Introduction
Transit time distributions (TTD) are a fundamental characteristics of catchment hydrological function, which can indicate the dynamics of water movement and solute transport in watersheds (Kirchner et al., 2000; McDonnell and Beven, 2014). The TTD gives conceptual, integrated understanding of the nature of flow paths that transform precipitation inputs (e.g. rainfall, snowmelt) to runoff at the catchment outlet, and their associated temporal dynamics. Mean transit time (MTT) defines the average travel time between water entering and leaving a catchment, which provides indication of water movement under different meteorological condition and for contrasting watershed characteristics (Hrachowitz et al., 2009). The MTT can be determined from the TTD; however, estimates of MTT can often be biased, especially for spatially heterogeneous catchments with heavy tailing of old water distributions (Kirchner, 2016a; Seeger and Weiler, 2014). Consequently, the young water fraction (Fyw), i.e. the average fraction of stream flow that is younger than a specified threshold age (typically between two and three months), has recently been proposed as a more reliable and stable descriptive metric of the TTD for spatially heterogenous and non-stationary conditions by Kirchner (2016a, b). A distinct advantage of this approach is that the estimation of Fyw does not need to assume a particular shape of the underlying TTD.
Currently, there are two main broad approaches for water transit time and age estimation: data-driven and model-driven approaches (McCallum et al., 2014). In the former approach, TTDs are estimated by fitting conservative tracer concentrations using inverse lumped parameter models, and then MTTs are determined from corresponding TTDs (McGuire and McDonnell, 2006; Peralta-Tapia et al., 2016). In the latter approach, the water transit times and age can be tracked directly by a tracer-aided hydrological model with calibrated parameters related to flow and tracer simulations (Hrachowitz et al., 2013; Soulsby et al., 2015; Remondi et al., 2018). Although it is difficult for data-driven approaches to constrain the spatial variation of flow transit times and water age distributions within a catchment, the influence of climate and landscape properties can also be assessed by catchment inter-comparison of the MTT and Fyw determined by the data-driven method. For example, comparison of MTT determined by data-driven methods across a range of environmental gradients catchments, Hrachowitz et al. (2009, 2010) and Heidbüchel et al. (2013) helped disentangle the relative influence of controls due to meteorological conditions and watershed characteristics (e.g. drainage density, topographic wetness index, soil cover and storage capacity, antecedent moisture conditions and precipitation event characteristics) on stream water ages. Similarly, the relationships between Fyw and terrain, soil, and land-use indices, as well as the precipitation characteristics have been examined in 22 Swiss catchments (von Freyberg et al., 2018). In addition, Jasechko et al. (2016) calculated the Fyw of streamflow of 254 relatively large catchments around the world, and they found there is no significant correlation between Fyw and annual rainfall, but there is an inverse correlation with the average topographic gradients inferring deeper vertical infiltration in steeper catchments.
Alternatively, where transit times and water ages are tracked using tracer-aided hydrological models (e.g. Benettin et al., 2015a; Hrachowitz et al., 2013, 2016; McMillan et al., 2012), although such models are usually conceptual, they have stronger skill in capturing the spatio-temporal variability of catchment transit times and water age due to non-stationarity conditions and spatial heterogeneity. These approaches have contributed to the enhanced understanding of spatial variation in runoff generation and solute transport processes and have shown how this influences the dynamics of transit times and water ages at the catchment scale (Birkel et al., 2012; Soulsby et al., 2015). Process-based models, with more complex structures and parameterisation, can provide more physically-based descriptions of catchment hydrological processes, and can give more integrated understanding of tracking flow transit times and water ages and analysing how hydrometeorological conditions and spatial heterogeneity may affect the TTD and Fyw (Kuppel et al., 2018a; Remondi et al., 2018). However, the high parameterisation of such models can increase uncertainty, unless detailed data on watershed states (soil moisture storage, groundwater levels etc.) and hydroclimatic inputs are available for multi-criteria model calibration, which limit the application of this method to more intensively instrument catchments (Kuppel et al., 2018b).
As both climate and landscape characteristics interact to determine transit times and water ages, understanding how catchment morphological properties and external meteorological forcing control TTDs and water age remains challenging. Most studies are site specific and focused in humid temperate catchments, so generalization to different geographical regions is rarely possible (Burt and McDonnell, 2015; Birkel and Soulsby, 2015; Maxwell et al., 2016). For example, karst regions cover 12% of the Earth’s surface and are the main source of drinking water to over 25% of the world’s population (Ford and Williams, 2013). However, due to the high spatial variability of the hydrodynamic properties and hydrological connectivity of the karst critical zone, the TTD and water age of catchment water fluxes have significant spatial and temporal variability (Zhang et al., 2019). Unfortunately, there is relatively little research on this issue (see Chen et al., 2018).
Because of the unique nature of karst geology and geomorphology, and characteristic features such as vertical shafts, caves and sinkholes, the spatial heterogeneity of drainage systems is high. The complex underground mixed-flow systems in karst aquifers include low velocity flows within the matrix and small fractures, and high velocity flow within large fractures and conduits (White, 2007; Worthington, 2009), which lead to a highly dynamic spatio-temporal variability of hydrological processes (Bakalowicz, 2005; Ford and Williams, 2013; Hartmann et al., 2014a). Hu et al. (2015) estimated the mean residence time of water at a karst epikarst spring with contributing area less than 1km2 in South China, based on detailed observations of hydrogen and oxygen isotopes. In their study, the MTT in epikarst spring was longer than one year, indicating that the epikarst had poor connectivity and high water retention, and could thus maintain continuous contributions to surface water. Hartmann et al., (2014b) simulated the time-variant transit time distributions of an Austrian karst system using a semi-distributed model, and showed that the variation in transit time in the karst area is very large, and can range from days to several years. However, these studies considered entire karst basins, and there is a need to understand the TTD and Fyw of water fluxes in different geomorphological units (e.g. hillslope and depression) or different mediums (e.g. dual flows in within the matrix and conduits) within karst landscapes, which are crucial for the understanding the interactions of hydrological processes and water quality in drainage waters.
The aim of this study is to address this research gap in the Chenqi catchment in SW China. The specific objectives are: (1) to use the output from a tracer-aided model to quantify the young water fraction (defined according to Kirchner, 2016a, b) of water storages in, and fluxes between, the main compartments of a complex karst landscape; (2) to examine the seasonal inter-relationships between storage and the young water fraction of the dominant water fluxes as hydrological connectivity changed; and (3) to assess the time variance of the water age and travel time distributions between the main seasons using flux tracking.