Recently, well, ok now that I get round to posting this, not so recently, as this happened at the end of last year …. we found a mistake in one of papers:
In the published paper, three debris properties stated to represent “whole-rock” values are inaccurately specified in the published simulations: namely, whole-rock thermal conductivity (0.94 W m-1 K-1), specific heat capacity (948 J kg-1 K-1), and density (1496 kg m-3; cf. Table 1 in Collier et al., 2014). The values for the first two parameters were estimated from whole-rock values of typical facies present on the Miage Glacier, corrected for observed porosity and an estimated moisture content. The thermal conductivity was computed by averaging 25 point measurements on debris at the Miage Glacier in 2005 using the “residual” approach, as explained in Brock et al. (2010). Therefore, all three values represent “effective” debris properties rather than whole-rock properties and include some influence of moisture content and porosity. We incorrectly used these values to calcualte bulk debris thermal conductivity from it, in effect accounting for the voids and void fill material (water/air/ice) twice over.
Careless, but we are only human.
Luckily, science is an ongoing process, and so the deal with finding an error in your work is to publish an erratum, which we have done, and you can read it in full here.
The way we dealt with our mistake was to assess the impact of these inaccurate parameter choices on our published results. To do this, we performed a Monte Carlo simulation using new ranges for the three whole-rock properties, which bracket published values for common rock types. Monte Carlo simulations involve running a numerical model many times (in this case 20,000 times) with different configurations and then looking at the spread of these results as an indication of the impact of the parameter choices on the modelled system behaviour.
The modelled cumulative sub-debris ice melt is increased by a factor of 2–2.5 compared with the published simulations due to the increase in whole-rock thermal conductivity (see below), and the overestimation of sensible heat flux evident in the published paper is reduced using these more correct parameter ranges (e.g., for the case of a simulated dry debris cover this was reduced from 65 W m-2 in the published paper to 10-50 W m-2), which is nice.
A time series of cumulative sub-debris ice melt (kg m-2) for the (a) 2008 and (b) 2011 simulations and the CMB-RES (blue curves) and CMB-DRY (grey) cases. The single lines show the results from the originally published simulations, while the filled polygons indicate the range of solutions from the Monte Carlo simulations.
The main conclusions of our paper still hold true when we do the simulations with the corrected debris properties:
1. Sub-debris ice melt is reduced when moisture is considered, largely due to heat extraction by the latent heat flux and also through changes in the debris thermal properties
2. When moisture is considered in the simulations for 2008, total cumulative mass balance is more negative – for thermal conductivity values below 2 W m-1 K-1 – since surface vapour fluxes compensate reduced sub-debris ice melt. Conversely, considering moisture in the simulations for 2011 reduces the mass loss due to the formation of ice near the base of the debris, which reduces heat transfer to the underlying ice regardless of the debris property specification.