Three-dimensional tree canopy depth and wind dynamics govern how forests capture rainfall across the globe. Standard earth system models assumed that flat leaf surface area alone determined how much water canopies catch. Instead, falling rain strikes the full vertical depth and width of forest crowns while moving air circulates moisture through the branches.
Raindrops strike the outer canopy and filter downward through layers of foliage and branches. The three-dimensional crown acts like a deep thicket rather than a flat sheet, trapping water across its entire volume. Blowing wind pushes droplets against branch surfaces, shifting how moisture clings to foliage before reaching the ground. Interception changes in strong, non-linear ways as rain intensity and canopy structures shift.
Researchers created an artificial intelligence framework called Feature-Constrained Deep Symbolic Regression to derive explicit equations for canopy interception. The new morphology equations increased site-level interception loss Kling–Gupta efficiency by 0.27 to 0.39 compared to traditional leaf area methods. Water-balance tests confirmed enhanced evapotranspiration at nearly 70 percent of global flux sites and stronger runoff dynamics across six major river basins.
Explicit representation of three-dimensional canopy architecture allows scientists to produce reliable predictions of terrestrial water balances. Adding this scheme into the CoLM land surface model improved runoff calculations for 71.5 percent of United States catchments and reproduced river discharge variability.