A Simple Geographical Model of Salt Marsh Vegetation with Respect to Tidal Channel Networks

Eric W. Sanderson1, Theodore C. Foin2 & Susan L. Ustin1
1Department of Land, Air and Water Resources, University of California, Davis, CA 95616, USA;
2Division of Environmental Studies, University of California, Davis, CA 95616, USA

Abstract

Previous work has shown that the distribution of plant species in a salt marsh near Petaluma, CA, is strongly influenced by the location and size of tidal channels. We developed a simple mathematical model to measure the "channel influence" at each p oint as a cumulative function based on inverse squared distance to channel, length of potentially influential channel, and channel size. Using this function and probabilities estimated from the vegetation data, we simulated the two-dimensional pattern of species abundance across the entire site. Comparisons of predictions with field data of overall species cover and distribution showed similar patterns of cover for most species. The results indicate that the distribution of vegetation in the salt marsh , both major zones of marsh dominants and patches of minor species, can be described by a single, empirical factor: cumulative distance from tidal channels of different size.

Tables:

Table 1. Model Weighting Scheme Based on Channel Size.
Assignment of channel order weighting factors is based on the approximate order of magnitude of channel cross-sectional area and maximum discharge measurements.
Channel Order
Assigned Weight
Mean Channel Cross-Sectional Area (m2)
Estimated Maximum Discharge Rate (m3s-1)*
1
0.1
0.06
0-0.03
2
0.5
0.14
0-0.03
3
1
1.12
0.04
4
10
2.35
0.29-0.82
5
100
33
4.5-17.8
6
1000
385
93.5-186.9
Measured by Sanderson et al (1998b)

*Estimated Maximum Discharges based on data from Leopold et al. (1993)

 
Table 2. Overall Species Cover Predicted by Model is Similar to Vegetation Map.
Comparison of the overall amount of cover recorded in the vegetation map of Sanderson et al (1998a) and the simulated model distributions for those species based on the coincident area. The rank order of overall amounts of cover recorded in Sanderson et al (1998a) was used to rank order the predictions of species in the model. See text for details.
Species Total Cover from 

Vegetation Map*

 Total Cover from Simulations 

 
Salicornia virginica
82.95%
77.24%
Spartina foliosa
1.45%
0.69%
Scirpus robustus
11.07%
9.38%
Grindelia stricta
0.67%
0.39%
Lepidium latifolium
0.43%
0.16%
Jaumea carnosa
0.20%
2.34%
Rumex crispus
0.03%
0.02%
Achellia millefolium
0.02%
0.02%
Cuscuta salina
<0.01%
0.03%
Distichlis spicata
not observed
0.20%
Atriplex patula
not observed
<0.01%
*from vegetation map created by Sanderson et al. (1998a).

 
Table 3. Reference Channel Influence (CISD) Values for a Straight Isolated Channel.
Reference Cumulative Inverse Squared Distance (CISD) values for sampling points placed 1, 10 and 20 meters from an isolated and straight channel of given order.
 
Channel Order
Reference CISD Values from Isolated Straight Channel
1 m
10 m
20 m
1
16
10
8
2
18
12
10
3
21
13
11
4
24
16
14
5
27
19
17
6
30
22
20

 
Table 4. Accuracy Assessments of Simulation Model Results.
Accuracy assessments derived from pixel by pixel comparisons of vegetation map of Sanderson et al (1998a) and the simulated model distributions for those species based on the coincident area. Both the map and the simulations were resampled to coincident 10 m grid cells and aggregated into zero, low (0-25%), and high (25-100%) classes. Error matricies were calculated, but not shown, in order to calculate accuracy assessment measures.
Species*
Percentage Cells Correctly Predicted
Average Error of Omission
Average Error of Commission
Salicornia virginica
92.61%
64.08%
61.97%
Spartina foliosa
94.83%
61.37%
51.49%
Scirpus robustus
72.66%
60.74%
61.69%
Frankenia salina
77.34%
67.10%
63.52%
Grindelia stricta
88.67%
33.14%
65.10%
Lepidium latifolium
94.83%
64.10%
31.40%
Jaumea carnosa
79.06%
34.75%
66.76%
Rumex crispus
98.77%
33.33%
0.41%
Achellia millefolium
99.26%
33.33%
0.25%
Cuscuta salina
99.51%
0.16%
33.33%
*no comparison data available for Distchlis spicata and Atriplex patula.

 
Table 5. Similar Vegetation Clusters are Derived from Field and Model Results.
Comparison of the composition of species assemblages calculated from simulated transects in the model distributions to species assemblages observed by Sanderson et al (1998a).

Field Data from Sanderson et al (1997) (n = 900)
 
Cluster A Cluster B Cluster C Cluster D Cluster H Cluster I Cluster J
n = 39
n = 138
n = 21
n = 11
n = 55
n = 616
n = 15
Spartina 65.6% Scirpus 78.0% Grindelia 79.1% Grindelia 69.0% Frankenia 66.3% Salicornia 84.8% Jaumea 74.3%
Scirpus 2.7% Salicornia 56.0% Salicornia 29.8% Frankenia 65.9% Salicornia 52.2% Jaumea 
5.2%		 Distichlis 32.6%
  Lepidium 1.2% Jaumea 
5.0%		 Salicornia 52.2% Distichlis 1.7% Lepidium 1.1% Salicornia 25.2%
    Lepidium 3.2% Jaumea 7.6%     Frankenia 1.0%
    Frankenia 1.9%        
Simulated Data (n = 2187)
 
Cluster a Cluster a’ Cluster a/b Cluster b Cluster c
n = 47 n = 42 n=46 n = 89 n = 36
Spartina 70.7% Spartina 73.3% Salicornia 69.8% Scirpus 73.4% Scirpus 76.4%
  Scirpus 44.8% Spartina 54.4% Spartina 16.8% Grindelia 49.1%
    Scirpus 21.6% Salicornia 4.8% Salicornia 23.3%
         
         
 
Cluster c’ Cluster b/h Cluster h Cluster i Cluster j
n = 14 n = 12 n = 140 n = 1258 n = 36
Salicornia 77.6% Scirpus 75.5% Salicornia 79.3% Salicornia 81.2% Salicornia 71.8%
Grindelia 47.6% Frankenia 46.8% Frankenia 55.4% Jaumea 3.1% Jaumea 56.4%
Jaumea 9.5% Salicornia 20.2% Jaumea 1.5 %   Frankenia 1.2%
Distichlis 3.7%        

 

Figures:

Publications and Presentations:

Abstract for 1997 Ecolocial Society of America Annual Meeting Presentation
 
Return to Eric's Homepage
Eric Sanderson's Current Research/ UC Davis / ewsanderson@ucdavis.edu 1/21/98

1998, Center for Spatial Technologies and Remote Sensing (CSTARS)
University of California, Davis