Sparse sensing to dense assessment: Incorporating spatial autocorrelation for assessing flood impacts

Area of Science:

• * Hydrology and Remote Sensing
• * Geospatial analysis and disaster management

Background:

• * Rapid and comprehensive flood impact assessment is crucial for effective emergency relief operations.
• * Sensor-driven approaches provide rapid flood data but are limited by sparse distribution, hindering comprehensive assessments.
• * Existing methods struggle to provide complete flood impact coverage due to sensor limitations.

Purpose of the Study:

• * To develop a systematic approach for rapid and comprehensive flood impact assessment.
• * To integrate sparse sensor data with spatial autocorrelation for improved flood impact modeling.
• * To enhance the coverage of flood impact assessments beyond sensor-covered areas.

Main Methods:

• * Extraction of flood impact data from sparsely distributed sensors.
• * Modeling spatial autocorrelation of flood-impacted areas using flood evolution and geographic principles.
• * Parameter learning for the spatial autocorrelation model via gradient descent.
• * Inference of flood impacts in sensor-uncovered areas using modeled spatial autocorrelation.

Main Results:

• * The proposed spatial autocorrelation model demonstrates effective generalization capabilities.
• * Accurate inference of flood impact probabilities for neighboring highway segments.
• * Significant extension of flood impact assessment coverage compared to sensor-only approaches.
• * Maintained high accuracy in flood impact assessment with extended coverage.

Conclusions:

• * The integrated approach significantly enhances the spatial coverage of flood impact assessments.
• * This method provides a more comprehensive understanding of flood impacts, aiding emergency management.
• * The approach offers a valuable tool for disaster response by improving situational awareness.