Methods

Methodology

Data Preparation

In order to perform spatial operation on the data layers, all the layers were clipped into the study area. The study area includes the subbasins within the State of Alaska and excludes the ones falling into Yukon, Canada. Two of the large basins at the boundary were partly cut off and may cause some small errors (uncertainties and errors will be addressed in the ‘Discussion’ section).

Spatial Analysis

Soil

Description:
High values were given to soil types that would have slower infiltration rates and lower permeability to water. Soil types with larger grains or higher organic matter contents such as Duneland and Mollisols were given lower values than soils with high water tables (Entisols & Inceptisols) and Permafrost near the surface (Gelisols). Spodosols can be sandy and well-drained or poorly drained, however they are generally loamy- and coarse-textured so moderately high value was assigned to this soil class. Rough Mountainous Lands usually are stony with a shallow covering of soil and so we ranked it as most likely to flood. Based on our soil ranking criteria this was appropriate, but upon completing the analysis we felt it would have been beneficial to either remove this soil class or give it a lower ranking. It became apparent that the confounding factor of elevation could not be corrected for with just the addition of a slope layer.

Analysis:
In order to perform further operations on the soil data layer, it needs to be first converted from feature to raster (using the map unit ID, MUID, as the target feature). The soil classification information was obtained from the metadata key document. After the input of the soil classification information, we were able to rank the drainage efficiency on a scale of 1 to 7. The ranking process of soil types involves subjective determination that could result in errors and inaccuracies. [sources of error from the ranking process and conversion (feature to raster) are addressed in the discussion section].

Table 2. Soil type ranking

Fuzzy membership tool (type: linear) was used to standardize the possibility of flooding in terms of the types of soil.

Figure 1. Soil type ranking map [click on image to enlarge]

Water Accumulation Area

Description:
Floods are likely to occur in areas where water accumulates as a function of the slope.

Analysis:
Perform a flow direction calculation using a ‘Hydrology’ tool, ‘Flow Direction’. The DEM layer is used as the input layer for the calculation since this tool assumes that the flow direction is essentially based on the contour of the landscape. The 'Flow Accumlation' tool is then used to calculate the accumulation area

The accumulation layer was then clipped using the Extract by “Mask” tool. With the assumption that the water is more likely to accumulate in the top 97% of the accumulation areas, the accumulation layer (raster) was reclassified: the top 97% were given a value of 2 and the lowest 3% were given no value.

After the reclassification process, the raster layer was converted to vector for the buffer operation. A 5 km buffer was created around the accumulation vector layer with the assumption that the 5 km buffer area is considered to be the most susceptible to flood. A new field was created in the attribute table of the new buffer layer and a value of one was given to all the buffer values.

Using one of the ‘Analyst - Overlay tools’, ‘Union’, the buffered layer and the study region (db_studyregion) layer were combined together. Lastly, the resulting vector was converted in to a raster layer to be used as one of the criteria for the MCE.

Figure 2. Water accumulation area map

Drainage Density

Description:

Drainage density is defined as the total length of all the rivers and streams in the drainage basin divided by the total area of the drainage basin (Wikipedia). It represents the “density” of rivers that exist in a drainage basin. Most literature indicates that high “density” of rivers in a drainage basin is usually an indicator of high flood risk . In addition, a high drainage density is also associated with bare soil and arid areas with sparse vegetation cover (Pallard et al., 2009). This correlates with our soil and vegetation ranking system. High “density” of rivers within an area is usually an indicator of high flood risk because the presence of streams is an important factor that contributes to the occurrence of flood.

Analysis:

The operation ‘Spatial Join’ was done between the rivers layer and "DB_study area". Then summarize the sub-basin by sum of river lengths. Create a new field in table called "db_density" and calculate the drainage basin density by dividing total length of river networks by the area of subbasins.

Figure 3. Drainage density map (unit: km/km2)

Fuzzy membership tool (type: linear) was used to standardize the possibility of flooding in terms of drainage densities. The highest drainage density is set to be maximum value while the lowest value for drainage density is set to be the minimum.

Wetland

Description:

The presence of wetlands tends to mitigate flooding, however they are typically found near rivers where floods tend to occur. This issue will be addressed in the discussion - errors section.

Analysis:

Create a “new_field” in wetland layer attribute table. A value of 1 is assigned to all wetland cells. Join the subbasin study region to the wetland polygon using the ‘Union’ tool (when converting to a raster layer, “new_field” field was used as the value field).

Figure 4. Wetland distribution map

Fuzzy membership tool (type: linear) was used to standardize the possibility of flooding in terms of the presence of wetlands. The maximum is set to be zero and the minimum is set to be 1 since the presence of wetland will decrease the possibility of flood occurrence.

