Snow- and glacier-fed river basins

SPHY is being used in large Asian river basins with significant contribution of glacier melt and snowmelt to the total flow (Immerzeel et al., 2012, Lutz et al., 2012, 2014a,2016). The major goals of these applications are two-fold:

• Assess the current hydrological regimes at high resolution; e.g., assess spatial differences in the contributions of glacier melt, snowmelt and rainfall–runoff to the total flow.

• Quantify the effects of climate change on the hydrological regimes in the future and how these affect the water availability.

Rivers originating in the high mountains of Asia are considered to be the most meltwater-dependent river systems on Earth (Schaner et al. 2012). In the regions surrounding the Himalayas and the Tibetan Plateau, large human populations depend on the water supplied by these rivers (Immerzeel et al., 2010). However, the dependency on meltwater differs strongly between river basins as a result of differences in climate and differences in basin hypsometry (Immerzeel and Bierkens 2012). Only by using a distributed hydrological modeling approach that includes the simulation of key hydrological and cryospheric processes, and inclusion of transient changes in climate, snow cover, glaciers and runoff, can appropriate adaptation and mitigation options be developed for this region (Sorg et al. 2012). The SPHY model is very suitable for such goals, and has therefore been widely applied in the region.

For application in this region, SPHY was set up at a 1km spatial resolution using a daily time step, and forced with historical air temperature (, , ) and precipitation data, obtained from global and regional data sets (e.g., APHRODITE, (Yatagai et al. 2012); Princeton, (Sheffield, Goteti, and Wood 2006); TRMM, (Gopalan et al. 2010)) or interpolated WMO station data from a historical reference period. For this historical reference period, SPHY was calibrated and validated using observed streamflow. For the future period, SPHY was forced with downscaled climate change projections obtained from general circulation models (GCMs), as available through the Climate Model Intercomparison Projects (e.g., CMIP3, (Meehl et al. 2007); CMIP5, (Taylor et al., 2012)), which were used as a basis for the Assessment Reports prepared by the Intergovernmental Panel on Climate Change (IPCC).

In central Asia, SPHY was applied in a study (ADB 2012; Immerzeel et al., 2012; Lutz et al., 2012) that focused on the impacts of climate change on water resources in the Amu Darya and Syr Darya river basins. SPHY was used to quantify the hydrological regimes in both basins, and subsequently to project the outflow from the upstream basins to the downstream areas by forcing the model with an ensemble of five CMIP3 GCMs. The SPHY model output fed into a water allocation model that was set up for the downstream parts of the Amu Darya and Syr Darya river basins.

In the Himalayan Climate Change Adaptation Programme (HICAP), led by the International Centre for Integrated Mountain Development (ICIMOD), SPHY has been successfully applied in the upstream basins of the Indus, Ganges, Brahmaputra, Salween and Mekong rivers (Lutz et al. 2013; Lutz et al. 2014a). In this study the hydrological regimes of these five basins have been quantified and the calibrated and validated model (Figure 6) was forced with an ensemble of eight GCMs to create water availability scenarios until 2050. Table 3 lists the calibration and validation results. Based on the validation results, we concluded that the model performs satisfactorily given the large scale, complexity and heterogeneity of the modeled region and data scarcity (Lutz et al. 2014a). We use one parameter set for the entire domain, which inherently means some stations perform better than others. In the particular case of the upper Indus, another possible explanation could be uncertainty in air temperature forcing in the highest parts of the upper Indus basin (locations Dainyor bridge, Besham Qila and Tarbela inflow in Table 3), since especially in this area, the used forcing data sets are based on very sparse observations. SPHY allowed the assessment of the current contribution of glacier melt and snowmelt to total flow (Figure 7), and how total flow volumes and the intra-annual distribution of river flow will change in the future (Lutz et al. 2014a).

Figure 8: Average monthly observed and SPHY-simulated flow (1998-2007) for the Chatara major discharge measurement location in the Ganges basin (Lutz et al. 2014a). Metrics are calculated based on monthly timesteps.

Table 3: Station locations used for calibration and validation of the SPHY model in HICAP (Lutz et al., 2014a). Three stations were used for calibration for 1998-2007). Five stations were used for an independent validation for the same period. The Nash-Sutcliffe efficiency (NS) and bias metrics were calculated at a monthly time step.

Figure 9: The contribution of glacier melt (a), snowmelt (b), and rainfall (c) to the total flow for major streams in the upstream basins of the Indus, Ganges, Brahmaputra, Salween and Mekong during 1998-2007 (Lutz et al. 2014a).

For basins with snowmelt being an important contributor to the flow, besides calibration to observed flow, the snow-related parameters in the SPHY model can also be calibrated to observed snow cover. For the Upper Indus basin, the snow-related parameters degree-day factor for snow () and snow water storage capacity (SSC) were calibrated independently using MODIS snow cover imagery (Lutz et al., 2016). The same MODIS data set was used as in Immerzeel et al. (2009). From the beginning of 2000 until halfway through 2008, the snow cover imagery was averaged for 46 different periods of 8 days (5 days for the last period) to generate 46 different average snow cover maps. For example, period 1 is the average snow cover for 01–08 January for 2000 until 2008, whereas period 2 is the average snow cover for 09–16 January for 2000 until 2008, etc. The SPHY model was run for 2000–2007 at a daily time step and, for each km grid cell, the average snow cover was calculated for the same 46 periods as in the MODIS observed snow cover data set. Subsequently, these simulated snow cover maps were resampled to 0.05 spatial resolution, which is the native resolution of the MODIS product. Figure 8 shows the basin-average observed and simulated fractional snow cover for the 46 periods during 2000–2007 and Figure 9 shows the same at the 0.05 grid-cell level. As a final step, the baseflow recession coefficient () and routing coefficient (kx) were calibrated to match the simulated streamflow with the observed streamflow.

Figure 10: Observed and simulated average fractional snow cover in the upper Indus basin. The values represent the 9-year average for 46 (8-day) periods during 2000-2007.

Figure 11: (a) SPHY simulated snow cover 2000-2007 and (b) MODIS observed snow cover 2000-2007.