Updates to the Everglades Vulnerability Analysis (EVA) vegetation module
Dates
Publication Date
2023-08-16
Time Period
2023-07-15
Citation
D'Acunto, L.E., Hackett, C.E., Haider, S.M., Benscoter, A.M., and Romañach, S.S., 2023, Updates to the Everglades Vulnerability Analysis (EVA) vegetation module: U.S. Geological Survey data release, https://doi.org/10.5066/P9QKJR3G.
Summary
The Everglades Vulnerability Analysis (EVA) is a series of connected, modular Bayesian networks that predict the response of several Everglades indicators of ecosystem health to changes in hydrology, salinity, and the landscape. This release provides the code to update the vegetation module of EVA, validate the updated module, and provides the process and outputs of a sensitivity analysis of the module. Key updates include expanding the number of vegetation classes predicted from 6 to 11 classes, simplifying the inputs to the module, and increasing the number of vegetation observations used to parameterize the network. The validation of the module includes the process to calculate receiver operating characteristic curves and their [...]
Summary
The Everglades Vulnerability Analysis (EVA) is a series of connected, modular Bayesian networks that predict the response of several Everglades indicators of ecosystem health to changes in hydrology, salinity, and the landscape. This release provides the code to update the vegetation module of EVA, validate the updated module, and provides the process and outputs of a sensitivity analysis of the module. Key updates include expanding the number of vegetation classes predicted from 6 to 11 classes, simplifying the inputs to the module, and increasing the number of vegetation observations used to parameterize the network. The validation of the module includes the process to calculate receiver operating characteristic curves and their associated area under the curve values, multi-class Brier scores, and classification error loss from a 10-fold cross-validation on the network. The sensitivity analyses explore the period of record under scenarios of altered hydrology or salinity and determine the most likely vegetation outcome given the proportion of states within the period of record being explored.
These scripts can be used to generate outputs that predict the probability of vegetation type presence given a set of hydrologic, salinity, and landscape parameters derived from spatially explicit hydrology models and other geospatial data.
