Description
As part of an ongoing effort to update the ground snow load maps in the United States, this paper presents an investigation into snow densities for the purpose of predicting ground snow loads for structural engineering design with ASCE 7. Despite their importance, direct measurements of snow load are sparse when compared to measurements of snow depth. As a result, it is often necessary to estimate snow load using snow depth and other readily accessible climate variables. Existing depth-to-load conversion methods, each of varying complexity, are well suited for snow load estimation for a particular region or station network, but none are consistently effective across regions and station networks. This paper proposes a random forests regression model for estimating annual maximum snow loads in the conterminous United States that makes use of climate reanalysis data and overcomes the limitations of existing methods. The effectiveness of the random forest model is demonstrated through accuracy comparisons of existing depth-to-load conversion techniques using a compilation of national and state-level data sources. The accuracy comparisons show that the random forest model is competitive for all regions and station networks, while other methods are competitive for only certain regions or station networks. These results highlight the feasibility of developing a single depth-to-load conversion method that appropriately characterizes region and climate specific differences in the snow depth/load relationship across the conterminous United States. Such universal models are an essential component for creating a unified set of national snow load requirements that eliminate the case study regions currently defined in ASCE 7.
Author ORCID Identifier
Jesse Wheeler https://orcid.org/0000-0003-3941-3884
Brennan Bean https://orcid.org/0000-0002-2853-0455
OCLC
1266168095
Document Type
Dataset
DCMI Type
Dataset
File Format
.csv, .r, .rds
Viewing Instructions
The Random Forest described in Wheeler, et al. (in press), is provided in the file RF.RDS. Also included is an example R script entitled Rexample.R. To run the example, download all files into the same folder and set your working directory in R to be the same as the location of these files. Further instructions and comments are given as R comments in the previously mentioned script. Dataset can be downloaded as a complete .zip file, or as the individual files in their original formats.
Publication Date
8-27-2021
Funder
American Society of Civil Engineers / Structural Engineering Institute
Publisher
Utah State University
Award Number
American Society of Civil Engineers / Structural Engineering Institute 202827
Methodology
See ReadMe file.
Referenced by
Wheeler, J., Bean, B., & Maguire, M. (2022). Creating a universal depth-to-load conversion technique for the conterminous United States using random forests. Journal of Cold Regions Engineering, 36(1), 04021019. https://doi.org/10.1061/(ASCE)CR.1943-5495.0000270
Language
eng
Code Lists
See ReadMe file.
Disciplines
Applied Mathematics | Applied Statistics
License
This work is licensed under a Creative Commons Attribution 4.0 License.
Identifier
https://doi.org/10.26078/NKDT-T278
Recommended Citation
Wheeler, J., Bean, B., & Maguire, M. (2021). Supplementary files for "Creating a Universal Depth-To-Load Conversion Technique for the Conterminous United States using Random Forests" [Data set]. Utah State University. https://doi.org/10.26078/NKDT-T278
Checksum
0c095b439d634fac07d7e54dad0687bc
Additional Files
data.csv (6881 kB)MD5: 84efb967655e961ad77d36a5423b9875
README.txt (7 kB)
MD5: 05aedc1715b8b37bb2e1d115ba23bed5
Rexample.R (2 kB)
MD5: ea5c821530ab1e2663db79bda0675268
RF.RDS (3139 kB)
MD5: 17426dbc4462d7d9d54ff755e194704a
Comments
Files:
1. data.csv This file contains the data that was used to train and test the random forest. This data includes the following: Variables:
- ID: Station code along with identifier of which network the station can be found in.
- NAME: Name of the measurement station.
- STATE: State in which the measurement station is located.
- ELEVATION: Elevation, in M, of the measurement station.
- LONGITUDE: Coordinate of the measurement station.
- LATITUDE: Coordinate of the measurement station.
- YEAR: Year in which the observations of annual maximum snow depth and snow water equivalent (SWE) are considered.
- maxv_SNWD: Annual maximum value of snow depth, in mm.
- maxv_WESD: Annual maximum value of SWE, in mm.
- maxm_SNWD: The month in which the maximum value of snow depth was measured.
- maxm_WESD: The month in which the maximum value of SWE was measured.
- maxRatio: The ratio maxv_WESD/maxv_SNWD, which is the response variable in the random forest.
- dist2coast: Distance between measurement station and nearest coast (ocean or great lake) measured in M.
- PPTWT: Sum of Winter Precipitation (Dec - Feb) in mm.
- MCMT: Mean Coldest Month Temperature, in Celsius.
- MWMT: Mean Warmest Month Temperature, in Celsius - TD: MWMT
- MCMT - snow_month_SNWD: Month in which the maximum value of snow depth was measured, converted to a new scale with 1 corresponding to October and 9 corresponding with June - data: String indicator of whether the observation was used for training or testing purposes
2. RF.RDS - Contains a compile randomForest object. This object is the Random Forest that was discussed in detail in Wheeler, et al. (in press). Instructions on how to use this file can be found in the file Rexample.R
3. Rexample.R - A sample script that loads the randomForest object RF.RDS, loads the dataset data.csv, and makes SWE predictions at the locations present in the dataset data.csv. The script also shows how one could create a different randomForest object if they so desired, as well as provides a few plotting functions that can be used to perform inference on a randomForest object.