<p>Fumarolic gas composition and temperature record deep processes that generate and transfer heat and mass towards the surface. &#160;These processes are a result of the emplacement, degassing and cooling of magma and the overturning of the above hydrothermal system. &#160;A reasonable expectation, and too often an unproved assumption, is that fumarole temperatures and the deep heat sources vary on similar timescales. &#160;Yet signals from deep and shallow processes have vastly different temporal variations.&#160; This indicates that signals arising from deep activity may be masked or modified by intervening hydrothermal processes, such as fluid-groundrock reactions in which secondary minerals play a major role. &#160;Clearly, this complicates the interpretation of the signals such as the joint variation of fumarole vent temperature and geochemical ratios in terms of what is occurring at depth. &#160;So what do the differences between the timescales governing deep and shallow processes tell us about the intervening transport mechanisms?</p><p>At the volcanic dome of La Soufri&#232;re de Guadeloupe, the Observatoire Volcanologique et Sismologique de la Guadeloupe has performed weekly-to-monthly in-situ vent gas sampling over many years. &#160;These analyses reliably track several geochemical species ratios over time, which provide important information about the evolution of deep processes. &#160;Vent temperature is measured as part of the in-situ sampling, giving a long time series of these measurements. &#160;Here, we look to exploit the temporal variations in these data to establish the common processes, and also to determine why these signals differ. &#160;By fitting sinusoids to the gas-ratio time series we find that several of the deep signals are strongly sinusoidal. &#160;For example, the He/CH<sub>4</sub> and CO<sub>2</sub>/CH<sub>4</sub> ratios, which involve conservative components and mark the injection of deep and hot magmatic fluids, oscillate on a timescale close to 3 years. We also analyse the frequency content of the temperature measurements since 2011 and find that such long signals are not seen. &#160;This may be due to internal buffering by the hydrothermal system, but other external forcings are also present. &#160;From these data we build up a more informed model of the heat-and-mass supply chain from depth to the surface. &#160;This will potentially allow us to predict future unrest (e.g. thermal crises, seismic swarms), and distinguish between sources of unrest.</p>