Integrated Water Vapour (IWV) trend analysis from GNSS and NWP reanalyses: a homogenised long-term analysis over Granada

Abstract

In a context of climate change and global warming, the characterisation and operational monitoring of greenhouse gases is of uppermost importance for implementing mitigation strategies that could help to reduce the impact of the current climatic emergency in the surrounding ecosystems and society. Among these gases, water vapour can contribute to almost a 60% of the total greenhouse effect. Moreover, its interaction with solar and infrared radiation or its main role in cloud formation, make water vapour a key driver of most atmospheric thermodynamic processes and a crucial component of the Earth's radiative budget. Nevertheless, the large spatial and temporal variability of water vapour hinders the acquisition of reliable operational measurements. Remote sensing techniques such as the Global Navigation Satellite System (GNSS) have been proven to be an accurate and trustworthy alternative for integrated water vapour (IWV) retrievals, providing a valuable platform for continuous operational monitoring and thus enabling long-term characterisation. To further address this challenge, reanalysis data from Numerical Weather Prediction (NWP) models can significantly increase the temporal and spatial coverage of atmospheric variables datasets. In particular, ERA5 (fifth generation of European Centre for Medium-Range Weather Forecasts reanalysis) and MERRA-2 (Modern-Era Retrospective Analysis for Research and Applications, version 2) provide validated data for the city of Granada, in southeastern Spain, since 1980.The current study presents a comprehensive analysis of IWV trends retrieved from a 15-year GNSS database and an extended 45-year reanalysis dataset. Special attention is paid to time-series quality control and homogenisation. Small jumps or discontinuities due to GPS receiver updates or changes in the data assimilation strategies of NWP models, can introduce artificial artifacts in the time series and consequently lead to biased or misleading trend esimates. A modified Mann-Kendall test proposed by Coen et al. (2020) that applies a Variance-Corrected Trend-Free Pre-Whitening approach is evaluated against a General Least Square method with a full custom covariance matrix accounting for residual heteroscedasticity and autocorrelation. While both methodologies agree on the sign and uncertainties of the retrieved trends, some discrepancies are found in the magnitudes, reflecting the different nature of both algorithms and highlighting the sensibility of trend detection techniques. Positive increasing IWV trends of a 3% per decade on average were obtained from both datasets and algorithms, being significant to a 95% level when analysing the 45-year time series. Nonetheless, relevant behaviour differences are found between the 1980-2000 and 2000-2024 periods, unveiling the pronounced increasing in IWV experimented during the last 25 years. The results obtained are consistent with previous studies, both regarding the trend magnitude and the uncertainty range, reinforcing the capability of the GNSS technique and NWP models as robust tools for environmental and atmospheric monitoring of complex variables such as water vapour (Parracho et al., 2018; Yuan et al., 2023). However, they also unveil trend discrepancies which are inherent to the chosen retrieval methodologies and that must always be assessed.