Flood Inundation Prediction Model on Spatial Characteristics with Utilization of OLAP-based Multidimensional Cube Information

Transcription

1 , pp Flood Inundation Prediction Model on Spatial Characteristics with Utilization of OLAP-based Multidimensional Cube Information Ji-Hoon Seo, Yoon-Ju Lee, Hye-Jin Jo, Jin-Tak Choi Incheon University, 119 Academy-ro, Yeonsu-gu, Incheon, Republic of Korea Abstract The escalation of IBS (Intelligent Building System) by convergence technologies is currently in progress with the rapid growth of data communication technology and advanced modern construction system, and on this basis, new high-density urbanization is formed. The severe problem of the formation of high-density cities is leading to the phenomenon of spatial collapse by geographical deformation, global warming and destruction of the ecosystem by climate change. Notwithstanding the fact that high accuracies are shown in the past data in terms of information on precipitation, serious vulnerabilities are shown with respect to flood preparation on the subject of natural disasters by the anomaly climate, which tends to produce an exceedingly low level of accuracy, and consequently, adverse effects by which the loss of nations assets and human damage are caused came about. This thesis, to bring about improvement to the drawbacks addressed, suggests a design modeling of water level prediction in relation to regional flood by constructing an OLAP-based multidimensional cube with the use of historic regional precipitation and water level data. The historic precipitation and water level data by region design the prediction model of flood inundation by extracting the pattern of the water level information after topographical relations and intelligent buildings have been increased through the extraction of multidimensional table modeling and construct of the data warehouse for the analysis of the source data via pre-processing of documents. Keywords: Climate change, multidimensional cube, pattern algorithm 1. Introduction High-tech systems with integration of convergence technology are making progress with the increase of science technologies and discoveries on technologies accumulated, and in spite of the fact that new urbanization has increased considerably on the basis of U- CITY with emergence of intelligent buildings, the destruction of the ecosystem caused by climate change and topographical changes is emerging as a serious problem. Therefore, vulnerabilities on flood preparation on a national level with respect to natural disasters are shown and damages and losses involving human and properties are being caused. These human damages are different with the reasonable factor which implies that thinking can be minimized by removing insecure behaviors in advance and factors in relation to human errors attributable to human s negligence and errors suggested in Herbert William Heinrich s domino theory. That is to say, in addition to technological developments, patterns of the fundamental cause of mishaps in relation to natural disasters by climate change suggested as side effects is required to be predicted, and a considerable relationship is being maintained in geographical changes. Despite the fact that high-tech infrastructures of the city environment with respect to the prediction of flood inundation and water supply system for covering flood inundation which exercise influence to the geographical transformations are developing in conjunction with the high-density ISSN: IJSEIA Copyright c 2015 SERSC

2 phenomenon of new cities, the risk factor patterns of the water level is presently showing a parallel. In other words, Comparative analysis by geographical changes and data deduction via precipitation patterns are required in the current flood inundation and regional flood inundation of the past. There are limitations to predicting the water level of artificial spatial terrains in which today s new cities are erupted with measurement data of the water level for each region which were produced spontaneously before the creation of the new cities. Therefore, this thesis has drawn a conclusion on the patterns in relation to the water level data gained after new urbanization was formed with the use of the current artificial requirements and water level data of the past gained in relation to the state prior to topographical changes, and topographical flood disaster prediction model design using the OLAP(Online Analytical Processing)-based multidimensional tables was created by constructing multidimensional tables and preprocessing of the precipitation data. In this thesis, a data warehouse was constructed by preprocessing the current data and source data of the past, and the normalization of OLAP-based flood prediction dimensional tables was modeled by using the historic precipitation data. Fact tables are created on foundation of dimensional tables designed, and potential inundation for areas where flooding has occurred was analyzed by constructing dimensional cubes. Flood patterns by regional groups were drawn via the multidimensional tables and flood inundation prediction modeling with respect to geographical and environmental changes of the past and present was suggested. 2. Related Work The OLAP(On-line analytical processing) engine by and large shows the structure that suggest the logical scale of the multidimensional data in the form of cubes, and the cubes can be defined into two parts, dimension tables and fact tables, in which the data are modeled into one or more dimensions [1, 2]. The OLAP cube may have varied forms structured from accumulated data values at the data warehouse and can provide accurate numerical values in terms of decision making of the data [3]. In spite of the fact various software capable of conducting simulations on flood inundation levels such as LEC-RAS, FLUMEN, MIKE FLOOD(MIKE 11 + MIKE 21) have been released in studies in the field of environmental engineering and safety today, it has been indicated their prediction accuracies are still at inaccurate levels, for simulators used in these software utilize the data suggested on the foundation of the present and precipitation data of the past had been processed and thus fact data have not been laid on their foundation [4] Infrastructure Vulnerability in Relation to Climate Change According to the 2007 IPCC (Intergovernmental Panel on Climate Change) and various studies done in South Korea and overseas, diverse changes, including the rise of the global average temperature by 0.76 compared to year 1900 attributable to global warming caused by climate change, rise of the mean sea level by 1.8mm per year since 1961 and a 2.7% 10years decline in the North Pole's average sea ice extent since 1978, are being observed. With respect to Korea's six major cities, the regions' average temperature has risen 1.7%, precipitation rate has increased 19% and sea level has risen 8 cm over the last 43 years, thus damages arising from flood and increase of torrential downpour are increasing significantly, and it has been confirmed that over 90% of the cause of the damage made by natural disasters occurred in South Korea are attributable to flood and typhoons [5, 6]. In spite of the fact that high-tech systems are being adopted for such cases with technological developments, it signifies that water supply systems and infrastructure for flood preparation remain vulnerable, and vulnerabilities have been drawn in relation to infrastructure as a result to carrying out analyses on scales of 2 Copyright c 2015 SERSC

3 adaptability, climate exposure, sensitivity via correspondence variables of each region with the use of researches carried out in South Korea [7] Prediction System for Climate Change In spite of the fact that various studies for predicting flood water level by point using neuro-fuzzy systems on the foundation of data are under way for predictions of climate change, methods in which models using multiple linear regression utilizing change of conditions of the weather data and real-time data are suggested do exist. While predictions are possible in such a case by analyzing the data-oriented statistical factors, the negative aspects are that range of regional prediction is restrictive and prediction of locations at which terrains have deformed on the basis water levels of the past is unclear [8, 9]. 3. Proposed Method This thesis has processed the quantitative data made use of statistics in consideration of the characteristic that risk level of annual inundation can be influenced by topographical changes and geographical conditions of concentration areas due to formations of new towns, albeit regular precipitation was recorded in the same region when source data were successively classified with patterns shown in the past one year as standard. However, for anomalous patterns and construct of natural drainage system or river system of flood preparation facilities built by the formation of new towns which induces the flow of water through dispersion, data were calculated with the quantitative approach. The process of the multidimensional table prediction model of this thesis was designed using the procedures below Preprocessing of Precipitation Observation Data For data information, Water Management Information System (WAMIS)'s regional water level and precipitation data were collected, and data were classified by year on the basis of the sample data capable of real-time verification for the analysis of data's portability. Data preprocessing process is an essential factor in this thesis. The preprocessing process is required to place a certain degree of emphasis on reliability in relation to the quality of the data on the basis of a definite purpose toward prediction and usage. Accordingly, the quality of the initial data were enhanced in this thesis to lay focus on the relevance in terms of whether or not the usage of the data is valid and factors of accuracy are accurate. For the preprocessing process, precipitation data of each region were classified by time for the purpose of creating multidimensional tables and filtering process was carried out by assorting the only information necessary for the prediction system. Moreover, in the case of WAMIS, respective data is required to be integrated as the extraction of information on water level and precipitation are carried out separately, and then a filtering process in which the format of the data is modified is carried out. Filtered data are drawn as the training data capable of predicting the initial source data by normalizing the dispersion of the numerical data Processing of Missing Values of the Precipitation Data In order to precisely draw a conclusion for the preprocessing of the precipitation data, missing values and invalid values are required to be filtered. This is due to the fact that precipitation is a scale which represents the occurrence frequency of rainfall in a timeframe of 365 days or 1 year and thus blanks are present in between data tables. The problem is considered to be less serious if the probability of precipitation was higher than the probability of having rainless days during a one year period and blanks exist between precipitation data tables, however, if occurrence frequency was shown with regard to the Copyright c 2015 SERSC 3

