m9brady / smp_to_csv Goto Github PK

Need to remove ice features from the force profiles when estimating microstructure.

Assumptions

• Sharp and decrease in force between snow and discrete ice features
• Typically < 1 cm in thickness (find literature to support, from personal exp.)
• Should not interpolate to fill as the thermodynamic conditions are often contrasting between sides (ie ice conducts heat and can limit vapour transport).

The hard coded microstructure coefficients are from Proksch et al 2016 but the paper was written using an older version of the SMP (v3). A function is needed to take in observed SSA and/or Density and minimize bias by iterating the coefficients. The following is needed to do so:

• Create a general function read and format SSA/Density data
• Use xcorr functions to align the data as best possible
• Use an optimization routine to identity best fit coefficients using microstructure()
• Output error metrics and adjusted estimates

When the data is read in it should be filtered with quality flags. Appropriate first method introduced in [1]. Four class system to flag good data, data with linear trends, data with dampened micro-variance, and data with both error types. Should be able to assign flags at the window level when sub-setting is done.

[1] Pielmeier, C., & Marshall, H. P. (2009). Rutschblock-scale snowpack stability derived from multiple quality-controlled SnowMicroPen measurements. Cold Regions Science and Technology, 59(2), 178-184.

We often talk in terms of stratigraphy or layering in snow science. Identifying these units can be as much of an art, as it is science. Providing quantitative tools to identify the indices or depths of these units within the snow volume is essential for addressing of large volumes of SMP data. Some potential approaches:

• Unsupervised classification
• Supervised classification with field data input
• Wavelet decomposition

Instead of using a subset could we just use a bool mask?

Could initiate it as self.mask = filteredArr[:,1] < 0 or have a bit flag at each array index,

This would be useful for masking outliers, ice features and other issues as well.

Useful for filtering out any weird measurements if the SMP instrument reaches the soil surface

Some sites had depths of greater than 120cm, so after a layer of snow was removed another profile was taken at approximately the same positioning. May need to provide SMP profile concatenation/matching functionality for profiles that have been designated as such.