{"accessLevel": "public", "bureauCode": ["010:12"], "contactPoint": {"@type": "vcard:Contact", "fn": "Samuel H Austin", "hasEmail": "mailto:saustin@usgs.gov"}, "description": "Tables are presented listing parameters and fit statistics for 25,453 maximum likelihood logistic regression (MLLR) models describing hydrological drought probabilities at 324 gaged locations on rivers and streams in the Delaware River Basin (DRB). Data from previous months are used to estimate chance of hydrological drought during future summer months. Models containing 1 explanatory variable use monthly mean daily streamflow data (DV) to provide hydrological drought streamflow probabilities for July, August, and September as functions of monthly mean DV from the previous 11 months. Outcomes are estimated 1 to 12 months ahead of their occurrence. Models containing 2 explanatory variables use monthly mean daily streamflow data (DV) and monthly mean precipitation data (P) to provide hydrological drought streamflow probabilities for July, August, and September as functions of monthly mean DV and monthly mean P from the previous October, November, December, January, and February. Outcomes are estimated 5 to 12 months ahead of their occurrence. Models containing 3 explanatory variables use monthly mean daily streamflow data (DV), monthly mean precipitation data (P), and monthly mean maximum daily air temperature (T) to provide hydrological drought streamflow probabilities for July, August, and September as functions of monthly mean DV, monthly mean P, and monthly mean maximum T from the previous October, November, December, January, and February. Outcomes are estimated 5 to 12 months ahead of their occurrence. Models containing 4 explanatory variables use monthly mean daily streamflow data (DV), monthly mean precipitation data (P), monthly mean maximum daily air temperature (T), and monthly mean potential evapotranspiration data (PET) to provide hydrological drought streamflow probabilities for July, August, and September as functions of monthly mean DV, monthly mean P, monthly mean maximum T, and monthly mean PET from the previous October, November, December, January, and February. Outcomes are estimated 5 to 12 months ahead of their occurrence. Explanatory variable selections for multiparameter models were optimized using random forest statistical methods.\nSelected single-parameter and multi-parameter models are provided. Overall correct classification rates tend to improve and models become more complex as the number of model explanatory variables increases from 1 to 4. Parameters for models with 1 explanatory variable are listed in the table labeled: \u201cDRB-1_Variable_Equations.\u201d Parameters for models with 2 explanatory variable are listed in the table labeled: \u201cDRB-2_Variable_Equations.\u201d Parameters for models with 3 explanatory variable are listed in the table labeled: \u201cDRB-3_Variable_Equations.\u201d Parameters for models with 4 explanatory variable are listed in the table labeled: \u201cDRB-4_Variable_Equations.\u201d      \nParameters describing models containing 1 explanatory variable may be used to populate drought probability equations as follows: p =1/[1 + e^-(\u03b20+ \u03b21\u2022 DV)] where: e is the base of the natural logarithm, \u03b20 is an intercept parameter, \u03b21 is a slope parameter, DV is a factor variable describing monthly mean daily streamflow (ft3/s). \nParameters describing models containing 2 explanatory variables may be used to populate drought probability equations as follows: p =1/[1 + e^-(\u03b20+ \u03b21\u2022 DV+ \u03b22\u2022 P)] where: e is the base of the natural logarithm, \u03b20 is an intercept parameter, \u03b21 is a slope parameter, \u03b22 is a slope parameter, DV is a factor variable describing monthly mean daily streamflow (ft3/s), P is a factor variable describing monthly mean precipitation (in/day). \nParameters describing models containing 3 explanatory variables may be used to populate drought probability equations as follows: p =1/[1 + e^-(\u03b20+ \u03b21\u2022 DV+ \u03b22\u2022 P+ \u03b23\u2022 T)] where: e is the base of the natural logarithm, \u03b20 is an intercept parameter, \u03b21 is a slope parameter, \u03b22 is a slope parameter, \u03b23 is a slope parameter DV is a factor variable describing monthly mean daily streamflow (ft3/s), P is a factor variable describing monthly mean precipitation (in/day), T is a factor variable describing monthly mean maximum daily air temperature (degrees F). \nParameters describing models containing 4 explanatory variables may be used to populate drought probability equations as follows: p =1/[1 + e^-(\u03b20+ \u03b21\u2022 DV+ \u03b22\u2022 P+ \u03b23\u2022 T+ \u03b24\u2022 PET)] where: e is the base of the natural logarithm, \u03b20 is an intercept parameter, \u03b21 is a slope parameter, \u03b22 is a slope parameter, \u03b23 is a slope parameter, \u03b24  is a slope parameter, DV is a factor variable describing monthly mean daily streamflow (ft3/s), P is a factor variable describing monthly mean precipitation (in/day), T is a factor variable describing monthly mean maximum daily air temperature (degrees F), PET is a factor variable describing monthly mean potential evapotranspiration (in/day). \nDV data span the period of record at each gage, ranging from July 1, 1899 through July 31, 2018. P, T, and PET data span the period associated with each gage beginning July 1, 1981 and ending July 31, 2018. \nEquation goodness of fit parameters document model strength, identifying the utility of each relation. Receiver Operating Characteristic (ROC) AUC values, scaled from 0 to 1, identify each model\u2019s overall correct classification rate and are listed in the table labeled: \u201cDRB-AUC_TABLE.\u201d \nMLLR modeling of drought streamflow probabilities exploits the explanatory power of temporally linked water flows. Models with strong correct classification rates are provided for streams throughout the Delaware River Basin. Hydrological drought MLLR probability estimates inform understanding of drought streamflow conditions, provide warning of future drought conditions, and aid water management decision making. \nMore details of methods used may be found in: Austin, S.H., and Nelms, D.L., 2017, Modeling summer month hydrological drought probabilities in the United States using antecedent flow conditions: Journal of the American Water Resources Association, v. 53, p. 1133\u20131146, accessed November, 15, 2018, at https://doi.org/10.1111/1752- 1688.12562.", "distribution": [{"@type": "dcat:Distribution", "accessURL": "https://doi.org/10.5066/P9NL5PWZ", "description": "Landing page for access to the data", "format": "XML", "mediaType": "application/http", "title": "Digital Data"}, {"@type": "dcat:Distribution", "description": "The metadata original format", "downloadURL": "https://data.usgs.gov/datacatalog/metadata/USGS.60eeffe3d34e93b366704ec5.xml", "format": "XML", "mediaType": "text/xml", "title": "Original Metadata"}], "identifier": "http://datainventory.doi.gov/id/dataset/USGS_60eeffe3d34e93b366704ec5", "keyword": ["Connecticut", "Delaware", "District of Columbia", "New Jersey", "New York", "Pennsylvania", "Rhode Island", "USGS:60eeffe3d34e93b366704ec5", "Virginia", "decision support system", "drought", "forecasting", "risk assessment", "streamflow", "surface water hydrology", "water supply", "watershed management"], "modified": "2021-09-28T00:00:00Z", "publisher": {"@type": "org:Organization", "name": "U.S. Geological Survey"}, "spatial": "-75.934753418555, 38.595065237161, -74.242858887373, 43.000309866704", "theme": ["geospatial"], "title": "Terms, Statistics, and Performance Measures for Maximum Likelihood Logistic Regression Models Estimating Hydrological Drought Probabilities in the Delaware River Basin (2020)"}