Sap flux was measured using custom-built Granier's thermal dissipation probes. Microclimate observations were used to understand the drivers of stand transpiration. Air temperature (°C), relative humidity (R_h, %), wind speed (U, m s^-1), and incoming short-wave radiation (R_s, kW m^-2) were recorded using an automatic weather station (AWS) (Vantage-pro Davis Net, USA). On-site Precipitation (P, mm hr^-1) was recorded using an automated tipping-bucket rain gauge. Additionally, long-term trend analysis was done using daily precipitation (P_IMD, mm d^-1) extracted from the 0.25⁰ gridded precipitation product by the Indian Meteorological Department (IMD) for the pixel containing the sap flow site (Pai et al., 2014). In-canopy air temperature and relative humidity were recorded using microsensors (iButton hygrochrons, Maxim Int., USA). Forest reference evapotranspiration (E_F, mm hr^-1) was computed based on FAO’s Penman-Monteith method using meteorological data from the installed AWS and the crop coefficients developed for natural vegetation (for details see Allen et al. 1998). Hourly Vapour Pressure Deficit (D, kPa) was computed from air temperature and relative humidity. Soil water potential was recorded at 10 cm incremental depths from the topsoil to up to 30 cm depth using granular matrix-based (watermark) sensors (Virtual Electronics, Roorkee) and converted to volumetric water content using the site-specific van Genuchten water retention curve. Total soil moisture (S, mm hr^-1) was computed for the topsoil (0-30 cm depth) using the trapezoidal method. S was smoothened using a 3-step moving-average window to gap-fill stray missing values. A stilling well fitted with a capacitance water level recorder (Dataflows Systems LTD, New Zealand) was installed on the first-order stream draining the micro-watershed. Streamflow (Q, mm hr^-1) was computed from the water-level using a stream-specific rating curve and catchment area.

The dataset contains some missing values marked as "NA".