GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 3.3 GNGTS 2024 Both L2 and AGMS inversions improve the model retrieval when correct log informaton is added, but a very diferent behaviour occurs when wrong data are fed to the inversions: the L2 inversion shows a signifcant artfact at the locaton of the wrong resistvity log, while the AGMS inversion is almost insensitve to the outliers. The same inversion procedure was used on a SkyTEM dataset acquired in the Netherlands in 2022, 25 kilometres west of Amsterdam (Fig. 2, top lef inset), together with 94 borehole resistvity logs, 91 VES, acquired in the same area over a period ranging many decades, in which the volume of the fresh groundwater has changed considerably. Excessive water abstracton from deep wells between 1903 to 1957 caused depleton of fresh groundwater. In 1957 pumping stopped and infltraton with treated water from the river Rhine started. This enlarged the drinking water producton capacity and restored the fresh water volume in the deep aquifer (Geelen et al., 2017; Olsthoorn and Mosch, 2020). The wells can stll be used as a back-up system if the quality of the water in the river Rhine is not sufcient. That’s why the integraton of resistvity logs and VES with AEM data is difcult: data will confict not necessarily because of their diferent support volume or sensitvity, but because they were acquired over diferent periods of tme. Thus, with the AGMS inversion we aim at two distnct goals: improving the AEM inversion where borehole logs and VES informaton bring compatble informaton; identfy the confictng informaton, as a proxy of the variatons that occurred on the fresh-sea water balance over the decades. A 40 m x 80 m XY horizontal discretzaton and log-increasing depths from 5 to 400 m were used for the inversion, with the same three-cycle inversion scheme utlized for the synthetc case. Only borehole logs and VES data were rejected in the last cycle, the aim being to identfy the informaton confictng with the AEM data, which were carefully processed. Fig. 2 presents the rejecton rate for both log data and VES data with the AGMS joint inversion, in comparison with the rejecton rate computed afer an AEM-only inversion, in which log and VES data do not concur in the model defniton. The rejecton of log data is not applied to entre logs, but value by value along the borehole depth. So, the rejecton rate indicates for each borehole log the fracton of values rejected. The overall rejecton rates are presented also in Table 1. Table 1: Comparison between rejecton rates with AGMS joint inversion of AEM, VES and log data and with AEM-only inversion Total data Data rejected with AGMS Rejecton rate % Data rejected with AEM-only Rejecton rate % Borehole logs 33646 4399 13 12646 38 VESs 1815 1159 64 1475 81