Additional elevation information were collected to fill gaps in the available data and to update terrain information due to recent and future developments. Other necessary data for instance Landuse or Landcover, Rainfall, flood depth and extent and flood risk perception are needed in flood modeling and development impact assessment. This study is divided into 4 main phases, namely, 1) Data preparation and analysis, 2) DTM and DSM modelling, 3) flood model calibration and modelling and 4) development impact assessment (see figure 3.1). The first phase focuses on elevation data preparation, analysis and integration. The second phase of the research methodology concerns on the construction of DTM and DSM based on different sources of elevation data, which were derived from both primary and secondary sources. The primary data collection aims at filling the gaps in the available data and to update terrain changes in the study area as a result of recent developments. In general, the elevation data is derived in various forms, for instance points, line and polygons, and these data are then aggregated into ground terrain and man-made features. Further aggregation is made to separate the elevation data into two terrain situations; current and future situations. The DTMs are produced using different terrain interpolation methods. The best product is selected and integrated with man –made features to produce DSMs of the study area. The 1D2D SOBEK flood model is used to simulate 5 recurrence intervals flood events. The flood calibration is made base on flood depth information derived from recent field observations (this data were collected by Saut Sagala and Peters Guarin Graciela) after the flood event caused by the Super typhoon Nanmadol (with an equivalent to 10 years return period flood). The flood calibration is based on two aspects; surface roughness and building structure. The calibrated surface roughness and the suitable building representation will be used for further flood modelling. The surface roughness value of the study is based on landuse or landcover. This updated information is used together with development plans to create recent and future landuse or landcover in Naga City. The final phase emphasizes on development impact assessments based on detailed investigation of flood characteristics before and after the development. This is supported by additional assessment focuses on changes of flood hazard areas for current and future situation of Naga City. The definition of flood hazard is based on the combination of flood velocity and flood depth.  Figure 3.1: Overall research methodology 3.1 DTM and DSM construction DTM and DSM of the study area were generated in 4 major steps; 1) elevation data preparation and analysis, 2) elevation data interpolation, 3) accuracy assessment and reporting and finally 4) integrating natural terrain with man made terrain to produce DSM. 3.1.1 Elevation data preparation and analysis The elevation dataset for DTM generation is derived through the integration of various elevation data sources and these data vary in both horizontal and vertical accuracies (see appendix 1). The arrangement of elevation data will influence the shape of the variogram model. In this case, the clustered elevation data, which were found around dunes, road embankments and other local abrupt changes in topography, were removed from the dataset (Blomgren, 1999). Wilson and Atkinson (2003) in their research, “Prediction the uncertainty of DEM on flood inundation modelling”, used the ordinary kriging to interpolate the elevation data in the floodplain area. The elevation data was the combination between the contour lines and the elevation points that were derived from GPS measurement. The original experimental semi-variogram of the contour lines had quite general shape. However this general shape or trend was reduced (increased variance at shorter lags than globally) when the elevation points derived from GPS measurements were added to the dataset. Ten set elevation data were used and integrated to produce DTM. These datasets vary in scale and contour interval which remarks difference in horizontal (planimetric) and vertical accuracies. The problem of integrating elevation data from different sources with different scales and accuracies lies on the fact that the elevation values in the combined dataset may lie close to each other. The challenge is to identify an appropriate approach to prioritize the datasets, to identify which of those datasets represent the true terrain elevation and to combine the entire datasets. Thus, the datasets are prioritized with 2 steps; 1) Prioritization based on Nominal horizontal and vertical accuracies (based on the National Standard Data Accuracy (NSSDA)), 2) Prioritization based on data forms (spot heights and contour lines) and production date (see table 3.1). Table 3.1: Available elevation data sources for DTM generation
The contour lines are converted to points and then combined with other point form data (spot heights). Data with higher priority score would replace the lower priority score data. The replacement is done when 2 or more elevation points fall within 3 meters radius. The effect of the integration of the multi-sources elevation data is assessed by means of semi-variogram analysis. The assumption is points that are close together should have less difference or high autocorrelation. Thus, high nugget value would reveal strong effect of disagreement between the elevation datasets. Certainly, the nugget effect could also attribute to the complexity of terrain features. However it was found that, the effect of data disagreement still appear when datasets with complex geomorphological features were removed. Several attempts with different integration method were used (see table 3.2) to decrease the nugget value. However, for the sake of the terrain complexity information, the nugget value was reduced from 2.9 to 2.2. Table 3.2: The value of Semi-variogram model parameters for each dataset; the 2nd and the 3rd datasets will be used for the DTM interpolation
At this stage, the integrated elevation datasets with low nugget value are assumed to have less disagreement between elevation dataset, less overlapping dataset, good elevation data in representing the real terrain and inevitably contain degraded complex terrain features. 3.1.2 Elevation data interpolation The DTM interpolation method should in general preserve the detailed terrain information while reducing the effect datasets disagreement. The interpolation of the ground terrain is done with 4 interpolation techniques; 1) Kriging, 2) TIN,3) Polynomial trend surface and 4) ANUDEM (see figure 3.2). A Geostatistical interpolation or kriging interpolation method is similar to a probabilistic interpolation technique, in which the weights are derived from the surrounding sample points. However, the weights are not only based on the distance, but also on the strength of the overall correlation among the measured points (Maune, 2001). The basic interpolation assumption is that, values at a short distance are more likely to be similar than at a larger distance. Demirhan et al., (2003) in his study had focused on the performance of several interpolation methods with presence of noise and sampling pattern. He had pointed out that the “Ordinary Kriging is the most robust interpolations method against noise”. On the other hand, the deterministic DTM interpolation approach tries to fit a mathematical function to a set of elevation samples of known coordinate (x and y). This interpolation technique could be done either through an exact interpolation or smooth interpolation (Meijerink et al., 1994).  Figure 3.2: Natural terrain derived from 6th Polynomial degree (a), TIN-based terrain modelling (b), ANUDEM (c), Ordinary Kriging (100 m conversion interval) (d), and Ordinary Kriging (5 m average block) (e), visualized in 3D The smooth interpolation method is suitable with the assumption that point measurement data are regarded as true value or with less error. On the other hand, with coarse data accuracy, the smooth interpolation scheme might be the best way to level out the error to some degree. The fitting process uses various types of mathematical functions usually known as polynomial functions at a certain degree of complexity to fit the surface through the sample points. The nature of TIN modelling is splitting the surface into triangular element planes (Meijerink et al., 1994). More detailed definition, “TIN is a digital terrain model based on irregular array of points which forms a sheet of non-overlapping contiguous triangle facets” (Maune, 2001). The next interpolation technique is ANUDEM. ANUDEM is a software package known as the Australian National University Digital Elevation Model developed by Hutchinson (Geodata and Geoscience Australia, 2002). It was designed and optimised to create a hydrologically correct terrain models. “It’s unique in both input and output for building a good terrain model”(Maune, 2001). The input data is not only confined to point data, but also lines which represent streams and ridges for drainage, and polygons as a lake boundary to produce a DEM that is virtually free of spurious sinks and pits |