Forecasting precipitable water vapor at unmonitored coastal sites using a sparse GNSS network and NWP forecasts

Ground-based Global Navigation Satellite System (GNSS) precipitable water vapor (PWV) provides essential observational constraints for coastal convective forecasting. However, existing statistical and deep-learning models rely on multi-year single-station archives, restricting utility to permanently instrumented locations and leaving unmonitored coastal zones dependent on uncorrected numerical weather prediction (NWP). Here, we establish a sparse-GNSS-backbone regional forecasting framework that projects observation-informed calibration beyond sensor footprints by coupling GNSS anchors with continuous NWP tendencies through geographically biased graph attention. Trained with stochastic whole-station history dropout, the architecture operates across a two-tier service paradigm: Level 1 regional coverage projecting calibration across unmonitored footprints without local GNSS hardware or archives, and Level 2 local enhancement providing precision upgrades where receivers exist. In China, Level 1 reduces raw NWP 1–24 h mean absolute error across 19 unmonitored targets by 12.8% (7.0% over calibrated NWP), where on-site GNSS records yield a 9.7% precision upgrade; across the instrumented source backbone, Level 2 reduces error by 7.5% over local fusion. An independent 138-station benchmark along the US coast confirms cross-continental portability: a 30-station backbone reliably projects calibration across 108 verification targets (reducing error by 6.1% over calibrated NWP), with skill sustained across 200–300 km. Under severe sensing disruptions, the architecture maintains a 9.9% error reduction during tropical cyclones, and its advantage increases to 19.9% (China) and 30.7% (US) during extended telemetry outages. This framework demonstrates observational leverage, transforming sparse GNSS arrays into regional predictive assets.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-19
DOI
https://doi.org/10.21203/rs.3.rs-11322350/v1
Primary Topic
GNSS positioning and interference
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Forecasting precipitable water vapor at unmonitored coastal sites using a sparse GNSS network and NWP forecasts

Han Su, GanZhi Luo, Jie Zhao, Peng Min et al.
Research Square
GNSS positioning and interference
preprint

Forecasting precipitable water vapor at unmonitored coastal sites using a sparse GNSS network and NWP forecasts

Han Su, GanZhi Luo, Jie Zhao, Peng Min, You Yin, Linrui Liu
preprint en

Abstract

Ground-based Global Navigation Satellite System (GNSS) precipitable water vapor (PWV) provides essential observational constraints for coastal convective forecasting. However, existing statistical and deep-learning models rely on multi-year single-station archives, restricting utility to permanently instrumented locations and leaving unmonitored coastal zones dependent on uncorrected numerical weather prediction (NWP). Here, we establish a sparse-GNSS-backbone regional forecasting framework that projects observation-informed calibration beyond sensor footprints by coupling GNSS anchors with continuous NWP tendencies through geographically biased graph attention. Trained with stochastic whole-station history dropout, the architecture operates across a two-tier service paradigm: Level 1 regional coverage projecting calibration across unmonitored footprints without local GNSS hardware or archives, and Level 2 local enhancement providing precision upgrades where receivers exist. In China, Level 1 reduces raw NWP 1–24 h mean absolute error across 19 unmonitored targets by 12.8% (7.0% over calibrated NWP), where on-site GNSS records yield a 9.7% precision upgrade; across the instrumented source backbone, Level 2 reduces error by 7.5% over local fusion. An independent 138-station benchmark along the US coast confirms cross-continental portability: a 30-station backbone reliably projects calibration across 108 verification targets (reducing error by 6.1% over calibrated NWP), with skill sustained across 200–300 km. Under severe sensing disruptions, the architecture maintains a 9.9% error reduction during tropical cyclones, and its advantage increases to 19.9% (China) and 30.7% (US) during extended telemetry outages. This framework demonstrates observational leverage, transforming sparse GNSS arrays into regional predictive assets.

Research Square
Chengdu Jincheng College (CN), Sichuan Normal University (CN)
GNSS positioning and interference
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon