clusterSPcenters

Wrapper for calculating centroids averageSP or medoids medoidSP when clustering, with efficient optimization when iterating kdba clustering calculations

Snow profiles describe the vertical (1D) stratigraphy of layered
snow with different layer characteristics, such as grain type, hardness,
deposition date, and many more. Hence, they represent a data format similar
to multivariate time series containing categorical, ordinal, and numerical
data types. Use this package to align snow profiles by matching their
individual layers based on Dynamic Time Warping (DTW). The aligned profiles
can then be assessed with an independent, global similarity measure that is
geared towards avalanche hazard assessment. Finally, through exploiting data
aggregation and clustering methods, the similarity measure provides the
foundation for grouping and summarizing snow profiles according to similar
hazard conditions. In particular, this package allows for averaging large
numbers of snow profiles with DTW Barycenter Averaging and thereby
facilitates the computation of individual layer distributions and summary
statistics that are relevant for avalanche forecasting purposes.
For more background information refer to Herla, Horton, Mair,
and Haegeli (2021) <doi:10.5194/gmd-14-239-2021>, Herla, Mair, and Haegeli
(2022) <doi:10.5194/tc-16-3149-2022>, and Horton, Herla, and Haegeli (2024)
<doi:10.5194/egusphere-2024-1609>.

Florian Herla

sarp.snowprofile.alignment

Snow Profile Alignment, Aggregation, and Clustering

Pascal Haegeli

Simon Horton

Paul Billecocq

SFU Avalanche Research Program 

clusterSPcenters function

<dl><dt>SPx</dt>
<dd>a sarp.snowprofile::snowprofileSet to be clustered</dd>
<dt>clustering</dt>
<dd>vector of integers (from 1:k) indicating the cluster to which each snow profile is allocated</dd>
<dt>config</dt>
<dd>a list providing the necessary hyperparameters for distance and average calculations. Use clusterSPconfig functions for convenience!</dd>
<dt>centers</dt>
<dd>type of center to determine, either <code>centroids</code> (default) where an average profile is computed for each cluster
or <code>medoids</code> where the index of the medoid profile is identified</dd>
<dt>clustering_old</dt>
<dd>same as <code>clustering</code> but from the previous iteration of a kdba (if provided then this function only computes new centroids when the profiles within a cluster changed)</dd>
<dt>cents</dt>
<dd>a sarp.snowprofile::snowprofileSet of centroids from the previous iteration; provide <code>NULL</code> if not available</dd>
<dt>distmat</dt>
<dd>a precomputed distance matrix of class dist (for faster medoid calculations)</dd></dl>

Arguments

Author

Compute centroids/medoids for clustered snow profiles — clusterSPcenters

<dl>

<dt>SPx</dt>
<dd>a sarp.snowprofile::snowprofileSet to be clustered</dd>


<dt>clustering</dt>
<dd>vector of integers (from 1:k) indicating the cluster to which each snow profile is allocated</dd>


<dt>config</dt>
<dd>a list providing the necessary hyperparameters for distance and average calculations. Use clusterSPconfig functions for convenience!</dd>


<dt>centers</dt>
<dd>type of center to determine, either <code>centroids</code> (default) where an average profile is computed for each cluster
or <code>medoids</code> where the index of the medoid profile is identified</dd>


<dt>clustering_old</dt>
<dd>same as <code>clustering</code> but from the previous iteration of a kdba (if provided then this function only computes new centroids when the profiles within a cluster changed)</dd>


<dt>cents</dt>
<dd>a sarp.snowprofile::snowprofileSet of centroids from the previous iteration; provide <code>NULL</code> if not available</dd>


<dt>distmat</dt>
<dd>a precomputed distance matrix of class dist (for faster medoid calculations)</dd>

</dl>

clusterSPcenters: Compute centroids/medoids for clustered snow profiles

Description

Usage

Value

Arguments

Author

See Also