GNGTS 2018 - 37° Convegno Nazionale

GNGTS 2018 S essione 3.2 703 and canopy water. In order to verify this hypothesis, we compute the ground displacement due to the hydrological loading using MERRA2 (Gelaro et al. , 2017) and GLDAS (Rodell et al. , 2004) models. Even if these models do not account for surface water, including rivers, and deep groundwater processes, they simulate the mass transfer at Earth’s surface and other environmental parameters by integrating satellite- and ground-based observations into land surface models through data assimilation techniques. With respect to the modelling results, Fig. 3a depicts the original continuous GPS time series from June 2012 to October 2016 (inter-seismic deformation time span) and its linear trend, calculated on the basis of a linear fit analysis. This fit, which indicates a mean subsidence rate of -1.9 +/- 0.1 mm/yr, has been used to preliminarily de-trend the original signal (Fig. 3b). In Fig. 3b the de-trended series is compared with the 6-months centered moving average and the hydrological modelling results. MERRA2, GLDAS/Noah v1.0 and GLDAS/Noah v2.1 models well fit the residual data and their results slightly differ due to some modelling basic conditions (e.g. spatial resolution, land surface model, layer geometry, etc.). All the models simulate periodicities at 6-months, 1- and 2-years with amplitudes ranging 6-10 mm. Although the standard statistical analysis (moving average) is able to detect the seasonal periodicities as well as the physically-based models, it underestimates the amplitude (Fig. 3b). An example of the linear fitting, after periodic trend removal with the best fitting model (GLDAS2), is shown in Fig. 3c. In this case, the most reliable mean subsidence velocity is -2.3 +/- 0.12 mm/yr. The used hydrological models have been also applied to other two GPS stations located very close to the Po River (i.e, Taglio di Po and Porto Tolle in Fig. 1). Since a single model mesh covers the entire Delta (spatial resolution: 0.5° x 0.625° for MERRA2 and 0.25° for GLDAS), the periodic oscillations simulated in these sites are very similar to the one obtained in Codigoro area. As expected, due to the Po River loading, the computed oscillations therein do not match the observations as other processes are not accounted by the modelling (e.g. surface water loading and ground settlements due to the river). In conclusion, the inter-annual (1- 2 years) variability observed in the hydro-meteorological data and due to weather and climatic processes, strongly influences the groundwater trend within the soil canopy and affects the large-scale oscillations visible in the geodetic data at CODI station. Moreover, the hydrological modelling confirms that far from the main river Fig. 3 - CODI time series: a) original data and linear fit; b) residual data after linear de-trending and fits by using moving average (red line) and physically-based models (light-green, green and brown lines); c) fitting of the final linear trend. See the text for details.