You are here:
Development of approaches to predict the distribution of Zostera marina and Z. japonica in Pacific Northwest estuaries
Citation:
LEE, II, H., M. R. FRAZIER, D. REUSSER, C. A. BROWN, P. J. CLINTON, B. DUMBAULD, E. Saarinen, AND D. R. YOUNG. Development of approaches to predict the distribution of Zostera marina and Z. japonica in Pacific Northwest estuaries. Presented at Sea Level Rise Meeting, Newport, OR, February 01 - 02, 2011.
Impact/Purpose:
The dominant species of submerged aquatic vegetation (SAV) in Pacific Northwest (PNW) estuaries is the intertidal and shallow subtidal seagrass, Zostera marina. Beds of Z. marina constitute a critical habitat, including providing habitat for juvenile salmon.
Description:
The dominant species of submerged aquatic vegetation (SAV) in Pacific Northwest (PNW) estuaries is the intertidal and shallow subtidal seagrass, Zostera marina. Beds of Z. marina constitute a critical habitat, including providing habitat for juvenile salmon. Additionally, the nonindigenous Z. japonica has invaded many PNW estuaries with potential positive and negative impacts on ecosystem function and diversity. While it is generally recognized that the distribution and abundance of both species of Zostera will change in response to sea level rise (SLR), the current version of the Sea Level Affects Marsh Model (SLAMM) does not predict changes in SAV. To address this limitation, we have begun evaluating two general approaches to predicting the distribution of both Zostera species in PNW estuaries. The first approach is to generate species distribution models (SDMs; aka: niche models) using point survey data from two studies of intertidal populations of Zostera. The first study was an EPA probabilistic survey of intertidal habitats in 13 estuaries in Oregon and California, where the abundance of Zostera was quantified at approximately 50 to 100 random sites per estuary. The second study was an intensive survey of Willapa Bay, Washington by the USDA where visual estimates of seagrass density and sediment characteristics were made at about 4200 sites. Using each of these studies we will evaluate the predictive power of a suite of SDMs to predict the likelihood of presence based on both site specific environmental parameters (e.g., percent fines of the sediment, presence of burrowing shrimp) and geomorphic/estuarine scale attributes (e.g., distance to mouth, salinity regime). The SDM modeling approaches being compared include: logistic regression, classification and regression trees, nonparametric multiplicative regressions, boosted regression trees, multivariate adaptive regression splines, and an ensemble modeling program, BIOMOD, which combines several of these approaches including neural nets. We will evaluate the models by accuracy, data needs, and whether models developed from the intensive Willapa Bay study are transferable to the EPA regional scale study and vice versa. The second general approach is to predict the probability of occurrence of SAV by simple shoreward shifts of existing SAV distributions due to projected bathymetric change. These GIS models are being calibrated from Z. marina distributions from previous aerial surveys in Oregon estuaries as well as intertidal/subtidal depth distributions of Z. marina in three estuaries. Where not available as input into this GIS model, intertidal bathymetry is interpolated from National Wetland Inventory (NWI) habitat classes. To generate a SAV module for SLAMM, we will evaluate the practicality and generality of the models generated from both the SDM and GIS approaches. It is possible that the most accurate model may utilize site-specific environmental parameters that are not generally available as data layers (e.g., percent fines). In such cases, we will evaluate whether substitution of class values (e.g., mud, sand), for which data layers tend to be more readily available, provide similar predictions or whether a different modeling approach would be required. The ultimate goal of the effort is to develop practical tools to improve the utility of SLAMM to help inform adaptive management responses to SLR and climate change.