was observed which has caused surface rupture for several kilometres (Figure 6 & 7). All these observations are plotted on LISS III image and it shows that Quaternary deposits, Rann and mud flats were more affected by surface deformation and liquefaction, and overall subsidence and movement were observed towards north (in direction to the Great Rann), due to faulting and movement in the direction of the Great Rann.
GIS based analysis
Ground motion due to earthquake is characterised by the Peak Ground Acceleration (PGA), which can be generated in the form of GIS-based class maps. The spatial distribution of ground motion can be determined using a deterministic ground motion analysis. For a given event magnitude, attenuation relationships are used to calculate ground shaking demand for rock sites which is amplified by factors based on local soil conditions. The attenuation relationships used in the present study are based on Joyner and Boore (1981) formula that was modified for this region. The final PGA values calculated for the area using modified equation are cross-checked with actual measured values and show good correlation. This ground motion is the basis to estimate the ground failure associated with an earthquake. In the present study, probability of liquefaction and Permanent Ground Displacement (PGD) due to lateral spreading were calculated. The peak ground acceleration was converted to Modified Mercalli Intensity Scale using the Trifunac and Brady (1975) Equation, based on which damage to Pucca and Kachha houses were predicted.
Liquefaction is a soil behaviour phenomenon in which a saturated soil looses a substantial amount of strength due to high pore-water pressure generated by and accumulated during strong earthquake ground shaking. The relative susceptibility ratings indicate that recently deposited relatively unconsolidated soils such as Holocene-Age river channel, flood plain, delta deposits and uncompacted artificial fills located below the groundwater table have high to very high liquefaction susceptibility. Sands and silty sands are particularly susceptible to liquefaction. Silts and gravels are also susceptible to liquefaction, and some sensitive clay has also exhibited liquefaction-type strength losses.
Permanent ground displacements due to lateral spreads or flow slides and differential settlement are commonly considered as significant potential hazards associated with liquefaction. The initial step of the liquefaction hazard evaluation is to characterize the relative liquefaction susceptibility of the soil/geologic conditions of a region or sub region. Geological map units and general soil condition, based on these characteristics, characterize susceptibility; a relative liquefaction susceptibility rating (e.g., very low to very high) was assigned to each surface geological unit.
The susceptibility of the soil, the amplitude and duration of ground shaking and the depth of groundwater primarily influence the likelihood of experiencing liquefaction at a specific location. It is recognised that in reality, natural geologic deposits as well as man-placed fills encompass a range of liquefaction susceptibilities due to variations of soil type (i.e. grain size distribution), relative density, etc. Therefore, portions of a geologic map unit may not be susceptible to liquefaction and this should be considered in assessing the probability of liquefaction at any given location within the unit. In general, we expect non-susceptible portions to be smaller for higher susceptibilities. This “reality” is incorporated by a probability factor that quantifies the proportion of a geologic map unit deemed susceptible to liquefaction (i.e., the likelihood of susceptible conditions existing at any given location within the unit). The likelihood of liquefaction is significantly influenced by ground shaking amplitude (i.e. peak horizontal acceleration, PGA), ground shaking duration as reflected by earthquake magnitude, M, and groundwater depth. Thus, the probability of liquefaction for a given susceptibility category can be determined by the following relationship:
Relationships between liquefaction probability and peak horizontal ground acceleration (PGA) for each susceptibility class are defined based on the state-of-practice empirical procedures as well as the statistical modelling of the empirical liquefaction catalog for representative penetration resistance characteristics of soils within each susceptibility category. The high probable areas (mostly in Rann and Kandla region) on the resultant Liquefaction Probability map correspond to higher incidence of liquefaction phenomena observed on satellite image and on ground due to the recent earthquake.
The expected permanent ground displacements due to lateral spreading can be determined using the following relationship:
This relationship for lateral spreading was developed by combining the Liquefaction Severity Index (LSI) relationship with the ground motion attenuation relationship and modified for this study area. The ground shaking level has been normalised by the threshold peak ground acceleration PGA(t) corresponding to zero probability of liquefaction for each susceptibility category. The resultant Ground Displacement Class Map indicates that permanent ground displacement could be in the range of 0.5m to 3.0m which shows good correlation with most of the ground observations.
Prior to instrumental observation and the introduction of the magnitude scale M, “size” of an earthquake was measured by its effects on the surface by a qualitative scale of intensity of earthquake. The intensity at a point depends not only upon the strength of the earthquake (magnitude) but also upon the distance from the epicentre to the point and the local geology at that point. It is, in a sense, a qualitative expression of the ground movement and the resulting damage caused by the earthquake. Using the Trifunac and Brady relationship intensity map in MMI scale was prepared, based on which damages to Pucca houses and Kachcha houses were predicted for each village and results were cross checked using UNDP damage data for part of the Kachchh region. However, in this estimation, it must be kept in mind that the damage estimate for Kachha and Pucca houses are taken for equivalent building categories given in the UN Radius Programme Manual, therefore, some deviation is expected, further there exists wide variation in construction in different parts of India. Therefore, the results of such studies are indicative only and should be correlated to actual damage with ground verification.
The study has highlighted that remote sensing (at the present level of technology) can be used as a tool for broad assessment of the damage and liquefaction phenomena associated with Bhuj earthquake. The detection of damage to large buildings shows that it is only matter of resolution which stands on the way to damage assessment in our thickly populated clumsy urbane areas. However, using some kind of statistical measures, damage can be assessed in urbane areas within very short time period, which can be very useful in carrying out relief measures. The GIS-based analysis indicates that within minutes of knowing the epicentre location, it is possible to calculate the PGA, liquefaction probability and ground displacement. Although the accuracy of these maps can be debated, the aim is not to produce very accurate precise maps, but to provide some kind of spatial information basis for giving a certain direction to mitigation and relief measures during the critical time.
At least in the present case it is demonstrated that we could have anticipated the damage in areas close to Bachau, Manfara, Chobari, Amarsar etc. within minutes, provided a soil map or a surface geology map was available in digital format (which can be easily done). During relief operation, the time taken to send medical aid and other help to proper places was basically the time lost in the whole operation. Such type of GIS based analysis could have saved some time and probably many limbs and lives. Apart from this, the methodology adopted in the present study can be utilised for microzonation purposes which is the need of the hour considering the fact that many villages and towns are to be rebuilt in relatively safe areas.