Find research datasets worth reusing

Soil rock fragment content (fragvol; % volume) soil maps of the Upper Colorado River Basin

<p>The data here were originally posted to facilitate timely and transparent peer review. The final public data release with formal metadata is now available from at the following location:</p> <p>Nauman, T.W., and Duniway, M.C., 2020, Predictive soil property maps with prediction uncertainty at 30 meter resolution for the Colorado River Basin above Lake Mead: U.S. Geological Survey data release,<a href="http://https//doi.org/10.5066/P9SK0DO2">&nbsp;https://doi.org/10.5066/P9SK0DO2</a>.</p> <p>Associated publication:</p> <p>Nauman, T. W., and Duniway, M. C., 2020, A hybrid approach for predictive soil property mapping using conventional soil survey data: Soil Science Society of America Journal, v. 84, no. 4, p. 1170-1194.&nbsp;<a href="https://doi.org/10.1002/saj2.20080">https://doi.org/10.1002/saj2.20080</a>.</p> <p>Version 3: Updated cross validation graphs with correct units labeled as ln(%vol+1).</p> <p>Version 2: Unfortunately, errors were found in the original training data preparation in version 1. This version corrects those errors and has resulted in cross validation accuracy increases (R²) from ~0.4 to ~0.55-0.6 for the shallower depths.</p> <p>Repository includes maps of soil rock fragment (&gt; 2mm) content&nbsp;(fragvol) as defined by United States soil survey program.&nbsp;</p> <p>These data are preliminary or provisional and are subject to revision. They are being provided to meet the need for timely best science. The data have not received final approval by the U.S. Geological Survey (USGS) and are provided on the condition that neither the USGS nor the U.S. Government shall be held liable for any damages resulting from the authorized or unauthorized use of the data.</p> <p>This data should be used in combination with a soil depth or depth to restriction&nbsp;layer map (both layers that will be released soon as part of this project)&nbsp;to eliminate areas mapped at deeper depths than the soil actually goes.&nbsp;This is a limitation of this data which will hopefully be updated in future updates.&nbsp;&nbsp;</p> <p>The creation and interpretation of this data is documented in the following article. Please note this article has not been reviewed yet and this citation will be updated as the peer review process proceeds.</p> <p>Nauman, T. W., Duniway, M. C., 2020. A hybrid approach for predictive soil property mapping using conventional soil survey data. Soil Science Society of America Journal.</p> <p>File Name Details:</p> <p>ACCURACY!! Please see manuscript and Github repository (https://github.com/naumi421/SoilReconProps) for full details on accuracy. We do provide cross validation (CV) accuracy plots in this repository for both the overall sample (NRCS field pedons plus NRCS laboratory pedons; file ending _CV_plots.tif) and for just the CV results at laboratory pedons (file ending _CV_SCD_plots.tif). These plots compare CV predictions with observed values relative to a 1:1 line. Values plotted near the 1:1 line are more accurate. Note that values are plotted in hex-bin density scatter plots because of the large number of observations (most are &gt;3000).</p> <p>Elements are separated by underscore (_) in the following sequence:</p> <p>property_r_depth_cm_geometry_model_additional_elements.extension</p> <p>Example: fragvol_r_0_cm_2D_QRF_bt.tif&nbsp;</p> <p>Indicates soil rock fragment content (fragvol) at 0 cm depth using a 2D model (separate model for each depth) employing a quantile regression forest. This file is the raster prediction map for this model. There may be additional GIS files associated with this file (e.g. pyramids) that have the same file name, but different extensions. The _bt indicates that the map has been back transformed from ln or sqrt transformation used in modeling.</p> <p>**also for 15 and 30 cm, some&nbsp;files also had &quot;_ART_SG100covs_bt.tif&quot;: these are the property prediction layers for those depths. These irregularities are&nbsp;due to file names not getting updated before upload, and issues creating a new version with corrected filenames (3/26/2019).</p> <p>The following elements may also exist on the end of filenames indicating other spatial files that characterize a given model&#39;s uncertainty (see below).</p> <p>_95PI_h: Indicates the layer is the upper 95% prediction interval value.</p> <p>_95PI_l: Indicates the layer is the lower 95% prediction interval value.</p> <p>_95PI_relwidth: Indicates the layer is the 95% relative prediction interval (RPI). The RPI is a standardization of the prediction interval that indicates that model is constraining uncertainty relative to the original sample. RPI values less than one represent uncertainty is being improved by the model relative to the original sample, and values less than 0.5 indicate low uncertainty in predictions. See paper listed above and also Nauman and Duniway (In revision) for more details on RPI.</p> <p>References</p> <p>&nbsp;Nauman, T. W., and Duniway, M. C., In Revision, Relative prediction intervals reveal larger uncertainty in 3D approaches to predictive digital soil mapping of soil properties with legacy data: Geoderma</p>