The following reference (the link provides a free pdf) and associated Conclusions (which I like more than the abstract), emphasizes the importance of kilometer scale variations in ice shelf melting. This work indicates that averaging these kilometer scale variations results in non-conservative (w.r.t. public safety) results regarding the risk of accelerating SLR:
http://www.the-cryosphere-discuss.net/7/1591/2013/tcd-7-1591-2013.pdfPine Island Glacier ice shelf melt distributed at kilometre scalesby: P. Dutrieux, D. G. Vaughan, H. F. J. Corr, A. Jenkins, P. R. Holland, I. Joughin, and A. Fleming; The Cryosphere Discuss., 7, 1591–1620, 2013;
www.the-cryosphere-discuss.net/7/1591/2013/; doi:10.5194/tcd-7-1591-2013
Conclusions:
"Previous work has indicated high melt-rates near the grounding line of PIG (Payne et al., 2007; Rignot and Jacobs, 2002) and the presence of basal channels in the ice (Bindschadler et al., 2011; Mankoff et al., 2012; Vaughan et al., 2012). Our observations show that the pattern of melting on PIG ice shelf is highly complex. Within 10km of the grounding line, the melt rate is at least 100myr
−1. Only 20 km downstream this reduces to 30myr
−1. Between 2008 and 2011, basal melting was largely compensated by ice advection, allowing us to estimate an average loss of ice to the ocean of 87 km3 yr
−1, in close agreement with 2009 oceanographically-constrained estimates. Close to the grounding line, melting is concentrated in the basal channels and carves out those channels at 80myr
−1. Further downstream, melting on the keels is 30myr
−1 faster than in the channels, which explains the gradual loss of channels in the downstream part of the ice shelf and the inversion of the surface elevation anomalies relative to free floatation. The gradual regime shift in channel melt could be explained by the initial formation, near the grounding line, of buoyant meltwater plumes rising up the ice base and most efficiently entraining heat to the channel crests, and a decrease in the heat entrainment efficiency downstream as the slope weakens, the ice base shallows and the warm water source gets further away. At some stage, the plumes within the channels deliver less heat to the ice shelf than the warmer deeper waters bathing the channel keels. With the advent of ice surface DEMs of even higher resolution (few meters) taken at regular time intervals, we can expect that the methodology developed here will reveal unforeseen details about the distribution of surface elevation changes and by inference of basal melt where the underlying assumptions are valid, thereby increasing our understanding of atmosphere-ice-ocean interaction dynamics and their temporal and spatial variability.
Our observations of the area close to the grounding line therefore indicate melt rates that are 80% higher in channels than on neighbouring keels, and point to high spatial variability in the melt-rates across the ice shelf, indicating strong modulation of ice-ocean interactions at kilometre scales. This implies that in-situ observations need to be interpreted within their contextual position relative to the channels. Possibly the most important implication of this work concerns the modelling of sub-ice shelf cavities. Accurately representing sub-kilometre scales using conventional ocean models is challenging even for dedicated regional studies, and will remain impossible for global coupled climate models for some time to come. One approach to solving this problem is to use unstructured computational meshes to focus the model resolution on features of interest, such as these channels (Kimura et al., 2013; Timmermann et al., 2012). A more conventional alternative would be to parameterise their effect on the larger scales that models are able to resolve. For either of these approaches to be successful, an essential prerequisite is a detailed observational understanding of the channels, for which the present study provides a significant advance."