Flood Susceptibility Mapping and Engineering Exposure Assessment of Zhob District, Balochistan, Using Sentinel-1 SARValidated AHP and Frequency Ratio Modeling

Jamil Rehman 

Zhob District in northern Balochistan was substantially affected by the 2022 Pakistan monsoon
floods, yet, unlike comparable districts mapped in the immediate aftermath of the same event
(Larkana, Nowshera, Pishin, Quetta), it has lacked a dedicated, district-specific flood
susceptibility assessment. This study addresses that gap using a hybrid Analytic Hierarchy
Process (AHP) and Frequency Ratio (FR) approach, validated against an independently derived
flood extent layer obtained from multi-temporal Sentinel-1 SAR change detection. Flood extent
was identified by comparing an August 2022 flood-peak composite against a multi-year, samemonth baseline (August 2019–2021), covering approximately 10.1 km² within the study area
after speckle filtering, dual-polarization thresholding, connected-component cleanup, and a
terrain-slope exclusion mask.
Rather than adopting factor-direction assumptions from lowland-floodplain literature, each
conditioning factor's relationship to observed flooding was tested empirically via Frequency
Ratio. This revealed that slope and Topographic Wetness Index behave contrary to the
conventional flood-susceptibility pattern in this setting: steep terrain (FR = 3.80 in the steepest
class) and low-TWI, fast-draining terrain (FR = 3.31 in the lowest class) show the strongest
positive association with flooding, indicating a flash-flood regime originating in steep uppercatchment tributaries rather than gradual floodplain inundation. Factor directionality in the
AHP model was corrected accordingly. The resulting susceptibility model achieved strong
validation performance: 90.2% of SAR-confirmed flood pixels fell within the model's High and
Very High classes, which together cover only approximately 40% of the study area — a
discrimination ratio of 2.25 relative to random assignment, and consistent with validation levels
reported for comparable regional AHP studies despite representing a fundamentally different
flood mechanism.
The susceptibility surface was overlaid with OpenStreetMap road and settlement data to
produce an engineering-oriented exposure ranking, identifying specific road segments —
predominantly local and tertiary roads in the northwestern district rather than the primary N50
highway — and settlements at elevated flood risk. These results are intended to support
drainage-structure review and maintenance prioritization by local infrastructure authorities.
The principal contribution of this study is methodological as much as descriptive: it
demonstrates that empirically testing, rather than assuming, conditioning-factor directionality
is necessary in flash-flood-dominated mountainous catchments, where applying lowlandderived AHP conventions would misclassify the terrain most responsible for observed
flooding.