The storm surge is the non-tidal component of coastal sea-level. It responds to the atmosphere both through the direct effect of atmospheric pressure on the sea-surface, and through Ekman transport induced by wind-stress. Tide gauges have been used to measure the sea-level in coastal cities for centuries, with many records dating back to the 19th-century or even further, at times when direct pressure observations were scarce. Therefore, these old tide gauge records may be used as indirect observations of sub-seasonal atmospheric variability, complementary to other sensors such as barometers. To investigate this claim, the present work relies on tide gauge records of Brest and Saint-Nazaire, two portal cities in western France, and on the members of NOAA's 20th-century reanalysis (20CRv3) which only assimilates surface pressure observations and uses numerical weather prediction model. Using simple statistical relationships between storm surges and pressure maps, we show that the tide gauge records reveal part of the 19th-century atmospheric variability that was uncaught by the pressure-observations-based reanalysis. In particular, weighing the 80 reanalysis members based on tide gauge observations indicates that a large number of members are very unlikely, which induces corrections of several tens of Hectopascals in the Bay of Biscay. These findings support the use of early tide gauge records in sensor-scarce areas, both to validate old atmospheric reanalyses and to better probe old atmospheric sub-seasonal variability.