Skip to main content
Log in

Insights into Salt Marsh Plant Community Distributions Through Computer Vision and Structural Equation Modeling

Estuaries and Coasts Aims and scope Submit manuscript

Abstract

Community structure and dynamics are influenced by numerous abiotic and biotic factors requiring large datasets to disentangle, which are often difficult to obtain over the spatiotemporal scales necessary for meaningful analysis. The approach outlined here illustrates one potential solution to this problem by leveraging computer vision methods to gain accurate, in-depth community data from ~ 10,000 photographs of salt marsh plants across an elevation gradient at Sapelo Island, GA, USA. A convolutional neural network (ResNext101) trained to detect the 6 dominant plant species achieved high accuracy for all species, allowing mapping of high-marsh plant communities over gradients in elevation and pore-water salinity. To statistically analyze the high-resolution mapping data, we constructed a structural equations model using the generated data as informed by prevailing ecological theory for salt marshes in the Southeastern United States. Model fit to data was strong, with R2 values for five of six plant species > 0.7. The distribution of the rare understory perennial Limonium carolinianum, however, was not accurately predicted by the model. Modeled effects of abiotic factors elevation and soil salinity were commensurate with the literature. Biotic interactions also largely conformed to ecological understanding of Southeastern marshes, but a potentially novel positive interaction between Borrichia frutescens and Batis maritima was observed. Overall, this approach shows promise as a method of efficiently generating and statistically analyzing community data for sessile species at scales not previously possible. This study contributes to a growing body of work developing integrated computer vision and big data techniques for ecological field work.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6

Data Availability

Data are available from the authors upon request.

References

  • Alber, M., E.M. Swenson, S.C. Adamowicz, and I.A. Mendelssohn. 2008. Salt marsh dieback: An overview of recent events in the US. Estuarine, Coastal and Shelf Science 80: 1–11.

    Article  Google Scholar 

  • Antlfinger, A.E. 1981. The genetic basis of microdifferentiation in natural and experimental populations of Borrichia frutescens in relation to salinity. Evolution 1056–1068.

  • Antlfinger, A.E., and E.L. Dunn. 1983. Water use and salt balance in three salt marsh succulents. American Journal of Botany 70: 561–567.

    Article  Google Scholar 

  • Barré, P., B.C. Stöver, K.F. Müller, and V. Steinhage. 2017. LeafNet: A computer vision system for automatic plant species identification. Ecological Informatics 40: 50–56.

    Article  Google Scholar 

  • Beijbom, O., P.J. Edmunds, C. Roelfsema, J. Smith, D.I. Kline, B.P. Neal, M.J. Dunlap, V. Moriarty, T.-Y. Fan, and C.-J. Tan. 2015. Towards automated annotation of benthic survey images: Variability of human experts and operational modes of automation. PLoS ONE 10: e0130312.

    Article  Google Scholar 

  • Bertness, M.D. 1991. Zonation of Spartina patens and Spartina alterniflora in New England salt marsh. Ecology 72: 138–148.

    Article  Google Scholar 

  • Bertness, M.D., and A.M. Ellison. 1987. Determinants of pattern in a New England salt marsh plant community. Ecological Monographs 57: 129–147.

    Article  Google Scholar 

  • Bertness, M.D., L. Gough, and S.W. Shumway. 1992. Salt tolerances and the distribution of fugitive salt marsh plants. Ecology 73: 1842–1851.

    Article  Google Scholar 

  • Bivand, R.S., and D.W.S. Wong. 2018. Comparing implementations of global and local indicators of spatial association. TEST 27: 716–748.

    Article  Google Scholar 

  • Blasco-Moreno, A., M. Pérez-Casany, P. Puig, M. Morante, and E. Castells. 2019. What does a zero mean? Understanding false, random and structural zeros in ecology. Methods in Ecology and Evolution 10: 949–959.

    Article  Google Scholar 

  • Bollen, K.A., J.B. Kirby, P.J. Curran, P.M. Paxton, and F. Chen. 2007. Latent variable models under misspecification: Two-stage least squares (2SLS) and maximum likelihood (ML) estimators. Sociological Methods & Research 36: 48–86.

    Article  Google Scholar 

  • Bortolus, A., P. Adam, J.B. Adams, M.L. Ainouche, D. Ayres, M.D. Bertness, T.J. Bouma, J.F. Bruno, I. Caçador, and J.T. Carlton. 2019. Supporting Spartina: Interdisciplinary perspective shows Spartina as a distinct solid genus. Ecology 100: e02863.

