Improvement of the decision tree
5.4 Improvement of the decision tree
It is important that the decision tree should not be case specific, and should be used in many locations by considering all factors to assess multi-hazards susceptibility. It should be generic and simple so that it can
be easily applied by non-technical experts in most of the cases. As it is mentioned earlier, it is assumed from field experiences, in case of Nepal, it will be used by DISCO officer who has local level technical knowledge but doesn’t have any expert knowledge. Based on the field testing of initial decision tree as well as from realization of necessity of minimizing the problems in the tree, as mentioned in section 5.3, the decision tree was needed for modification. The conflicting factors of the initial decision tree have been focused on clarified and modified in the proposed decision tree by incorporating historical events inventory, and observation signs as described in Table 5.2. Additionally, the results of Flow-R runout assessment has been incorporated in the proposed tree to identify whether the unit is located in the runout flow paths or not.
Table 5.2: Proposed modification for the conflicting factors of initial decision tree
Factors in initial decision tree
Clarified in the proposed decision tree
Presence of Rocky cliff upslope
Potential indicators of recent event is included by observing the presence of different color in rock scars such as “Does it have different color rock scars or open joints”
Vegetation anomalies
To visualize vegetation anomalies, such as brokendamaged trees or presence of different aged vegetation comparing adjacent area
Evidence of fallen rocks
Potential indicators of event will be considered based on whether fallen rocks look fresh color or covered with some vegetation.
Either from observations of fallen blocks or information from local people
Does the unit have concave
It can be understood when there are deposition with
sections or hummocky parts?
colluvium materials.
Are there landforms that might be
Possible to collect from already mapped landslides
formed by landslides
information (landslides inventory) or from discussion with local people.
Does the channel show signs of
By checking the evidence by river deposition or information
large changes of discharge
from local people
The terminology of different hazard assessment was modified as susceptibility assessment of hazardous events as there was no evaluation of frequency or magnitudeintensity of hazardous events. Rather, it was considered based on the relative likelihood of future hazardous event, solely depended on the existing situation of the local site. In the proposed decision tree (Figure 5.7), there are three level of susceptibility of The terminology of different hazard assessment was modified as susceptibility assessment of hazardous events as there was no evaluation of frequency or magnitudeintensity of hazardous events. Rather, it was considered based on the relative likelihood of future hazardous event, solely depended on the existing situation of the local site. In the proposed decision tree (Figure 5.7), there are three level of susceptibility of
A second major improvement is the inclusion of Flow-R model results of rockfall and debris flows runout flow paths in the improved decision tree. The information of runout flow paths is checked for susceptibility assessment in decision tree. In addition, automated terrain unit maps are used for multi-hazard susceptibility
assessment. As because, it has the information of class of slope steepness (low: <15 0 , moderate: 15 0 -35 0 ,
steep: 35 0 -50 0 and very steepcliff: >50 0 ), land cover type as well as the existence of rockfalldebris runout.
Although, geomorphological information is not used in automated terrain unit analysis or in the decision tree, it is desirable, however, it requires expert geomorphological knowledge. Additionally, manual procedure for generating geomorphological based terrain unit might be too time consuming which might
be unrealistic in Nepal case. In general, there are five different steps in the proposed decision tree for assessing the susceptibility of multi-hazards which is shown in Figure 5.6. Checking all the alternatives one after another, the inclusive susceptibility to multi-hazards will be assessed for each terrain unit.
Historical
TU with
Unit bounded by
Unit bounded by
Unit is located
evidence of
different slope
steep slopes
steep slopes
next to drainage
hazardous event
• Present or
• Checking
• Can a hazard
• Assessing
• Assessing
not
whether a
from a steep
whether the
whether the
hazard can
upslope unit
unit will be
unit is
intitiate in the
reach the unit
affected due
susceptible to
unit itself
to downslope
flood
unit hazard
Figure 5.6: Illustrations of different levels in the proposed decision tree for multi-hazard susceptibility assessment On the other hand, considering different types of hazard different questions related to different causal
factors are organized in the decision tree. The questions related the causal factors will be answered by observing the existing field condition. As example, for rockfall susceptibility at flat unit, it is needed to check three factors such as (i) Is the unit located inside the rockfall path (ii) Does it have rock scars or open joints (iii) Is there any evidence of fallen rocks or vegetation anomalies. Depending on the answers of these questions, the unit is assessed as high or moderate rockfall susceptibility. In this way, a combination of scientific knowledge with local knowledge can be made using a simple decision tree. To overcome the problem of multi-hazards for each unit, linkage has been made between different types of hazards. The improved decision tree is shown in Figure 5.7. The workflow of proposed decision tree is like the initial one
i.e the questions will be answered by YES or NO. If the answer is YES it will go towards right direction and if it is NO, it goes downward.
It has also been considered that for some factors such as landslide inventory, model output of rockfall and debris flow runout assessment, Digital Elevation Model and derivatives such as slope steepness classes, land cover information, data on buildings and roads, and automated terrain units are needed before going to the field. On the other hand, if an apps can be developed based on the proposed decision tree, the simple GIS analysis will be done within the apps in the field (this will be described in detail in chapter-7).
Susceptibility class
Figure 5.7: The proposed decision tree of multi-hazards susceptibility assessment