Figure 5

Spectrogram of thermistor data. Color shading shows the dominant periods of thermocline IGWs throughout our experiment. At each time, we use 8 days of temperature data to compute power density spectra for each sensor deployed at 14 m depth and plot the spectrum averaged between all instruments. Dashed lines show the theoretical periods \(T=\frac{L}{N}\sqrt{{(\frac{2\pi }{L})}^{2}+{(\frac{n\pi }{h})}^{2}}\) of mode 1 (n = 1) IGWs with wavelengths L = 4150, 3800, 2720 and 2000 m (top to bottom) in a basin with constant depth h = 40 m. Before the landslide of August 18, the temporal variability of measured N² (Fig. 1D) sufficed to produce a good agreement between theory and observations. After the landslide, however, our fits fail and most IGW energy shifts towards an oscillatory mode best described by L = 4020 m (white line).