There is still limited understanding of where stream water originates, their flow paths, how water sources mix, and for how long water transits montane tropical catchments. Here, we used a simple gamma convolution integral model (GM), ensemble hydrograph separation (EHS) and a tracer‐aided model (TAM) to assess runoff generation, mixing processes and water ages in the pristine tropical rainforest Quebrada Grande catchment in Costa Rica. Model simulations are based on a four‐year record (2016–2019) of continuous hydrometric and stable isotope observations. Comparative model tests included multi‐objective calibration (2017–2019) and validation (2016) using stream discharge and isotope data as well as an independent model evaluation using groundwater and soil water isotope data. GM and TAM agreed on the dominance of young water in streamflow that was less than 95 days old for 75% of the study period. The EHS suggested a young water fraction threshold of 12 ± 2 days with a transit time distribution that approximates the best‐fit GM. These short water ages are the result of high annual rainfall even during drier years such as 2019 with 4300 mm/a and consistent quick near‐surface runoff generation with limited mixing. A supra‐regional loss (~55%) of likely older groundwater was detected. The TAM‐based hydrograph separation (streamflow KGE > 0.78, δ2H KGE > 0.90) suggested an average near‐surface water contribution of more than 60% to streamflow emphasizing the dominance of quick flow paths. This tropical rainforest represents one of the quickest streamflow responses of mostly young water of pristine catchments globally.

There is still limited understanding of how waters mix, where waters come from and for how long they reside in tropical catchments. In this study, we used a tracer-aided model (TAM) and a gamma convolution integral model (GM) to assess runoff generation, mixing processes, water ages and transit times (TT) in the pristine humid tropical rainforest Quebrada Grande catchment in central Costa Rica. Models are based on a four-year data record (2016 to 2019) of continuous hydrometric and stable isotope observations. Both models agreed on a young water component of fewer than 95 days in age for 75% of the study period. The streamflow water ages ranged from around two months for wetter years (2017) and up to 9.5 months for drier (2019) years with a better agreement between the GM estimated TTs and TAM water ages for younger waters. Such short TTs and water ages result from high annual rainfall volumes even during drier years with 4,300 mm of annual precipitation (2019) indicating consistent quick near-surface runoff generation with limited mixing of waters and a supra-regional groundwater flow of likely unmeasured older waters. The TAM in addition to the GM allowed simulating streamflow (KGE > 0.78), suggesting an average groundwater contribution of less than 40% to streamflow. The model parameter uncertainty was constrained in calibration using stable water isotopes (δH), justifying the higher TAM model parameterization. We conclude that the multi-model analysis provided consistent water age estimates of a young water dominated catchment. This study represents an outlier compared to the globally predominant old water paradox, exhibiting a tropical rainforest catchment with higher new water fractions than older water.