Uncertainty analysis for evaluating the accuracy of snow depth measurements

In comparison with manual observations, bias errors were calculated to show the spatial distribution of snow depth by considering snow depth measured from four snow stakes located on the easternmost side of the site as reference. The bias error of snow stakes on the west side of the site was largest. The uncertainty of all pairs of stakes and the average uncertainty for each base were 1.81 and 1.52 cm, respectively. The 15

The paper "Uncertainity analysis for evaluation the accuracy of snow depth measurements" by Lee et al. presents a systematic investigation on the accuracy of different snow depth measurements. Several types of snow depth sensors (ultrasonic, laser, manual measurements) were obtained at the CARE site during a winter season and the measurements were compared (same types of sensors and different sensors) with statistical methods (bias, error propagation). The paper is understandable and written well. The conclusions drawn are supported by the data. The topic of the manuscript is important for the snow science community and especially for snow hydrology and therefore well suitable for publication in HESS. However, I have some suggestions with -as I believe -could further improve the paper: 1) I think that a paper that aims to analyze uncertainty and accuracy of snow depth C1941

Interactive Comment
Full Screen / Esc

Printer-friendly Version
Interactive Discussion Discussion Paper measurements should also have a look at the measurement precision of the individual sensors. This could be done by repeatedly measuring a flat, regular artificial surface and / or by selecting a short time series when a flat snow surface did not change 2) It is not clear how the periods without snow in the beginning of the season were treated. Were they removed for the analysis? As measuring snow is what matters in this study (and not grass) I think that these periods should be removed and possibly be analyzed separately 3) The results found in the paper should be discussed in more detail. What do the results mean in a broader context and how do they relate to other studies?
Minor comments: Title and abstract sound like "a new method for uncertainty analysis was developed". In fact the authors applied a well-established method to analyze snow depth sensors. I suggest that is expressed more clearly.
P4159 L23 Fischer 2008 is missing in references L5-6 I suggest to include a statement (possibly in the summary) if the analysis showed that these WMO criteria are met by the sensors tested P4160 L5fff: I suggest to clearly define the terminology used: e.g. what exactly is meant by "uncertainty", "random error", "propagation of error" and so on. The different types of errors could be listed and their physical meaning could be provided (e.g. instrumental, random, spatial variability. . .).
P4162 L19,20 MAE and RMSE are not used in the rest of the paper. There is no need to show the formulas.
P4163 L8 I think that spatial variability would be the better term in that context (instead of spatial distribution) L9 include a break after "at each target." L19ff why do you now use x1 and x2 instead of x and y as in the earlier equations? P4163 L5fff: The paper deals with accuracies of different sensors and how they compare. In my opinion, then functionality of those sensors is an essential information to the reader in that context. I therefore suggest including some more information on the sensors (e.g. measurement principles, sensor accuracy, application. . . of each sensor).
P4166 L2; I think that "spatial variability" would be the better expression (instead of distribution) P4167 L2ff: I do not agree: Data quality is mandatory for a study that deals with sensor accuracy and the data must therefore be processed with the best method possible, especially when the manufacturers propose a specific processing. If -as the authors state -a QC method is available, that is believed to perform better, that method should be tested and possibly applied. I do not see reason to postpone that to later SPICE activities! L9 what is meant with "temporal variation" -as stated the manual measurement take about 20 minutes; I do not see the process which causes significant temporal variation during that short period of time-Heavy snow fall might induce some variation during 20 min but if I understood correctly manual measurements have not been performed during snow fall.
P4170 L6; what is meant with "significant" -statistically significant? Please clarify L7 The value stated in the text (10.7 cm) deviates from the one in Table 3 (10.8 cm) L8 I suggest to add, that the random error is expressed by the BRRMSE P4170 L19 This could be investigated by removing all snowfall events from the data and to check how this changes the distributions and the uncertainty?
L 19-22 this section is not sufficiently described. What about the distributions of e.g. FEL 11A -FEL 20 or FEL11A -12A -they look "bimodal", too. Anyway, I think that bimodal is not a very clear description for the distributions. The obvious shifts of the distributions should also be discussed.
L26 spatial variability instead of spatial distribution P4171 L20: The paper does not really "introduce a methodology" but is rather applies an established method. Please reformulate.
L22: No results of MAE and RMSE are shown in the paper L24: spatial variability instead of spatial distribution L4172: The discussion should be enlarged: What do the presented results mean for C1944 the research area and how do they compare to earlier studies? L7: please add that the data were aggregated to 1 min Table1: I suggest adding a column showing the measurement precision as stated by  the producers   Table 2, 3: units are missing