Predicting the effects of sea level rise and salinity changes on west coast tidal marsh plant and avian communities
|
|  Yüklə 7.88 Mb.
|
Modeling Species DistributionsSpecies distribution models (SDM), also known as niche models or bioclimatic models, have seen increasing popularity in recent years as tools for predicting potential shifts in species’ distributions as a result of climate change (Pearson and Dawson 2003; Thuiller 2004; Araujo et al. 2005). This empirical approach has distinct practical advantages in that it tends to provide more realistic (data-driven) predictions than theoretical/analytical models, has greater precision than mechanistic or process-based models, and can also provide a high level of generality given proper inputs and informed ecological assumptions (Guisan and Zimmerman 2000). However, most SDM work has been done at a broad, continental or regional scale, often at spatial resolutions of grid cells 1 km2 or greater. Furthermore, the great majority of such modeling work has been done for upland terrestrial habitats. Very little species distribution modeling work has been conducted explicitly for coastal systems, which necessitate a relatively fine-scale approach, due to their limited narrow extent. Although several researchers have conducted spatial evaluations of SLR on the availability and quality of shorebird habitat (Galbraith et al. 2002; Austin and Rehfisch 2003), we know of only one example of an SDM used to predict climate change-induced shifts in coastal or estuarine species (Rehfisch et al. 2004). There are several common approaches to SDMs, which can be categorized as simple bioclimatic envelope models such as BIOCLIM (Busby 1991) and DOMAIN (Carpenter et al. 1993); statistical models such as generalized linear models (GLM; McCullagh and Neder 1989), generalized additive models (GAM; Hastie and Tibshirani 1990), and classification and regression trees (CART; Breiman et al. 1984); or machine learning approaches, such as genetic algorithms for rule-set prediction (GARP; Peterson 2001), artificial neural networks (ANN; Ripley 1996), and maximum entropy (MaxEnt; Phillips et al. 2006). In general, statistical approaches are considered the most rigorous and are usually used with species occurrence datasets that contain both presence and absence data, while envelope models and some machine learning approaches are most suitable for presence-only occurrence data, such as museum specimens or natural heritage databases. However, there is wide variation in the performance of these models, and this depends on a large number of factors that are difficult to control. Recent comparative studies have suggested that novel methods such as MaxEnt (Elith et al. 2006) and model-averaged CARTs (Lawler et al. 2006) have the highest rates of prediction success in some contexts. However, standard GLMs and GAMs are widely used, have strong statistical foundations, identify functional relationships, are relatively easy to interpret, and perform well in comparison tests (Wintle et al. 2005). We have chosen to use a combination of MaxEnt, GLM, and GAM modeling methods. OBJECTIVES AND HYPOTHESESUsing a combination of field sampling and data analysis, species distribution modeling, and experimental manipulations, we propose to address the following overall question: How will tidal marsh extent and community processes respond to a range of future SLR and salinity scenarios? In our study we will focus on the following specific questions:
We will address this question through (a) intensive vegetation sampling at six mature marsh sites across the salinity gradient, and analysis of within-marsh patterns of distribution, diversity, and productivity; (b) extensive sampling of vegetation distributions across the Bay-Delta, and spatial modeling of estuary-wide patterns of distribution, diversity, and productivity; (c) experimental evaluation of the differential impact of salinity and inundation on six dominant tidal marsh plant species to determine threshold sensitivities to these factors; and (d) observational study of upstream seed dispersal to evaluate species-specific dispersal limitations and order of colonization.
We will address this question by using an extensive dataset of avian occurrence and abundance across the Bay-Delta, in conjunction with plant distribution, diversity, and productivity data, as well spatial environmental data layers, to develop and compare various empirical models of avian distribution and abundance.
We will address this question by applying future predictions of SLR and salinity shifts to spatial models developed using current plant distribution data. A range of climate change scenarios will be used to assess conservative to more extreme predictions of future conditions. Model-predicted future distributions will be compared with current distributions, and with results of greenhouse salinity and inundation experiments, to describe an envelope of future change potential.
We will address this question by applying future predictions of SLR and salinity shifts, as well as predicted plant species distributions, primary productivity, and plant species diversity, to models of avian distribution and abundance, predicting future suitable habitat areas for bird species of interest. |