    Article  Google Scholar 

  • Brodrick, P.G., A.B. Davies, and G.P. Asner. 2019. Uncovering ecological patterns with convolutional neural networks. Trends in Ecology & Evolution 34: 734–745.

    Article  Google Scholar 

  • Cadena, C., L. Carlone, H. Carrillo, Y. Latif, D. Scaramuzza, J. Neira, I. Reid, and J.J. Leonard. 2016. Past, present, and future of simultaneous localization and mapping: Toward the robust-perception age. IEEE Transactions on Robotics 32: 1309–1332.

    Article  Google Scholar 

  • Chapman, V.J., 1977. Wet coastal ecosystems. Elsevier Scientific Pub. Co..

  • Chen, L.-C., Y. Zhu, G. Papandreou, F. Schroff, and H. Adam. 2018. Encoder-decoder with atrous separable convolution for semantic image segmentation. In Proceedings of the European conference on computer vision (ECCV) 801–818.

  • Christin, S., É. Hervet, and N. Lecomte. 2019. Applications for deep learning in ecology. Methods in Ecology and Evolution 10: 1632–1644.

    Article  Google Scholar 

  • Cliff, A., and J. Ord. 1981. Spatial processes: models and applications. Pion Limited.

  • Correll, M.D., W. Hantson, T.P. Hodgman, B.B. Cline, C.S. Elphick, W. Gregory Shriver, E.L. Tymkiw, and B.J. Olsen. 2019. Fine-scale mapping of coastal plant communities in the Northeastern USA. Wetlands 39: 17–28.

    Article  Google Scholar 

  • Crain, C.M., B.R. Silliman, S.L. Bertness, and M.D. Bertness. 2004. Physical and biotic drivers of plant distribution across estuarine salinity gradients. Ecology 85: 2539–2549.

    Article  Google Scholar 

  • Cressie, N. 2015. Statistics for spatial data. John Wiley & Sons.

    Google Scholar 

  • D’Urban Jackson, T., G.J. Williams, G. Walker-Springett, and A.J. Davies. 2020. Three-dimensional digital mapping of ecosystems: A new era in spatial ecology. Proceedings of the Royal Society B: Biological Sciences 287: 20192383.

    Article  Google Scholar 

  • Dale, M.R., and M.-J. Fortin. 2002. Spatial autocorrelation and statistical tests in ecology. Ecoscience 9: 162–167.

    Article  Google Scholar 

  • Dale, V.H., A.E. Lugo, J.A. MacMahon, and S.T. Pickett. 1998. Ecosystem management in the context of large, infrequent disturbances. Ecosystems 1: 546–557.

    Article  Google Scholar 

  • Dethier, M.N., E.S. Graham, S. Cohen, and L.M. Tear. 1993. Visual versus random-point percent cover estimations: 'objective' is not always better. Marine Ecology Progress Series 93–100.

  • Dray, S., R. Pélissier, P. Couteron, M.-J. Fortin, P. Legendre, P.R. Peres-Neto, E. Bellier, R. Bivand, F.G. Blanchet, and M. De Cáceres. 2012. Community ecology in the age of multivariate multiscale spatial analysis. Ecological Monographs 82: 257–275.

    Article  Google Scholar 

  • Ellison, A.M. 1987. Effects of competition, disturbance, and herbivory on Salicornia europaea. Ecology 68: 576–586.

    Article  Google Scholar 

  • Fletcher, D., D. MacKenzie, and E. Villouta. 2005. Modelling skewed data with many zeros: A simple approach combining ordinary and logistic regression. Environmental and Ecological Statistics 12: 45–54.

    Article  Google Scholar 

  • Friedrichs C.T., and J.E. Perry. 2001. Tidal salt marsh morphodynamics: a synthesis. Journal of Coastal Research 7–37.

  • Gaspard, G., D. Kim, and Y. Chun. 2019. Residual spatial autocorrelation in macroecological and biogeographical modeling: A review. Journal of Ecology and Environment 43: 19.

    Article  Google Scholar 

  • Grace, J.B. 2006. Structural equation modeling and natural systems. Cambridge University Press.

    Book  Google Scholar 

  • Grace, J.B., T.M. Anderson, H. Olff, and S.M. Scheiner. 2010. On the specification of structural equation models for ecological systems. Ecological Monographs 80: 67–87.

    Article  Google Scholar 

  • Grace, J.B., and K.A. Bollen. 2005. Interpreting the results from multiple regression and structural equation models. The Bulletin of the Ecological Society of America 86: 283–295.

    Article  Google Scholar 

  • Grace, J.B., D.R. Schoolmaster Jr., G.R. Guntenspergen, A.M. Little, B.R. Mitchell, K.M. Miller, and E.W. Schweiger. 2012. Guidelines for a graph-theoretic implementation of structural equation modeling. Ecosphere 3: 1–44.

