Inundation observations and inundation model predictions for vernal pools of the northeastern United States

This data release includes data-processing scripts, data products, and associated metadata for a study to model the hydrology of several hundred vernal pools (i.e., seasonal pools or ephemeral wetlands) across the northeastern United States. More information on this study is available from the project website. This data release consists of several components: (1) an input dataset and associated metadata document ("pool_inundation_observations_and_climate_and_landscape_data"); (2) an annotated R script which processes the input dataset, performs inundation modeling, and generates model predictions ("annotated_R_script_for_pool_inundation_modeling.R"); and (3) a model prediction dataset and associated metadata document ("pool_inundation_predictions") for the model predictions generated by the annotated R script. The inundation modeling procedures are explained in detail in the metadata documents and annotated R script and are described briefly here. The input dataset includes approximately 3,000 field observations of inundation from approximately 450 vernal pools located across the northeastern United States. Observations of inundated depth, length, and width for pools were collected between May and July, from 2004 through 2016. From these observations, four binary inundation metrics were constructed. The H1 metric classified pools as inundated if any amount of water was observed, i.e. inundated depth and area greater than 0. The H2, H3, and H4 metrics classified pools as inundated if they had inundated depth greater than or equal to 5cm and area greater than or equal to 5m2, depth greater than or equal to 10cm and area greater than or equal to 15m2, and depth greater than or equal to 15cm and area greater than or equal to 25m2, respectively. For each inundation metric (H1 through H4), boosted regression tree models were constructed using the inundation metric as the response variable and pool attributes, weather and climate variables, and landscape characteristics as explanatory variables. These models were then used to generate inundation predictions for the H1 through H4 metrics at several seasonal time points and under a variety of weather and climate scenarios.