Landcover (Vegetation)

Description:
Needle-leafed forests are established in areas with the lowest flood frequencies with Broadleaf forests occurring at low flood frequency areas as well. Theoretically, areas with less floods tend to allow plants with a longer life history to establish. Plants that thrive in high disturbance areas typically grow faster and are the first to colonize disturbed lands, they are called ruderal species. It is likely that herbaceous plant would be more likely to persist in a flood disturbed area compared to Shrublands because woody plants take longer to grow. Lichen tends to grow on rocky places and very slowly, so are less likely to be found in areas with a high flood-frequency. Barren land is most likely to flood because there are no plants to intercept precipitation and no macropores in the soil that would have be created by plant roots allow drainage. We assume ice and snow cover to have the highest flood frequency because rain on snow events tends to generate the most water run-off. Unfortunately this assumption along with categorizing rough mountainous land as the likeliest to cause flooding may have skewed our results, but the creation of the slope layer was to mitigate this effect.

Analysis:

Summarize the data by landcover category and create a summary table. Determine the rankings (see vegetation justification) and join the table to the layer.

Table 3. Vegetation ranking table

Figure 5. Vegetation and rankings map

In order to perform fuzzy membership operation on the vegetation layer, it needs to be converted from vector to raster first.

Fuzzy membership tool (type: linear) was used to standardize the possibility of flooding in terms of the types of vegetation.

Precipitation

Description:

In order for precipitation to be used as one of criteria for the multi-criteria evaluation, the original data layer needs to be adjusted and classified. Since the body of precipitation (ie. rain and snow) tends to travel downhill, it would be inaccurate to use the actual values associated with each cell to calculate flood risk, so instead, the sum of annual precipitation was calculated for each basin as a preparation for further spatial analysis.

Analysis:

An operation of ‘Spatial Join’ was done between the precipitation raster layer and subbasins and the sum of annual rainfall values within each water basin; the resulting layer represents the amount of total rainfall for each basin.

Figure 6. Precipitation map

Fuzzy membership tool (type: linear) was used to standardize the probability of flood occurrence in terms of the amount of annual rainfall. The maximum value was set to be the higher annual rainfall while the minimum was set to be the lowest value for annual rainfall. Generally, high level of precipitation is associated with higher possibility of flood occurrence whereas lower level of precipitation is associated with smaller possibility of flood occurrence.

Slope

The slope is considered the most important factor in our MCE analysis. It was created from the DEM layer that was previously clipped to our study area.

Figure 7. Slope map

Fuzzy membership tool (type: linear) was used to standardize the possibility of flooding in terms of the slope based on DEM.

Multi-criteria Evaluation (MCE)

The Pairwise Comparison Method is one of the methods for determining weights of criteria in an MCE. This method only allows the comparison of only two criteria at once. We can convert subjective assessments of relative importance into a linear set of weights (Heywood et al., 1993). In our MCE, AHP was chosen to be used for the determination of the weights for the criteria.

Analytical Hierachy Process Tool (AHP)

The ranking of the criteria for the MCE is mainly based on the level of limitation and errors that are present in each criterion layer. For instance, during data manipulation processes for the land cover (vegetation) and for the soil type layer, there were many sources of uncertainties or errors present that may have an effect on the final result of our analysis. Therefore, they were rated to be the less important factors due to their relatively high level of uncertainties. Since the ranking of the criteria was based on the relative importance of one factor comparing to others, the determination can be largely depended upon the interest of investigation. Other possibilities of the ranking can also be used when the purpose of the analysis is different or when the quality of data is different under a different circumstance.

Figure 8. AHP results

Using one of the ‘Spatial Analyst Tools- Overlay’, perform the MCE with the ‘Weighted Sum’ tool (the same tool is used for all four sensitivity analysis outputs). After creating the maps showing the MCE results, all the layers were reclassified to 30% and 70% of their values, but only the top 30% were shown on the MCE final result map.

Post- MCE Analysis:

First convert all the reclassified layers from raster to polygon. In the attribute table of the new polygon layers, create a new field named “area” and use the calculate geometry tool to calculate the areas. Then create a new layer and calculate the area of each basin. Using the 'Spatial Join' tool to join the "db_studyregion" layer and the polygon layers. In the attribute table, use the field calculator to calculate the percentage of the area affected by the flood to the area of each subbasin.

Sensitivity Analysis:
Similar procedures are repeated for our four sensitivity analyses.

Potential Damage on Human Properties

To find the towns affected by the most top 30% of the flood risk area (the most intensely affected upper 30%), simply use the 'Select by Attribute' tool to select the points (towns) that fall within the upper 30% on the flood risk area map (vector).

To find the infrastructure affected: Using the 'Transformation' tool, re-project the original infrastructure layer to maintain spatial alignment. Then the resulting layer was clipped to the upper 30% of the flood risk area (polygon layer). A 'new field' was then created to calculate the length of all the infrastructure types (eg. highways, electrical lines, etc.). Lastly, summarize the result and obtain a table displaying the total lengths affected for each type of infrastructure.