4 missing values for the reason that over 10% were rainless days based on the sample precipitation data of the entire 365 days, the problem is required to be solved before reliability of the data can be increased. In addition, taking account of national and regional conditions, regions where levels of precipitation increase dramatically during the rainy season between the month of June and August exist, whereas regions where it is raining all year round also do exist. Such cases remain as data's missing values, and there is a need to maintain the data transformation as integer type. Methods such as removal of all items inclusive of missing values, replacement of missing values with appropriate values and forwarding the process of missing values to the analytical phase can be utilized to settle the missing values. The method in which all items inclusive of missing values are removed from the data is not reliable since over 60% of the entire sample data is required to be removed. The method in which missing values are forwarded to the analytical phase is prone to generating errors in the normalization process of forming multidimensional tables. In this thesis, 0 was assigned as the missing value of the time at which it did not rain. Although zero was assigned as the missing value in principle, it did not necessarily imply of the fact that it did not rain. Although zero will be suggested if the weather information implies cloudiness or slight precipitation, there is no meaning in defining a separate categorical variable. Figure 1. Missing Values of the Precipitation Data Since the application via the construct of the data warehouse (DW) is meaningless if there is no data recorded in the tuple, calculation was carried out with zero as a phase to settle the missing value when precipitation probability was 0% Processing of Invalid Values of the Precipitation Data A range of variables and invalid values exist in the precipitation data. Invalid value represents a value that is beyond common sense, or in a case where positive number values are negative, a faulty input of the data type. As regards the precipitation data, the significance included in the precipitation is required to be realized, and attribute and type are required to be classified distinctively. The semantics of precipitation is comprised of a variety of information, such as rain, hail, Scotch mist, snow, sleet, snowstorm and et cetera, for it signifies all coagulating matters, including liquid and solid matters which 4 Copyright c 2015 SERSC

5 reach the ground from the air. Therefore, distinction is made between liquid type precipitation and solid type precipitation, and since solid type precipitation is also measured with the depth of the water dissolved, it is also included in the precipitation for observation. However, since hail and snow are factors that do not affect flood inundation to a great extent, a procedure for the classification of the attributes of the weather was carried out. Figure 2. Invalid Value Filter Processing of Outlier of the Precipitation Data In this thesis, outliers were considered in relation to climate change. An analysis which can explain the changes in the precipitation levels in comparison with the identical hour of the identical day of the previous year is required, and through such an analysis, outliers can be detected for comparison with the regional water level data Data Analysis Process For a comparative analysis of the precipitation data with information on water level inundation, regional locations are required to be selected, selections need to be classified into the past or current data and identical formation patterns need to explored for classification as a factor for the construct of the multidimensional modeling. The data analysis process in thesis is as follows. STEP 1 Sorting of the identical regions and assortment by calculation of the past meteorological information. STEP 2 For meteorological information, time, date and region were classified by using the clustering method and dimension tables and fact tables are calculated. STEP 3 Normalization of object entities and creation of multidimensional cubes. STEP 4 Calculation of prediction value of precipitation scale via analyses of pattern data utilized multidimensional cubes 3.3. Structuralization of the Data Warehouse for Data Prediction The operating procedure model of the data warehouse has the structuralization capable of predicting precipitation data via the analyses of external data. Storage space for saving Copyright c 2015 SERSC 5

6 information regarding geographical changes corresponding to the data of the existing region is required to be maintained to be able to utilize the historical data on precipitation, and to this end, a scale for the process of quantitative data regarding inundation level to geographical conditions of the past and the present is required. In this thesis, MIKE FLOOD has been used for the data design on quantitative spatial geography model of the past and the present. Figure 3. Model of Structuralization Process of Data Warehouse ETL is used for the purpose of extraction, transformation and loading of the data and filtering and storing the initial data, which are the precipitation data and water level data of the past, and is managed by the disk sharing method. Figure 4. As is and to-be Models on Geographical Changes 6 Copyright c 2015 SERSC

7 3.4. Design of Multidimensional Prediction Model The structuralization of the data warehouse was constructed with star schema, one of the methods of relational modeling. For each category, dimension tables of position data, dimension tables of weather data, dimension tables of precipitation data and dimension tables of time were used for the modeling of fact tables of the multidimensional space. Figure 5. Definition of the Multidimensional Cube Model on the Precipitation Forecast Star schema and snowflake schema are used for the multidimensional modeling, and each has its own advantages and disadvantages. In this thesis, a modeling was constructed on the basis of the directivity of the speed occurred through the minimization of the join and simplification of the model. Snowflake schema minimizes data redundancy and can increase the pliability of the process; however, modeling factors are complex and generates factors that decrease the speed. The downside of star schema is that there is an inconsistency in the data and the process of the redundant data is indefinite, but reliability of the data had been secured via the filtering process. 4. Performance Evaluation In this thesis, an international new town corresponding to Region I distance S in South Korea was selected, and prediction test of the precipitation data was carried out based on a region of an area of 93 m3. Copyright c 2015 SERSC 7

