Validating and Comparing Open InSAR Vertical Land Motion Products: Lessons from Three Cases
Abstract
Open, satellite-derived vertical land motion (VLM) products are increasingly used to estimate relative sea-level rise and coastal hazard, often well beyond the studies that produced them. Researchers often infer the reliability of these VLM products from agreement with other products rather than validating them against independent observations. Researchers then sometimes attribute differences between products to processing without testing. Here, we examine three cases: 1) We independently processed Sentinel-1 data for 17 of the 48 coastal cities in a widely cited open dataset provided by Tay et al. (2022) over nearly the same observation window, and compared both products with Global Navigation Satellite System (GNSS) vertical velocities. After transforming both to IGS14, our rates agree with GNSS at R² = 0.87, whereas the transformed Tay et al. rates agree at R² = 0.3. Tay et al. velocity maps also show patterns consistent with known signatures of coregistration, phase-unwrapping, and pair-selection errors. 2) We show that a recent comparison of InSAR VLM products for coastal California by Govorcin et al. (2025) attributes their differences to Blackwell et al. (2020) to spatial filtering without controlling for observation epoch, using rate values that do not appear in the source, quantifying resolution, or accounting for localized horizontal motion. That untested attribution has since been cited as evidence of inconsistency among InSAR products. 3) Third, a comparison for New York City by Buzzanga et al. (2023) contrasts their products from different epochs with Ohenhen et al. (2023) to conclude that an Ohenhen et al. dataset lacks hot spots, although its time series show those hot spots as transient, and reprocessing with the Ohenhen et al. method over the same epoch as Buzzanga et al. recovers them. In cases 2 and 3, disagreement was attributed to processing. Therefore, we propose a six-step protocol for validating and comparing VLM products, along with a community-maintained open archive to make these checks routine.
Related
- Comparing and validating modelled snow stability metrics
- Variable contributions of vertical land motion to sea-level change inferred at tide gauges
- Assessing Spatial and Temporal Variability of Vertical Land Motion Along Coastal Massachusetts Using Sentinel-1 InSAR and GNSS Time Series Analysis
- Assessing the vertical representativeness of ERA5 humidity for interpreting satellite upper-tropospheric humidity trends
- Can vertical mixing arrest ocean frontogenesis?
BibTeX
@misc{shirzaei2026validating,
title = {{Validating and Comparing Open InSAR Vertical Land Motion Products: Lessons from Three Cases}},
author = {Manoochehr Shirzaei},
howpublished = {EarthArXiv},
year = {2026},
month = oct,
doi = {10.31223/x58n6q},
url = {https://eartharxiv.org/repository/view/15424/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Validating and Comparing Open InSAR Vertical Land Motion Products: Lessons from Three Cases
AU - Manoochehr Shirzaei
PY - 2026
DA - 2026/10/09
PB - EarthArXiv
DO - 10.31223/x58n6q
UR - https://eartharxiv.org/repository/view/15424/
AB - Open, satellite-derived vertical land motion (VLM) products are increasingly used to estimate relative sea-level rise and coastal hazard, often well beyond the studies that produced them. Researchers often infer the reliability of these VLM products from agreement with other products rather than validating them against independent observations. Researchers then sometimes attribute differences between products to processing without testing. Here, we examine three cases: 1) We independently processed Sentinel-1 data for 17 of the 48 coastal cities in a widely cited open dataset provided by Tay et al. (2022) over nearly the same observation window, and compared both products with Global Navigation Satellite System (GNSS) vertical velocities. After transforming both to IGS14, our rates agree with GNSS at R² = 0.87, whereas the transformed Tay et al. rates agree at R² = 0.3. Tay et al. velocity maps also show patterns consistent with known signatures of coregistration, phase-unwrapping, and pair-selection errors. 2) We show that a recent comparison of InSAR VLM products for coastal California by Govorcin et al. (2025) attributes their differences to Blackwell et al. (2020) to spatial filtering without controlling for observation epoch, using rate values that do not appear in the source, quantifying resolution, or accounting for localized horizontal motion. That untested attribution has since been cited as evidence of inconsistency among InSAR products. 3) Third, a comparison for New York City by Buzzanga et al. (2023) contrasts their products from different epochs with Ohenhen et al. (2023) to conclude that an Ohenhen et al. dataset lacks hot spots, although its time series show those hot spots as transient, and reprocessing with the Ohenhen et al. method over the same epoch as Buzzanga et al. recovers them. In cases 2 and 3, disagreement was attributed to processing. Therefore, we propose a six-step protocol for validating and comparing VLM products, along with a community-maintained open archive to make these checks routine.
ER -