Simulating streamflow in ungauged basins under a changing climate: The importance of landscape characteristics

Highlights

• Significant relationships exist between landscape forms and future streamflow shifts.

• Landscape characteristics control the sensitivity of catchments to climate change.

• Regionalization is useful to predict landuse and climate effects in ungauged basins.

• Virtual experiments reveal the importance of forest cover for future streamflow.

Abstract

In this paper we explored how landscape characteristics such as topography, geology, soils and land cover influence the way catchments respond to changing climate conditions. Based on an ensemble of 15 regional climate models bias-corrected with a distribution-mapping approach, present and future streamflow in 14 neighboring and rather similar catchments in Northern Sweden was simulated with the HBV model. We established functional relationships between a range of landscape characteristics and projected changes in streamflow signatures. These were then used to analyze hydrological consequences of physical perturbations in a hypothetically ungauged basin in a climate change context. Our analysis showed a strong connection between the forest cover extent and the sensitivity of different components of a catchment’s hydrological regime to changing climate conditions. This emphasizes the need to redefine forestry goals and practices in advance of climate change-related risks and uncertainties.

Introduction

Agreement in the science community is wide spread that changing climate conditions in response to increasing atmospheric concentrations of greenhouse gases substantially affect the water cycle (Milly et al., 2005). In the last century, streamflow patterns have undergone profound changes as a consequence of shifts in temperature, atmospheric water vapor and precipitation (IPCC, 2014). A continued warming of oceans and atmosphere in combination with shifting precipitation trends within the next decades (Alexander et al., 2006, IPCC, 2014) will severely perturb regional hydrology towards the end of the 21st century. The available scientific literature on hydrological climate change impacts in different world regions consistently suggests changing amounts of annual river streamflow (Nijssen et al., 2001), shifts in flood peak magnitudes and timing (Hirabayashi et al., 2013) as well as alterations in flow duration curves (Arora and Boer, 2001). Annual streamflow in snow-dominated catchments in mid to higher latitudes is typically projected to increase (Andréasson et al., 2004, Bergström et al., 2001, Graham et al., 2007a, Graham, 2004, Nijssen et al., 2001), while most tropical and mid-latitude basins will likely experience a reduction (Arora and Boer, 2001, Nijssen et al., 2001). But these findings are mostly based on individual catchments (e.g. Graham et al., 2007a) or on comparisons of catchments located in different climate zones (e.g. Bergström et al., 2001 or Nijssen et al., 2001). The question arises as to whether it is reasonable to draw such generalized conclusions in terms of future hydrological changes for ungauged basins within a larger region based on a selection of representative catchment studies. One could argue that nearby catchments within the same climate zone and underlying geology should function in similar ways, which means that conclusions can be drawn for nearby ungauged basins with the same climate and geological conditions. On the one hand, this hypothesis is supported by the fact that spatial proximity has been found to be a good indicator of catchment similarity (Carey et al., 2010, Sawicz et al., 2011), especially in humid runoff-dominated regions (Patil and Stieglitz, 2012). On the other hand, a recent study by Karlsen et al. (2016b) discovered highly variable hydrological functioning of neighboring and rather similar catchments under current climate conditions. The authors reasoned that these variations are controlled by different landscape characteristics such as topography, geology, soils and land cover. In fact, it has long been established that principal streamflow-generating processes are not only controlled by external climatic conditions, but also by physical properties (Beven, 2002, Buttle, 1998, Dunne, 1978, Trancoso et al., 2016). A catchment’s physical landscape characteristics result from perpetual adaptive ecological, geomorphic and land-forming processes (Sivapalan, 2005). These physical attributes heavily influence – together with external climate conditions - the present-day hydrological functioning as well as the variability at multiple temporal and spatial scales (Wagener et al., 2007). Such relationships between hydrological behavior and the landscape have been confirmed in many different parts of the world by a number of studies, strongly suggesting that surface runoff, infiltration, water retention and snowmelt dynamics in a catchment are controlled by

• topographic features such as elevation, aspect or slope (Frisbee et al., 2012, Kokkonen and Jakeman, 2002, Peña-Arancibia et al., 2010, Seyfried and Wilcox, 1995, Williams et al., 2009),

• soil characteristics such as soil types, profiles or depth (Correa et al., 2016, Farmer et al., 2003, Seyfried and Wilcox, 1995, Tetzlaff et al., 2007, Williams et al., 2009),

• catchment area (Buttle and Eimers, 2009, Lyon et al., 2012, Post and Jakeman, 1999),

• land use and land cover (Lyon et al., 2012, Ochoa-Tocachi et al., 2016, Schoonover et al., 2006),

• vegetation type (Brown et al., 2005, Farmer et al., 2003, Ochoa-Tocachi et al., 2016, Seyfried and Wilcox, 1995, Zhang et al., 2001) and

• geology (Correa et al., 2016, Seyfried and Wilcox, 1995).

Consequently, the hydrological behavior of catchments in the same region with the same stationary climate conditions varies due to catchment-specific landscape characteristics. Teutschbein et al. (2015) took these findings one step further and used a diagnostic approach to demonstrate that neighboring catchments in the boreal region do not respond uniformly to the same change in climate conditions. Similar indications for considerably different streamflow responses to a changing climate across subcatchments were also found for other geographical regions, such as North America (Chang and Jung, 2010), Central Europe (Kunstmann et al., 2004) or East Africa (Musau et al., 2015). This emphasizes the complexity of the interlinkages between changing climate conditions, landscape characteristics and hydrological behavior, which make it difficult to generalize potential impacts on other catchments within a larger region.

We claim that landscape characteristics are not only important under stationary climate conditions, but also play a fundamental role for the sensitivity of a catchment to changing climate conditions. Further, we hypothesize that it is possible to predict the combined effects of changes in physical catchment and climate conditions on hydrological processes in ungauged basins given that the underlying functional relationships between landscape characteristics and hydrological shifts in a changing climate can be established.

To explore these theories, the following procedure (Fig. 1) was implemented based on a meso-scale low-land catchment and its associated, partly nested subcatchments in the boreal region in Northern Sweden: After collecting all required data (step 1), present and future hydrological regimes were simulated with help of a hydrological model (step 2). Streamflow signatures were calculated for both present and future streamflow simulations to identify signature changes (step 3). Using a regionalization approach, these signature changes were then linked to catchment-specific landscape attributes with help of functional relationships (step 4). Thereafter, the identified functional relationships were validated, i.e., tested for their ability to provide information about the change in an hypothetically ungauged nearby catchment with known physical catchment characteristics (step 5). Finally, we created two virtual forest change cases (step 6) and made an attempt to predict the combined influence of landuse and climate change in hypothetically ungauged basins. Such an analysis is critical to understanding how watersheds will hydrologically respond to combined landuse transformations and future climatic perturbations, which directly reflects upon our ability to reliably project future streamflow in ungauged basins.