Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FLNet: Flood-Induced Agriculture Damage Assessment using Super Resolution of Satellite Images

Sanidhya Ghosal, Anurag Sharma, Sushil Ghildiyal, Mukesh Saini

TL;DR

FLNet tackles flood-induced crop damage assessment for smallholder-dominated regions by bridging the resolution gap between Sentinel-2 and high-resolution imagery. It upsamples NDVI from 10 m to 3 m using an EDSR-based SR trained on PlanetScope data, then computes a high-resolution change map $ΔNDVI$ used by a UNet to classify damage. On BFCD-22, FLNet achieves a Full Damage F1 of 0.89, matching high-resolution imagery while avoiding costly data. This approach offers a scalable, cost-effective path for rapid, farm-level disaster assessment and relief planning.

Abstract

Distributing government relief efforts after a flood is challenging. In India, the crops are widely affected by floods; therefore, making rapid and accurate crop damage assessment is crucial for effective post-disaster agricultural management. Traditional manual surveys are slow and biased, while current satellite-based methods face challenges like cloud cover and low spatial resolution. Therefore, to bridge this gap, this paper introduced FLNet, a novel deep learning based architecture that used super-resolution to enhance the 10 m spatial resolution of Sentinel-2 satellite images into 3 m resolution before classifying damage. We tested our model on the Bihar Flood Impacted Croplands Dataset (BFCD-22), and the results showed an improved critical "Full Damage" F1-score from 0.83 to 0.89, nearly matching the 0.89 score of commercial high-resolution imagery. This work presented a cost-effective and scalable solution, paving the way for a nationwide shift from manual to automated, high-fidelity damage assessment.

FLNet: Flood-Induced Agriculture Damage Assessment using Super Resolution of Satellite Images

TL;DR

FLNet tackles flood-induced crop damage assessment for smallholder-dominated regions by bridging the resolution gap between Sentinel-2 and high-resolution imagery. It upsamples NDVI from 10 m to 3 m using an EDSR-based SR trained on PlanetScope data, then computes a high-resolution change map used by a UNet to classify damage. On BFCD-22, FLNet achieves a Full Damage F1 of 0.89, matching high-resolution imagery while avoiding costly data. This approach offers a scalable, cost-effective path for rapid, farm-level disaster assessment and relief planning.

Abstract

Distributing government relief efforts after a flood is challenging. In India, the crops are widely affected by floods; therefore, making rapid and accurate crop damage assessment is crucial for effective post-disaster agricultural management. Traditional manual surveys are slow and biased, while current satellite-based methods face challenges like cloud cover and low spatial resolution. Therefore, to bridge this gap, this paper introduced FLNet, a novel deep learning based architecture that used super-resolution to enhance the 10 m spatial resolution of Sentinel-2 satellite images into 3 m resolution before classifying damage. We tested our model on the Bihar Flood Impacted Croplands Dataset (BFCD-22), and the results showed an improved critical "Full Damage" F1-score from 0.83 to 0.89, nearly matching the 0.89 score of commercial high-resolution imagery. This work presented a cost-effective and scalable solution, paving the way for a nationwide shift from manual to automated, high-fidelity damage assessment.
Paper Structure (20 sections, 3 equations, 4 figures, 2 tables, 1 algorithm)

This paper contains 20 sections, 3 equations, 4 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Assessing Agricultural Flood Damage with Remote Sensing
  4. Super-Resolution in Remote Sensing
  5. Learned Cloud Removal Techniques
  6. Methodology
  7. Motivation
  8. Problem Formulation
  9. Network Architecture
  10. Super-Resolution with EDSR
  11. Damage Feature Generation and Labeling
  12. UNet-based Damage Segmentation
  13. Experiments and Results
  14. Dataset and Preprocessing
  15. Implementation Details
  16. ...and 5 more sections

Figures (4)

  • Figure 1: Illustrates the architecture of the FLNet model.
  • Figure 2: Pre- and post-flood NDVI over the study area in Muzaffarpur (October 2022). Healthy pre-flood vegetation (left) transitions to reduced post-flood vigor (right), motivating $\Delta$NDVI as a damage feature.
  • Figure 3: Comparison of $\Delta$NDVI for a representative subregion: (left) native 10 m Sentinel-2 (LR), (middle) super-resolved 3 m from Sentinel-2 (SR), (right) native 3 m PlanetScope (HR). Super-resolution sharpens parcel boundaries and reduces mixed-pixel smearing relative to native 10 m.
  • Figure 4: Qualitative output: UNet damage classification overlaid on RGB. Super-resolution sharpening improves boundary fidelity and retrieval of narrow Full-Damage features relative to native 10 m inputs.