Discussion of uncertainties and comparison to measured elevation changes

The variation of HAR precipitation as input for the MB model through the application of three differ-ent scaling factors (0.56±0.25) creates a range of uncertainty for every SEB and MB compondiffer-ent (see section 5.4). The overall uncertainty for the annual MB over the simulated period 2000-2011 reveals an inter-annual variability depending on the sensitivity of the different parameters on the precipita-tion amount (Fig. 5.6). The determinaprecipita-tion of the ‘correct’ scaling factor is difficult because in-situ measurements or remote sensing analyses for the evaluation are scarce. The comparison of simula-ted runoff and the occurrence of GLOF events in section 5.4.1 reveal that the inter-annual pattern of the input parameters is not in gross contradiction with observations but cannot solely explain the ob-served lake outbursts. The overall simulated MB for the reference run is less negative than the aver-age published by various authors for glaciers in the Himalayas, whereas the lower estimate with a precipitation scaling factor of 0.31 seems to be slightly too negative (see section 5.4.1). Furthermore, the MB model calculates a refreezing amount of 56% for the reference run what is far above the ob-served values for cold glaciers (Fujita & Ageta 2000, Fujita et al. 2007). Due to its location on the TP, the thermal regime of Halji glacier is assumed to be temperate to polythermal (see section 1.4.1).

The strong refreezing can be explained by a long-lasting snow cover on the glacier, revealed by the high albedo values (Fig. 5.4g) and the low snow lines (Fig. 5.7). With the lower precipitation estimate only 7% of surface and subsurface melt refreezes (Table 5.2). This value is more reasonable.

Kropacek et al. (2013) calculated surface elevation differences between 2004 and 2009 at Halji gla-cier from ICESat (Fig. 5.10). The obtained values also cover neighbouring glagla-ciers and are only valid for the indicated profiles. We chose the two ICESat profiles on Halji glacier that nearly overlap spati-ally (Fig. 5.10a). This was until recently the only data available for validation at Halji glacier. The mean elevation difference for the profiles for the period from November 2003 to October 2007 is ≈-4 m (Fig. 5.10b). The surface height change determined by the MB model for a similar profile over the period 1 October 2003 – 30 September 2007 is +0.8 m for the reference run and -3.1 m for the model run with 25% decreased precipitation (Table 5.3).

Fig. 5.10: Surface elevation differences at Halji glacier and its surroundings from ICESat, 2004-2009 (Kropacek et al. 2013); (a) Halji glacier (red outline) with ICESat tracks (coloured dots) and the selected profile for the comparison with the MB model output (blue dashed line); the MB profile along the central flow line is indicated as a grey dashed line as also indicated in Fig. 5.1; (b) average elevation differences de-termined from the ICESat tracks; the red lines indicate the two similar profiles in 2003 and 2007.

This finding supports the assumption that a precipitation scaling factor of 0.56 clearly overestimates the total precipitation amounts at Halji glacier when keeping Tair unchanged. A scaling factor of 0.31

leads to better results but mass loss seems still too low compared to the ICESat profiles. Considering the observed cold bias in HAR Tair at Naimona’nyi glacier (see section 1.4.1 and chapter 4), it is very likely that a similar offset is evident also at Halji glacier. Thus, the variation of HAR precipitation am-ounts needs to be complemented by variations of Tair. To limit the computational effort the MB mo-del with varying combinations of precipitation scaling factors and temperature offsets is run for a north-south profile along the western ICESat profiles that cross Halji glacier (blued dashed line in Fig.