    Article  Google Scholar 

  • Hester, M.W., I.A. Mendelssohn, and K.L. McKee. 2001. Species and population variation to salinity stress in Panicum hemitomon, Spartina patens, and Spartina alterniflora: Morphological and physiological constraints. Environmental and Experimental Botany 46: 277–297.

    Article  CAS  Google Scholar 

  • Hladik, C., and M. Alber. 2014. Classification of salt marsh vegetation using edaphic and remote sensing-derived variables. Estuarine, Coastal and Shelf Science 141: 47–57.

    Article  Google Scholar 

  • Hopkinson, B.M. 2022. Marsh_Ann: experiments on saltmarsh data for multi-label and single label classification. GitHub.

  • Hopkinson, B.M., A.C. King, D.P. Owen, M. Johnson-Roberson, M.H. Long, and S.M. Bhandarkar. 2020. Automated classification of three-dimensional reconstructions of coral reefs using convolutional neural networks. PLoS ONE 15: e0230671.

    Article  CAS  Google Scholar 

  • Huang, J., V. Rathod, C. Sun, M. Zhu, A. Korattikara, A. Fathi, I. Fischer, Z. Wojna, Y. Song, and S. Guadarrama. 2017. Speed/accuracy trade-offs for modern convolutional object detectors. In Proceedings of the IEEE Conference On Computer Vision and Pattern Recognition 7310–7311.

  • Lambert, D. 1992. Zero-inflated Poisson regression, with an application to defects in manufacturing. Technometrics 34: 1–14.

    Article  Google Scholar 

  • LeCun, Y., Y. Bengio, and G. Hinton. 2015. Deep learning. Nature 521: 436–444.

    Article  CAS  Google Scholar 

  • Legendre, P., and M.J. Fortin. 1989. Spatial pattern and ecological analysis. Vegetatio 80: 107–138.

    Article  Google Scholar 

  • Legendre, P., and L. Legendre. 1998. Numerical ecology, 2nd. English. Elsevier Science.

    Google Scholar 

  • Li, J., and A.D. Heap. 2014. Spatial interpolation methods applied in the environmental sciences: A review. Environmental Modelling & Software 53: 173–189.

    Article  Google Scholar 

  • MacKenzie, D.I., J.D. Nichols, G.B. Lachman, S. Droege, J. Andrew Royle, and C.A. Langtimm. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83: 2248–2255.

    Article  Google Scholar 

  • Martone, R.G., and K. Wasson. 2008. Impacts and interactions of multiple human perturbations in a California salt marsh. Oecologia 158: 151–163.

    Article  Google Scholar 

  • McIntire, E.J., and A. Fajardo. 2009. Beyond description: The active and effective way to infer processes from spatial patterns. Ecology 90: 46–56.

    Article  Google Scholar 

  • Meese, R.J., and P.A. Tomich. 1992. Dots on the rocks: A comparison of percent cover estimation methods. Journal of Experimental Marine Biology and Ecology 165: 59–73.

    Article  Google Scholar 

  • Mitsch, W.J., and J.G. Gosselink. 2015. Wetlands. John Wiley & Sons.

    Google Scholar 

  • Moor, H., K. Hylander, and J. Norberg. 2015. Predicting climate change effects on wetland ecosystem services using species distribution modeling and plant functional traits. Ambio 44: 113–126.

    Article  Google Scholar 

  • Moran, P.A. 1950. Notes on continuous stochastic phenomena. Biometrika 37: 17–23.

    Article  CAS  Google Scholar 

  • Morris, J.T. 1995. The mass balance of salt and water in intertidal sediments: Results from North Inlet, South Carolina. Estuaries 18: 556–567.

    Article  CAS  Google Scholar 

  • Nakagawa, S., and I.C. Cuthill. 2007. Effect size, confidence interval and statistical significance: A practical guide for biologists. Biological Reviews 82: 591–605.

    Article  Google Scholar 

  • Navarro, J.R. 2020. Analysis of the facilitative interaction between Batis Maritima and Avicennia Germinans as a mangrove restoration strategy. The University of Texas Rio Grande Valley.

  • Odum, W.E. 1988. Comparative ecology of tidal freshwater and salt marshes. Annual Review of Ecology and Systematics 19: 147–176.

    Article  Google Scholar 

  • Parashar, J., S.M. Bhandarkar, J. Simon, B.M. Hopkinson, and S.C. Pennings. 2021. Estimation of abundance and distribution of salt marsh plants from images using deep learning. In 2020 25th International Conference on Pattern Recognition (ICPR) 2635–2642. IEEE.