8 Figure 6. Precipitation Forecast Information of Region I during the Rainy Season The testing has calculated the prediction information of precipitation to the cube modeling of the summer season which tends to show higher precipitation levels due to meteorological nature of South Korea by extracting date information, weather information and regional information on the basis of partial data of the past three years applicable to the cube data. The graph being suggested in [Figure 6] is the information on the partial prediction of precipitation calculated from the cube model. The X value represents the date information and the Y value represents the prediction data of precipitation on the season. With regard to the forecast information of precipitation, forecast information can be assigned by applying an arithmetic operation to the accumulated flood inundation data of the past with OLAP multidimensional data models. The flood data suggested in this thesis can differentiate the scales regarding the degree of the increase of the water level in relation to topographical changes of the past by comparing recent water levels to quantitative data on water levels recorded for the identical region. However, with data on water level alone, there are limitations to quantitatively generating data on flood inundation which changes according to the topography of the present and the past. This is mainly due to the fact that diverse variables are present depending on the construction rate and utilization of water disposal-based infrastructures after the formation of the city and topographical factors before the formation of city. In order to quantify and increase the accuracy of such variables, a modeling is required to be suggested by supplementing a more detailed fact values on the foundation of data on discharge, flow velocity, installation drawings of drains and water supply systems installed across the city. 5. Conclusion This thesis has designed a prediction model of precipitation water level inundation via the multidimensional array modeling by extracting spatial and topographical information on deformation areas and precipitation data accumulated through historic observations. Normalization via the preprocessing process was accomplished on the basis of water level and precipitation data and a multidimensional table model was suggested via the data warehouse. The study further aims to conduct researches on the development of integrated systems for the prediction of precipitation based on spatial data in which information about rate of discharge, water disposal system for the increase of accuracy of geographical information, schema transformations for the increase of precision of normalization and data extension via the extension of fact tables can be additionally applied. 8 Copyright c 2015 SERSC

9 Acknowledgement This research was supported by Basic Science Research Program through the National Research Foundation of Korea(NRF) funded by the Ministry of Science, ICT and future Planning(NRF-2014R1A2A1A ). References [1] Pilot Software, An Introduction to OLAP: Multidimensional Terminology and Technology, [2] S. Agarwal, R. Agrawal, P. M. Deshpande, A. Gupta, J. F. Naughton, R. Ramacrishnan and S. Sarawagi, On the Computation of Multidimensional Aggregates, Proc. of VLDB Conference, (1996), pp [3] J. Chen and A. A. Hill, Modeling urb an flood hazards: just how much does DEM resolution matter, Applied Geography Conference, vol. 30, (2007), pp [4] S. Patro, C. Chatterjee, S. Mohanty, R. Singh and N. S. Raghuwanshi, Flood inundation modeling using MIKE FLOOD and remote sensing data, Journal of the Indian Society of Remote Sensing, X SCI(E), vol. 37, no. 1, pp [5] A. Calver, J. Crewett, H. Davies, R. Lamb and S. Crooks, Modelling Floods from Combined Surface and Subsurface Sources, Ministry of Agriculture, Fisheries and Food, FD0425, pp. 78. [6] J. F. Girard and J. M. Baltassat, Interreg III A, Flood1: The Role of Groundwater in Flooding Events, BRGM RP FR, pp. 76. [7] G. B. Kim, Y. C. Son, S. H. Lee, A. C. Jeong, E. J. Cha and M. J. Ko, Understanding of surface watergroundwater connectivity in an alluvial plain using statistical methods, The Journal of Engineering Geology, vol. 22, no. 2, pp [8] J. H. Lee, S. Y. Hamm, C. M. Lee, J. J. Lee, H. S. Kim and G. B. Kim, Numerical simulation of groundwater system change in a riverside area due to the construction of an artificial structure, The Journal of Engineering Geology, vol. 22, no. 3, (2012), pp Authors Seo-Ji Hoon, received his Bachelor degree in 2008 at Seoul National University of Science and Technology from department of Safety Engineering. He finished his MS and Ph.D. at Incheon National University from Department of Computer Science and Engineering in 2010 and 2015 respectively. His research interest includes Data Mining, Database Management and Sensor Networking. Yoon-Ju Lee, received her Bachelor degree in 2013 at Incheon National University from department of Computer Science and Engineering. Her research interest includes Database, Big Data and cryptography. Hye-Jin Jo, received her Bachelor degree in 2014 at Incheon National University from department of Computer Science and Engineering. Her research interest includes Database, Big Data and cryptography. Copyright c 2015 SERSC 9

10 Choi-Jin Tak, received his B.S. degree in Mathematics and his M.S. degree in Computer Science from Dongkuk University, He received Ph.D. degree in Electronics from Kyunghee University, he has been a Faculty Member at the Dept. of Computer Science of Incheon National University. His research interests include cryptography, database systems 10 Copyright c 2015 SERSC