5.10a). The results for the modelled surface height change between 1 October 2003 and 30 Sep-tember 2007 are given in Table 5.3. It is obvious that several combinations of precipitation scaling factors and air temperature offsets lead to reasonable results. This finding confirms the assumption made for Naimona’nyi glacier that the chosen combination is only one of several solutions (see sec-tion 4.5). Finally, we chose a precipitasec-tion scaling factor of 0.56 and a Tair offset of +3 K for further analysis because we assume Tair to be distributed rather homogenously over short distances. At Nai-mona’nyi the comparison of HAR Tair with AWS measurements revealed a cold bias of ≈+3 K. Further-more, the results of the MB model run with the respective temperature offset are in good agreement with other data available for Naimona’nyi glacier (see section 4.5). We assume a similar Tair offset to be evident also at Halji glacier.

The results presented in Table 5.3 indicate that the rate of surface height decrease strongly increases with decreasing precipitation and increasing Tair (Table 5.3). This physical response is possibly depen-ding on α that strongly impacts surface melt.

Table 5.3: Surface height change in m for the period 1 October 2003 – 30 September 2007 for the ICESat pro-file at Halji glacier (Fig. 5.10a) for several precipitation scaling factors and air temperature offsets. The green box indicates the combination that is chosen for further analysis.

±0K +1K +2K +3K

0.56 +0.8 -0.3 -2.4 -4.0

0.43 -0.1 -1.6 -4.3 -13.9

0.31 -3.1 -5.5 -15.1 -24.8

Based on the combination of precipitation scaling factor and Tair offset determined from the ICESat profile glacier-wide SEB and MB components are calculated by the MB model. The glacier-wide MB estimate in the model run with increased Tair by 3 K for the period 2001-2011 is -19765 kg m^-2 (-1976.5 kg m^-2 yr^-1). Recently, Nicolai Holzer from TU Dresden determined surface elevation changes and geodetic MB for Halji glacier between February 2000 and October 2013 from a DEM generated from optical tri-stereo Pleiades data (2013) relative to a SRTM DEM (2000) (personal communication, 2nd July 2014). An average ice density of 850±60 kg m^-3 was applied for the conversion into mass changes. N. Holzer estimated a MB of -380±300 kg m^-2 yr^-1 for the period 2000-2013 with a total sur-face height change of -5.9±4.43 m (-0.45±0.35 m yr^-1). These results indicate that mass loss calculated from the MB model with a parameter combination determined from the ICESat profile is too large compared to the remote sensing study of N. Holzer. In order to find a combination of precipitation scaling factor and air temperature offset that leads to a similar MB as determined by N. Holzer we calculated glacier-wide SEB and MB for the combinations 0.56/+2 K and 0.56/+1 K. The resulting an-nual average MB is -1075 kg m^-2 yr^-1 for the combination 0.56/+2 K and -374 kg m^-2 yr^-1 for the com-bination 0.56/+1 K. The latter comcom-bination leads to a total surface elevation change of -7.2 m (-0.72 m yr^-1). The results for MB and surface elevation changes of the MB model within this study are similar to the results determined by N. Holzer when assuming a precipitation scaling factor of 0.56 and a temperature offset of +1 K. The spatial comparison of calculated surface height changes from this study and the remote sensing analyses from N. Holzer reveals that the overall pattern is captured to a high degree (Fig. 5.11). However, several combinations of precipitation scaling factors

and air temperature offsets might lead to reasonable results. Therefore, further calculations with varying parameter combinations will be carried out in the near future.

Fig. 5.11: Spatial comparison of modelled surface height change of Halji glacier (precipitation scaling factor 0.56 / Tair +1 K), 2001-2011 (left), with the results provided by N. Holzer, TU Dresden (right) (difference Pleiades DEM 2013 / SRTM DEM 2000).

To visualize the effects of a temperature offset over all altitudes we calculated a northeast-south profile along the glacier flow-line (see Fig. 5.1 and Fig. 5.10a). MB and surface height change for the previously discussed model runs with three precipitation scaling factors and unchanged Tair are compared to the three model runs with varying temperature offsets (Fig. 5.12). The ICESat profile in-tersects the northeast-south profile in ≈5420 m a.s.l. (Fig. 5.10a and dotted line in Fig. 5.12).Similar to the findings at Naimona’nyi MB differences are largest in the lower glacier areas (Fig. 5.12). From the MB profiles and the corresponding DEM altitudes a rough estimation of the ELA can be provided.

