Uncertainty in climate change impacts on Southern Alps river flow: the role of hydrological model complexity

Abstract

Climate change scenario modelling for New Zealand indicates a series of hydrological changes can be expected. Hydrological modelling is a critical tool to assess the likely impacts of future climate change on river runoff. A hydrological model no matter the complexity can be viewed as a simplified representation of the real-world environment. The most comprehensive hydrological models, i.e. fully-distributed (e.g. TopNet and MIKESHE), are generally complex and require large amounts of input data, computer power and time. The use of a semidistributed model like HBV-light to perform essentially the same function can yield a relatively efficient method of scenario hydrological modelling. As such, the aim of this research is to assess the use of the HBV-Light hydrological model to simulate observed river flow data from within the Shotover Catchment, situated in Queenstown, New Zealand, and compare projected changes to future river runoff under climate change conditions to that of a more complex hydrological model (TopNet).

The HBV-Light model performed well in the Shotover Catchment when replicating an observed data set of 13 years (1972 ̶ 1984), returning a monthly NSE of 0.72 for the calibration period and 0.61 for the validation period (1985 ̶ 1997). The use of HBV-Light to simulate river flow under future climate change conditions, and the potential influence such changes could have on New Zealand’s freshwater resources were also assessed. HBV-Light, using an IPCC SRES A1B climate scenario, and an ensemble of 12 GCM’s suited to New Zealand’s climate conditions, projected runoff increases to the Shotover River in the magnitude of 13.43 % for 2040 over a 20-year average (2030 ̶ 2049), and 17.44 % in 2090 (2080 ̶ 2099). In comparison to the more complex model (TopNet) used assess the impact of climate change on river flow for the same future time periods using the same 12 GCM ensemble and IPCC SRES A1B climate scenario, HBV-Light simulates a yearly projection for 2040 that is 3.5 % higher than TopNet, 1.65 % lower in 2090. Monthly comparisons of future runoff show the HBV-Light model drastically over-simulates TopNets assessment of the Shotover Catchment (by 35 % in 2090 for August) and other research in the Clutha Catchment, in some cases by up to 50 %. HBV-Light under-simulated soil moisture by 300 mm/yr-1 and over-simulated AET consistently, but simulated snowmelt relatively well in comparison to TopNet. Therefore, the current research concludes that the role of hydrological model complexity outweighs the role of other modelling uncertainties such as input data. HBV-Light was run with physical input 3 data which is more accurate but limited in the spread of the catchment, whereas TopNet was run using interpolated data (VCSN) which covers the catchments upper reaches but is only an interpolation. TopNets performance was more in line with other research from the Clutha Catchment and its sub-catchments, such as the Matukituki and the Lindis. However, in the absence of large comprehensive data sets (VCSN) required to run complex models such as TopNet, HBV-Light could be an acceptable alternative for assessing future impacts to river runoff under climate change conditions in New Zealand.