ELECTRONIC SUPPLEMENTAL MATERIAL
Complex tidal marsh dynamics structure fish foraging patterns in the San Francisco Estuary
Denise D. Colombano*1,2,3, Thomas B. Handley4, Teejay A. O’Rear2,3, John R. Durand2,3, Peter B. Moyle2,3
1Department of Environmental Science, Policy, and Management University of California, Berkeley
130 Mulford Hall #3114 Berkeley, CA, 94720, USA
2Center for Watershed Sciences University of California, Davis One Shields Avenue
Davis, California, 95616, USA
3Department of Wildlife, Fish, and Conservation Biology University of California, Davis
One Shields Avenue
Davis, California, 95616, USA
4North Central Region Office
California Department of Water Resources 3500 Industrial Boulevard
West Sacramento, CA, 95691, USA
*Corresponding Author: denise.colombano@berkeley.edu
Prepared on November 16, 2020
Methods
Bathymetry modeling with soap-film smoothers
Soap-film smoothers were implemented using generalized additive mixed models (GAMMs), which are flexible hierarchical non-linear regression functions that allow smooth relationships between the predictor and response variables to vary among groups (Wood 2017; Pedersen et al. 2019). The boundary smoothing term was comprised of a cyclic spline with a fixed depth of zero along the shoreline. The major advantage of soap-film smoothing is that the spline averages the depth data from a local region without leaking information across land-based
boundaries, thus ideal for waterways with complex shoreline features
(Simpson 2016). A 1-m resolution grid of x,y coordinates within the channel polygon (not touching the boundary itself) was used as the location of knots supplied to the smoother, which was then coupled with known depth values as model inputs (Fig. S1). Fitted depths from the resulting model were then used to generate a raster of model predictions for the remaining x,y
coordinates within the channel boundaries (Fig. 2). Using Program R v3.6.1 (Team 2019), we conducted spatial analyses with the packages “sf” (Pebesma 2018) and “raster” (Hijmans and Van Etten 2019) and fit models using the “gam” function and the restricted maximum likelihood (REML) smoothing selection method in “mgcv,” which reduces overfitting and yields less variable estimates of smoothing parameters (Wood 2017).
Supplemental Tables
Table S1. Sampling effort for 139 gill-net surveys conducted over nine sampling days. Tide:
incoming vs. outgoing. Microhabitat type: subtidal-open-water confluence (SOC), subtidal-open- water (STO), intertidal-subtidal confluence (ISC), or marsh surface edge (MSE) (Kneib 2000).
Inconsistencies between tide predictions and those retroactively identified by TidalTrend (Donovan and Ayers 2019) led to some inconsistencies in sampling effort (e.g. 2015-08-12).
Date Tide SOC
#Nets STO
#Nets ISC
#Nets MSE
#Nets 2015-04-
29 Incoming 1 -- 1 2
2015-04-
29 Outgoing 1 4 1 4
2015-05-
05 Incoming 1 -- -- 2
2015-05-
05 Outgoing 1 2 2 8
2015-05-
12 Incoming 1 3 1 5
2015-05-
12 Outgoing 1 2 1 2
2015-05-
17 Incoming 1 2 -- 1
2015-05-
17 Outgoing 2 3 2 6
2015-05-
28 Incoming 1 4 1 2
2015-05-
28 Outgoing 1 1 1 6
2015-07-
08 Incoming 1 2 1 5
2015-07-
08 Outgoing 1 2 1 2
2015-07-
15 Incoming 1 -- 1 2
2015-07-
15 Outgoing 1 2 1 5
2015-07-
24 Incoming 1 2 1 2
2015-07-
24 Outgoing 1 2 1 4
2015-08- Incoming 1 -- -- --
12
2015-08-
12 Outgoing 1 4 2 9
Supplemental Figures
Fig. S1. Single-beam sonar bathymetry transects with measured channel depths (color gradient:
yellow=shallower, purple=deeper), and location of knots (black circles) specified in spatial GAMMs. Data: Handley 2015, UCD, unpublished data.
Fig. S2. Gut fullness model fits show effects of channel depth (elevation) by tide direction.
Fig. S3. (A) Fish-count model fits show effects of channel depth (elevation) by species.
Fig. S3. (B) Fish-count model fits show effects of tide height by species.
References
Donovan, J. M., and D. E. Ayers. 2019. TidalTrend software technical report. US Geological Survey, Dept. of Interior.
Hijmans, R. J., and J. Van Etten. 2019. raster: Geographic data analysis and modeling. (version R Program 3.0-2). Vienna, Austria: The R Foundation.
Kneib, R. T. 2000. Salt Marsh Ecoscapes and Production Transfers by Estuarine Nekton in the Southeastern United States. In Concepts and Controversies in Tidal Marsh Ecology, ed.
M. P. Weinstein and D. A. Kreeger, 267–291. Dordrecht: Springer Netherlands.
doi:10.1007/0-306-47534-0_13.
Pebesma, E. 2018. Simple Features for R: Standardized Support for Spatial Vector Data. The R Journal 10: 439. doi:10.32614/RJ-2018-009.
Pedersen, E. J., D. L. Miller, G. L. Simpson, and N. Ross. 2019. Hierarchical generalized additive models in ecology: an introduction with mgcv. PeerJ 7. PeerJ Inc.: e6876.
doi:10.7717/peerj.6876.
Simpson, G. 2016. Soap-film smoothers & lake bathymetries. From the Bottom of the Heap.
Team, R. C. 2019. R Project for Statistical Computing (2019). Vienna, Austria.
Wood, S. N. 2017. Generalized Additive Models: An Introduction with R, Second Edition. CRC Press.