[Paper Review] Scalable Surface Water Mapping up to Fine-scale using Geometric Features of Water from Topographic Airborne LiDAR Data

Hunsoo Song, Jinha Jung|arXiv (Cornell University)|Jan 16, 2023

Remote Sensing and LiDAR ApplicationsEnvironmental Science3 citations

TL;DR

This paper presents a fully automated, scalable method for high-resolution surface water mapping using topographic airborne LiDAR data by exploiting the natural geometric property that water surfaces are flat due to gravity. The method achieves accurate 3D water body extraction across diverse landscapes—including urban, coastal, and mountainous regions—without relying on optical reflectance, outperforming NDWI-based methods even without parameter tuning.

ABSTRACT

Despite substantial technological advancements, the comprehensive mapping of surface water, particularly smaller bodies (<1ha), continues to be a challenge due to a lack of robust, scalable methods. Standard methods require either training labels or site-specific parameter tuning, which complicates automated mapping and introduces biases related to training data and parameters. The reliance on water's reflectance properties, including LiDAR intensity, further complicates the matter, as higher-resolution images inherently produce more noise. To mitigate these difficulties, we propose a unique method that focuses on the geometric characteristics of water instead of its variable reflectance properties. Unlike preceding approaches, our approach relies entirely on 3D coordinate observations from airborne LiDAR data, taking advantage of the principle that connected surface water remains flat due to gravity. By harnessing this natural law in conjunction with connectivity, our method can accurately and scalably identify small water bodies, eliminating the need for training labels or repetitive parameter tuning. Consequently, our approach enables the creation of comprehensive 3D topographic maps that include both water and terrain, all performed in an unsupervised manner using only airborne laser scanning data, potentially enhancing the process of generating reliable 3D topographic maps. We validated our method across extensive and diverse landscapes, while comparing it to highly competitive Normalized Difference Water Index (NDWI)-based methods and assessing it using a reference surface water map. In conclusion, our method offers a new approach to address persistent difficulties in robust, scalable surface water mapping and 3D topographic mapping, using solely airborne LiDAR data.

Motivation & Objective

• To develop a robust, fully automated method for high-resolution surface water mapping that works consistently across diverse and complex landscapes.
• To overcome the limitations of optical image-based water mapping, especially in accurately detecting small water bodies (<1 ha) under variable atmospheric and surface conditions.
• To enable the extraction of both 2D water extent and 3D elevation information of water bodies from a single LiDAR point cloud, eliminating reliance on external data or post-processing.
• To produce accurate, hydrologically consistent 3D topographic models that include water bodies, reducing errors from temporal misregistration and DEM post-processing.

Proposed method

• The method leverages the physical principle that natural water surfaces are flat due to gravity, using local point density and planarity to identify potential water segments.
• It applies region merging based on geometric similarity (planarity and elevation consistency) to group LiDAR points into coherent water bodies, starting from seed points identified via local point density thresholds.
• Key parameters include Z (reference point for local density), ER (error threshold for planarity), and MS (maximum segment size), which control sensitivity and extent of water body expansion.
• The algorithm operates fully automatically with no need for site-specific parameter tuning, relying only on LiDAR point clouds to extract water bodies with their elevations.
• Water body boundaries are refined through iterative merging of adjacent regions that meet geometric criteria, ensuring flatness and elevation consistency.
• The method outputs a full 3D topographic model where water bodies are explicitly represented with their 3D geometry, enabling direct use in hydrological modeling.

Experimental results

Research questions

• RQ1 Can a LiDAR-based method achieve accurate and scalable surface water mapping across diverse landscapes without relying on optical reflectance or external data?
• RQ2 How does the geometric flatness of water surfaces enable robust detection in challenging environments such as urban areas with small water bodies and mountainous regions with shadows and snow?
• RQ3 To what extent can a fully automated method based on point cloud geometry outperform traditional NDWI-based optical image methods in detecting small water bodies?
• RQ4 Can the method produce accurate 3D water body representations, including elevation, as a by-product of the segmentation process?
• RQ5 How do key parameters (Z, ER, MS) influence the segmentation outcome, and can they be interpreted physically to ensure intuitive and explainable tuning?

Key findings

• The proposed method achieved higher accuracy than NDWI-based optical methods in detecting surface water across all tested landscapes, including urban, coastal, and mountainous areas, without any parameter tuning.
• In urban and mountainous regions with complex topography and small water bodies (e.g., puddles, ditches, ponds <1 ha), the method successfully extracted water bodies that were often missed by optical methods.
• The parameter ER (planarity error threshold) had a significant impact on segmentation results, with lower values (0.05) reducing over-estimation of water extent and higher values (0.15) increasing flooding artifacts.
• The MS (maximum segment size) parameter effectively controlled the expansion of small water bodies, with 100 m² allowing detection of smaller ponds while 1000 m² led to under-segmentation.
• The method produced a full 3D topographic model that includes both terrain and water bodies, eliminating the need for post-processing steps like hydro-flattening DEMs, thus reducing registration and temporal errors.
• Extensive testing over ~2,500 km² of diverse terrain confirmed the method’s scalability and robustness, demonstrating consistent performance across different land cover types and topographic complexities.