ALTIS: Automated Loss Triage and Impact Scoring from Sentinel-1 SAR for Property-Level Flood Damage Assessment

Amogh Vinaykumar, Prem Kamasani · 2026

Read ALTIS: Automated Loss Triage and Impact Scoring from Sentinel-1 SAR for Property-Level Flood Damage Assessment by Amogh Vinaykumar, Prem Kamasani (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-13803.txt.

Introduction

Vinaykumar and Kamasani present ALTIS, a pipeline that scores flood damage house by house from satellite radar. Radar sees through clouds, so maps can be ready within a day or two of a flood peak. ALTIS mixes change detection, depth estimates from terrain data, and parcel records for each property. Depth-damage links are tuned against real insurance claims from the national flood program. The output is a ranked triage list with confidence scores, tested on Hurricane Harvey data from Harris County, Texas.

Catalog record

CREATORAmogh Vinaykumar, Prem Kamasani
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE91167
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.13803

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

ALTIS: Automated Loss Triage and Impact Scoring from Sentinel-1 SAR for Property-Level Flood Damage Assessment Amogh Vinaykumar, Prem Kamasani Abstract Floods are among the costliest natural catastrophes globally, generating tens of billions of dollars in insured losses per event. Yet the property and casualty insurance industry’s post-event response remains heavily reliant on manual field inspection: a slow, expensive, and geographically constrained process. Satellite Synthetic Aperture Radar (SAR) offers a cloud-penetrating, all-weather imaging capability uniquely suited to rapid post-flood assessment, but existing research overwhelmingly evaluates SAR flood detection against academic benchmarks such as IoU and F1-score that do not capture insurance-workflow requirements. In this paper, we present ALTIS: a five-stage pipeline designed to transform raw Sentinel-1 GRD and SLC imagery into property-level impact scores within 24–48 hours of a flood event peak. Unlike prior approaches that produce pixel-level damage maps or binary flooded/non-flooded outputs, ALTIS delivers a ranked, confidence-scored property triage list directly consumable by claims management platforms. Our pipeline integrates (i) a multi-temporal SAR change detection stage using dual-polarization VV/VH intensity and InSAR coherence, (ii) physics-informed flood depth estimation by fusing flood extent with high-resolution digital elevation models, (iii) property-level zonal statistics derived from parcel footprint overlays, (iv) depth-damage curve calibration against NFIP historical claims, and (v) a confidence-scored triage ranking output. We formally define the novel task of Insurance-Grade Flood Triage (IGFT) and introduce two insurance-aligned evaluation metrics: the Inspection Reduction Rate (IRR) and the Triage Efficiency Score (TES). Using Hurricane Harvey (2017) across Harris County, Texas as a comprehensive case study, we present preliminary analysis grounded in validated sub-components and measured stage-level benchmarks suggesting that the ALTIS pipeline is designed to achieve an IRR of approximately 0.52 at 90% recall of high-severity claims, potentially eliminating over half of unnecessary field dispatches. These estimates are grounded in executed flood extent and depth validation results, published benchmarks for each algorithmic stage, and the physical properties of the Harris County flood domain. By blending SAR flood intelligence with the realities of claims management, ALTIS establishes a methodological baseline for translating earth observation research into measurable insurance outcomes. K eywords Synthetic Aperture Radar \cdot Flood Mapping \cdot Insurance Triage \cdot Sentinel-1 \cdot Depth Estimation \cdot Property-Level Assessment \cdot Hurricane Harvey \cdot Remote Sensing 1 Introduction Flood events represent the single largest driver of insured natural catastrophe losses globally. Between 2000 and 2023, floods caused an estimated $1.3 trillion USD in total economic losses, of which only a fraction was covered by insurance, a protection gap that has itself become a subject of growing regulatory concern [ 28 ] . In the United States alone, Hurricane Harvey (2017) generated over $125 billion in total damages, with National Flood Insurance Program (NFIP) claims exceeding $8.9 billion across more than 80,000 properties in Harris County, Texas [ 1 ] . The standard post-flood claims workflow has not fundamentally changed in decades: an insured files a First Notice of Loss (FNOL), an adjuster is dispatched to inspect the property, damage is assessed, and a payout is authorised. For major catastrophe events affecting tens of thousands of properties simultaneously, this workflow produces severe