# Maximum Likelihood Curved Surface Estimation of Multi-Baseline InSAR for DEM Generation in Mountainous Environments

## TL;DR

This paper introduces a new InSAR method for creating accurate elevation maps in mountainous areas.

## Contribution

The novel MLCSE method improves DEM accuracy in challenging mountainous environments using adaptive strategies.

## Key findings

- MLCSE achieves a mean elevation error of 7.89 m with 70.32% of errors below 10 m.
- MLCSE outperforms other InSAR methods in hilly, mountainous, and alpine regions.
- MLCSE meets international elevation accuracy standards for various terrain types.

## Abstract

Digital elevation model (DEM) generation using Interferometric Synthetic Aperture Radar (InSAR) in mountainous environments encounters challenges including signal acquisition difficulties, decorrelation, and highly variable topography. To address these challenges, we propose a novel approach termed maximum likelihood curved surface estimation (MLCSE), utilizing multi-baseline InSAR to enhance DEM accuracy in mountainous regions. First, multi-baseline InSAR with Sentinel-1 images is employed to acquire more accurate interferometric phases. Second, two strategies are implemented to improve maximum likelihood elevation estimation, which is particularly susceptible to topographic relief and decorrelation. These strategies include replacing fixed neighborhood size with adaptive neighborhood size selection and estimating parameters of the maximum likelihood local curved surface. Finally, the mean error of the MLCSE DEM results and the proportion of errors less than 10 m are 7.89 m and 70.32%, respectively. The results demonstrate that MLCSE surpasses other InSAR methods, achieving higher elevation estimation accuracy. MLCSE exhibits stable performance across the study areas, reducing elevation errors in hilly, mountainous, and alpine regions. Additionally, hydrological analysis of the elevation results reveals that MLCSE, using the adaptive neighborhood size selection strategy, outperforms other methods in both visual inspection and quantitative comparisons. Moreover, the elevation accuracy achieved by MLCSE meets the standards of the American DTED-2, the Level 2 standard of the 1:50,000 DEM (Mountain), and the Level 1 standard of the 1:50,000 DEM (alpine region) for spatial resolution and height accuracy.