Wind-driven forest seismicity encodes tree water status

Abstract. Wind-driven tree motion generates ground vibrations that can be recorded by seismic sensors. While wind provides the excitation, the resulting motion may also reflect the condition of the trees. We test whether seismic observations contain information on tree water status beyond that provided by meteorological and soil moisture measurements. We combine observations from three growing seasons (2024–2026) at the ECOSENSE forest in southwestern Germany: five broadband seismometers, meteorological and soil moisture measurements, and high-resolution stem-diameter measurements from four mature Douglas firs (Pseudotsuga menziesii) and four mature European beeches (Fagus sylvatica). From the stem-diameter records, we derive tree water deficit (TWD), which quantifies reversible stem shrinkage and serves as an indicator of tree water status. We use principal component analysis to characterize collective seismic dynamics and gradient-boosted tree models to test whether seismic observations improve TWD prediction beyond environmental predictors. The seismic network forms a continuous, structured latent space organized by wind. For the illustrated tree, TWD does not define a separate seismic regime but is associated with differences in how observations populate this latent space. Adding seismic predictors improves held-out TWD prediction for many tree–station combinations, both with and without soil moisture in the baseline, but gains vary substantially among trees and stations. For Douglas fir Pm1 and station C6A, adding seismic predictors to meteorology and soil moisture increases held-out R² from 0.64 to 0.80 and reduces RMSE by 26.4%. Other combinations show smaller gains or deterioration. These findings indicate that the seismic response of a wind-driven forest contains information associated with tree water status beyond the measured environmental conditions. Ground-based seismic sensing could therefore complement direct physiological measurements in multimodal forest monitoring.