operational bottlenecks. Adjusters are geographically constrained, road closures prevent timely access, and claim settlement frequently extends to weeks or months. Multiple studies have estimated that 40 to 60 percent of field inspections dispatched following major flood events are redundant, directed to properties that experienced no inundation or sustained only superficial damage below deductible thresholds [ 15 , 20 ] . Satellite remote sensing offers a promising path toward resolving these inefficiencies. Synthetic Aperture Radar (SAR) sensors, in particular the ESA Sentinel-1 constellation operating at C-band (5.405 GHz), produce both Ground Range Detected (GRD) and Single Look Complex (SLC) products that are unaffected by cloud cover or solar illumination, precisely the conditions present during active flood events. The theoretical basis for SAR flood detection is well established: open water surfaces exhibit specular reflection, producing near-zero backscatter that appears dark in SAR amplitude imagery [ 30 , 4 ] . Post-event inundation mapping can therefore be achieved by comparing pre-event and post-event SAR acquisitions and identifying pixels with a statistically significant decrease in backscatter intensity. In dense urban environments, where the majority of insurable property is concentrated, amplitude change alone is insufficient. The double-bounce scattering mechanism between floodwater and building facades can cause flooded urban pixels to appear anomalously bright rather than dark, inverting the naive detection logic [ 17 , 4 ] . The complementary signal of Interferometric SAR (InSAR) coherence, which decreases when floodwater destabilises previously stable building facades and street surfaces, provides an additional discriminative layer that amplitude methods cannot reliably supply in these settings [ 4 , 18 ] . Despite this strong theoretical foundation and a rapidly maturing research literature, a critical translational gap persists. Existing SAR flood detection studies evaluate model performance using metrics drawn from semantic segmentation: Intersection over Union (IoU), pixel accuracy, and F1-score computed against ground-truth flood extent masks [ 2 , 13 , 11 ] . These metrics quantify how accurately a model can delineate inundated from non-inundated areas at the pixel level. They do not measure what a property and casualty insurer operationally requires: which specific insured properties were flooded, to what depth, with what estimated structural damage, and in what order they should be prioritised for field dispatch versus remote settlement. This paper addresses that gap directly. This paper makes three primary contributions toward closing that gap. 1. We formally define the task of Insurance-Grade Flood Triage (IGFT) : given a portfolio of insured properties and a triggering flood event, rank properties by expected damage severity using satellite SAR imagery such that the highest-confidence, highest-severity claims are dispatched first and a demonstrably smaller fraction of unnecessary inspections is conducted. 2. We introduce two evaluation metrics aligned with insurance operational requirements. The Inspection Reduction Rate (IRR) measures the fraction of field dispatches eliminated at a specified recall threshold. The Triage Efficiency Score (TES) is a composite metric that trades off claim recall against dispatch reduction, making the business case for SAR-based triage directly legible to property and casualty carriers and claims operations leaders. 3. We implement the ALTIS pipeline end-to-end on Hurricane Harvey (2017) and present preliminary performance estimates grounded in validated sub-component benchmarks and the physical properties of the Harris County domain, using publicly available data and widely accessible research tooling, with code and processing scripts released to enable future comparison against this baseline. The remainder of this paper is organised as follows. Section 2 surveys related work across SAR flood detection, physics-informed depth estimation, and catastrophe insurance analytics. Section 3 motivates the choice of Sentinel-1 as the primary data modality and presents the optical–SAR contrast that constrains pipeline design decisions. Section 4 formalises the IGFT task definition and evaluation metrics. Section 5 describes the ALTIS pipeline architecture in detail. Section 6 covers deployment architecture and operational latency. Section 7 presents the experimental protocol and preliminary results. Section 8 addresses limitations, failure modes, and future directions. 2 Related Work 2.1 SAR-Based Flood Detection The use of synthetic aperture radar for flood mapping spans more than three decades, from early demonstrations on ERS-1 data through to contemporary deep learning systems trained on global benchmarks [

Read the complete text