PhD Defence Mr Saibal Ghosh
Dept. of Earth Systems Analysis
Title of Defence
Knowledge-guided Empirical Prediction of Landslide Hazard
Every year during monsoon large parts of the Indian mountains experience a substantial loss of properties and lives due to various types of landsliding events. For mitigating this disaster, predictive maps of landslide hazard and risk, preferably at medium scales (1:25,000 to 1:50,000) provide vital geo-information to the administrators/planners. Landslide hazard maps that are currently available in India are, at best, qualitative landslide susceptibility maps based mainly on the heuristic guidelines for quickly assessing large areas where landslides can occur. Due to the absence of suitable methods pertaining to the Indian geo-environment (e.g. the Himalayas), translating those maps into the actual expected impacts of landslides is a difficult task, which precludes the subsequent hazard and risk analysis. Therefore, the prime aim of this research is to propose an effective method for medium scale landslide hazard and risk analysis suitable for the Indian geo-environment which can be readily used by the public research Institutes/Organisations engaged in landslide research in India. The proposed method considered the variability and complexities of the one of the predominant landslide-prone terrains (e.g., the Darjeeling Himalayas) and the landsliding processes prevalent there, so that the aspects of variable controls of landsliding factors can be better understood and the potential adverse impacts of landslides to life and property can suitably be minimised. This proved challenging because landslide source data sets in India like many other countries are usually scarce and/or filled with incomplete information.
Predictive mapping of landslide susceptibility still remains to be the first and foremost step for any hazard and risk analysis. This research demonstrated that due to variable terrain conditions and different landsliding processes, predictive mapping of landslide susceptibility based only on heuristic/subjective guidelines, where a specified number of factors and/or specified factor-weights are considered could be of limited use. In contrast, the empirical methods of landslide susceptibility can perform better and can easily be transformed into maps that are able to portray the actual impacts of landslide events.
The empirical methods (e.g., statistical or mathematical) of landslide susceptibility mapping depend on two types of spatial associations: (a) spatial associations of individual spatial factors with historic landslides of a certain type; and (b) the relative importance of individual spatial factors with respect to one another in relation to landslide occurrences. Although, multivariate statistical methods can simultaneously model both the above two types of empirical spatial associations, yet some of the spatial factors selected and used by them according to certain (statistical) criteria are sometimes not representative of specific genetic processes associated with the type of landslides being studied. Whereas, bivariate statistical methods, although intuitive, can only model the first type of spatial association, and thereby, exhibit moderate prediction rates. Therefore this research demonstrates a suitable empirical method to study the spatial association to (a) select the most appropriate spatial factors that approximate realistic genetic associations with landslides of a certain type, and (b) objectively determine the importance of every spatial factor with respect to the others in relation to the landslide type under study and combine them using the multivariate methods. The proposed empirical technique can interactively and iteratively augment the empirical model processes that estimate susceptibility, which is subsequently used to prepare the landslide hazard and risk maps.
Scarcity of information on past landslides acts as a serious deterrent in correctly predicting the landslide hazard scenarios and thereby affects the quantitative risk assessment because quantitative landslide hazard and risk analysis requires the availability of sufficient historical landslide information in order to correctly estimate the spatial, temporal and magnitude probabilities. This could be another main reason why landslide scientists in India rely more on susceptibility maps to do a qualitative landslide impact assessment. Despite the above unavoidable constraints, this research proposes suitable methods for the determination of temporal and magnitude probabilities of landsliding events by generating and using event-based landslide inventory maps prepared with the available source datasets. Thus, this research presented a method for quantitative landslide hazard and risk analysis, by generating a series of landslide hazard scenario maps based on the magnitude of triggering events and landslide event-days from a continuous record of 40 years (1968-2007), and to study and analyse their corresponding variation in landslide density and temporal likelihoods. Based on the above plausible landslide hazard scenarios, this research presents suitable methods of medium-scale (1:25,000) landslide risk estimation with varying levels of uncertainty in expected loss assessment to buildings, population, and roads. The varying levels of uncertainty in such loss assessment were caused due to the obvious limitations in the past landslide and its damage data. With the above, the present research addresses the fact that in literature, not much research on landslide hazard and risk estimation in a data-scarce environment is available, where varying levels of uncertainties that are expected to propagate in quantitative landslide hazard and risk assessment are shown.
This research recommends that the predictive mapping for landslide susceptibility which remained to be the foremost step of any landslide hazard and risk analysis should be based on quantifying the spatial associations between spatial factors and the landslide type under study instead of relying solely on the heuristically or subjectively identified causal factors and its pre-defined weights and further, to lessen the uncertainty in hazard and risk estimation, this research demonstrates that maintaining a continuous landslide event database for use in future are of prime necessity.
Saibal Ghosh was born on 15th December 1967 in Kolkata, India. He completed his higher-secondary level (10+2) education from the Hare School, Kolkata, India in 1986. He did his graduation (Bachelor of Science) with a major in Geology from the Presidency College, Kolkata, India in 1989. He completed his Master of Science (M.Sc.) degree in Applied Geology in 1991 from the Indian Institute of Technology (IIT), Kharagpur, India and subsequently completed his Master of Technology (M.Tech.) degree in Applied Geology in 1993 from the same institute after successfully qualifying the Graduate Aptitude Test in Engineering (GATE) in 1991. He joined as a Geologist in the Geological Survey of India (GSI) on 14th June 1994 through UPSC’s Indian Geologists’ Examination and is still working there presently as a Senior Geologist. During his Masters level, he did his dissertation on the evolution of recent sedimentary structures in a deltaic environment (M.Sc.) and on the sedimentary evolution of the Gilbert-type deltaic formations of the Lower Gondwana rocks in the Talchir Gondwana Basin, Orissa, India (M.Tech.) respectively. In GSI, he has been associated with a number of hydroelectric projects in the north-east and eastern Himalayas of India (Arunachal Pradesh, Sikkim, Darjeeling) as a consultant engineering geologist. Apart from taking up geological investigations in various such engineering geology projects, he has also been engaged simultaneously in various regional and site-specific landslide investigations in Darjeeling-Sikkim Himalayas since 2001. In 2007, he has been selected by GSI as one of the research-fellows to pursue the present Ph.D. research on landslide hazard and risk at ITC, The Netherlands, under an Indo-Dutch Joint Research Agreement between ITC, GSI and NRSC. He is married and has one son.
Ghosh, S., Jetten, V.G. (Promotor) , van Westen, C.J. (assistant promotor) and Carranza, E.J.M. (assistant promotor) (2011) Knowledge guided empirical prediction of landslide hazard : e-book. PhD thesis University of Twente; summaries in Dutch and English. ITC Dissertation 190, ISBN: 978-90-6164-310-4.
|Event starts:||Tuesday 05 July 2011 at 13:30|
|City where event takes place:||Enschede|
|Country where event takes place:||Netherlands|