Probabilities of stream permanence for the contiguous United States based on drainage area

This dataset contains estimated probabilities of detecting at least of n zero days of streamflow for river locations across the contiguous United States as a function of drainage area. Also included are the 95-percent confidence intervals around the probability estimates. Probabilities are estimated by survival regression models, which are simple, non-parametric, empirically-based models that assume a decreasing probability of observing at least n days of zero streamflow as drainage area increases.

For each value of n, survival regression models are estimated for 19 modeling units within the contiguous United States. These modeling units are assumed to contain similar physiographic and hydroclimatic stream characteristics that are not yet available for stream reaches in the 3-Dimensional Hydrography Program (3DHP). A map containing the location of the modeling units, the streamgages and streamflow values used to develop the models, and an explanation of how the modeling units were determined can be found in the model use case located at https://water.usgs.gov/catalog/datasets/2e5c17d8-da95-4721-8f6b-345c063aa15a/. Also included in the model use case are noted limitations of the probability estimates for each n value and modeling unit. Particularly, for higher n values and modeling units with few observed zero streamflow days, survival regression may not be appropriate. Further, for some modeling units and n days, no zero streamflow values were observed in the training dataset and, therefore, probabilities and confidence intervals are not available.

Each child item contains modeling-unit-scale look-up tables describing the probabilities of detecting at least n days of zero streamflow, with n equal to 1, 3, 7, and 30 days of zero streamflow. There are 19 files within each child item with one file per modeling unit. Files are named with the convention “km.table_XX_YYday.txt” where XX is the 2- or 3-alphanumeric code for the modeling unit, and YY is n, the number of zero days of streamflow. If a file is missing for a modeling unit and/or a particular n day, there were no observed zero streamflow values from which to develop a model.