The MB model run for the northeast-south profile with a precipitation scaling factor of 0.56 and a temperature offset of +1 K results in a mean ELA between 2001 and 2011 of ≈5460 m a.s.l. (Fig. 5.12).

This value coincides with the ELA determined directly from the glacier-wide MB (5467 m a.s.l.).

The resulting glacier-wide SEB and MB components from the MB model run with a precipitation scaling factor of 0.56 and a Tair offset of +1 K are given in Fig. 5.13. The inter- and intra-annual pat-terns of the different parameters are similar to those explained in section 5.4.1. For the considered period 2001-2011, SWin (+240.8 W m^-2) and LWin (+209.3 W m^-2) dominate energy input followed by Qsens (+21.9 W m^-2). Energy sinks at the glacier surface are LWout (-257.6 W m^-2), SWout (-192.9 W m^-2), Qlat (-15.3 W m^-2), Qmelt (-8.8 W m^-2) and QG (-1.9 W m^-2). This results in a SWnet of +47.9 W m^-2. In general, the contributions of the single SEB components to the total energy flux change only slightly compared to the model run with unchanged Tair (Table 5.1). SWnet accounts for 36%, followed by LWnet (35%), Qsens (15%), Qlat (11%) and QG (3%).

Fig. 5.12: MB and surface height change for 2001-2011 along a northeast-south profile at Halji glacier for HAR precipitation and temperature offsets. The location of the profile is indicated in Fig. 5.1. The altitude in which the MB is zero (dashed lines) is the ELA. The profile of the chosen ICESat tracks (Fig. 5.10a) intersects the northeast-south profile in ≈5420 m a.s.l. (dotted line).

The glacier-wide MB estimate in the model run with increased Tair by 1 K for the period 2001-2011 is -3741 kg m^-2 (-374.1 kg m^-2 yr^-1). Surface melt (-8302 kg m^-2/-830.2 kg m^-2 yr^-1) is the largest factor of glacier-wide mass loss followed by sublimation (-2093 kg m-²/-209.3 kg m^-2 yr^-1) that dominates abla-tion in winter, when air temperatures are below 0°C and surface melt is absent (Fig. 1.12, 5.13b).

Subsurface melt (-37 kg m^-2/-3.7 kg m^-2 yr^-1) plays a minor role. Solid precipitation (+4462 kg m^-2/ +446.2 kg m^-2 yr^-1) and refreezing (+2229 kg m^-2/ +222.9 kg m^-2 yr^-1) contribute to the mass gain of the glacier. The amount of solid precipitation is less compared to the MB model run with the same sca-ling factor but unchanged Tair (Table 5.2). This is because the solid proportion of total precipitation decreases with increasing Tair. Over the total simulation period, effective melt (surface melt + subsur-face melt – refreezing) accounts for 74% and sublimation for 26% of the total mass loss. In total, 27%

of the surface and subsurface melt refreeze.

Glacier-wide annual MB and runoff as calculated from the MB model (0.56/+1 K) are shown in Fig.

5.14 and compared to the occurrence of GLOF. The results are similar to the respective pattern with unchanged Tair (Fig. 5.6).

Fig. 5.13: Glacier-wide monthly (a) SEB components (see section 2.3.1 for abbreviations) and (b) MB compo-nents from October 2000 to September 2011 at Halji glacier (precipitation scaling factor 0.56 / Tair

+1 K). Note that the first months (blue box) should not be considered.

Fig. 5.14: Glacier-wide annual MB and runoff for MB years (October – September) 2000-2011 estimated by the HAR forced MB model for Halji glacier (precipitation scaling factor 0.56 / T_air +1 K). The black dots mark the ‘non GLOF years’ (Kropacek et al. (2013)). Note that the MB in 2000/01 (blue box) should not be considered.