  • Pearl, J. 2012. The causal foundations of structural equation modeling. California Univeristy Los Angeles Department of Computer Science.

  • Pennings, S.C., and M.D. Bertness. 2001. Salt Marsh Communities. Marine Community Ecology 11: 289–316.

    Google Scholar 

  • Pennings, S.C., M.-B. Grant, and M.D. Bertness. 2005. Plant zonation in low-latitude salt marshes: Disentangling the roles of flooding, salinity and competition. Journal of Ecology 93: 159–167.

    Article  Google Scholar 

  • Pennings, S.C., and D.J. Moore. 2001. Zonation of shrubs in western Atlantic salt marshes. Oecologia 126: 587–594.

    Article  Google Scholar 

  • Pennings, S.C., and C.L. Richards. 1998. Effects of wrack burial in salt-stressed habitats: Batis maritima in a southwest Atlantic salt marsh. Ecography 21: 630–638.

    Article  Google Scholar 

  • Potts, J.M., and J. Elith. 2006. Comparing species abundance models. Ecological Modelling 199: 153–163.

    Article  Google Scholar 

  • Rand, T.A. 2000. Seed dispersal, habitat suitability and the distribution of halophytes across a salt marsh tidal gradient. Journal of Ecology 88: 608–621.

    Article  Google Scholar 

  • Schalles, J.F., C.M. Hladik, A.A. Lynes, and S.C. Pennings. 2013. Landscape estimates of habitat types, plant biomass, and invertebrate densities in a Georgia salt marsh. Oceanography 26: 88–97.

    Article  Google Scholar 

  • Schibalski, A., A. Lehtonen, and B. Schröder. 2014. Climate change shifts environmental space and limits transferability of treeline models. Ecography 37: 321–335.

    Article  Google Scholar 

  • Schonberger, J.L., and J.-M. Frahm. 2016. Structure-from-motion revisited. In Proceedings of the IEEE conference on computer vision and pattern recognition 4104–4113.

  • Schweiger, E.W., J.B. Grace, D. Cooper, B. Bobowski, and M. Britten. 2016. Using structural equation modeling to link human activities to wetland ecological integrity. Ecosphere 7: e01548.

    Article  Google Scholar 

  • Simon, J. 2022. SEM_submodel_scripts: scripts for conducting local structural equations modeling via an approach best suited for large, spatially-realized ecological data sets. GitHub.

  • Spiker, J.S., and T.A. Warner. 2007. Scale and spatial autocorrelation from a remote sensing perspective. Geo-Spatial Technologies in Urban Environments 197–213. Springer.

  • Touchette, B. 2006. Salt tolerance in a Juncus roemerianus brackish marsh: Spatial variations in plant water relations. Journal of Experimental Marine Biology and Ecology 337: 1–12.

    Article  CAS  Google Scholar 

  • Villa, A.G., A. Salazar, and F. Vargas. 2017. Towards automatic wild animal monitoring: Identification of animal species in camera-trap images using very deep convolutional neural networks. Ecological Informatics 41: 24–32.

    Article  Google Scholar 

  • Vince, S.W., and A.A. Snow. 1984. Plant zonation in an Alaskan salt marsh: I. Distribution, abundance and environmental factors. The Journal of Ecology 651–667.

  • Weinstein, B.G., S. Marconi, M. Aubry-Kientz, G. Vincent, H. Senyondo, and E.P. White. 2020. DeepForest: A Python package for RGB deep learning tree crown delineation. Methods in Ecology and Evolution 11: 1743–1751.

    Article  Google Scholar 

  • Welsh, A.H., R.B. Cunningham, C. Donnelly, and D.B. Lindenmayer. 1996. Modelling the abundance of rare species: Statistical models for counts with extra zeros. Ecological Modelling 88: 297–308.

    Article  Google Scholar 

  • Wiegert, R., L. Pomeroy, and W. Wiebe. 1981. Ecology of salt marshes: an introduction. The ecology of a salt marsh (pp. 3–19). New York: Springer.

  • Wiegert, R.G., and B. Freeman. 1990. Tidal salt marshes of the southeast Atlantic coast: a community profile. US Department of the Interior, Fish and Wildlife Service.

  • Xie, S., R. Girshick, P. Dollár, Z. Tu, and K. He. 2017. Aggregated residual transformations for deep neural networks. In Proceedings of the IEEE conference on computer vision and pattern recognition pp. 1492–1500.

Download references

Acknowledgements

This study was supported by the Georgia Coastal Ecosystems Long-Term Ecological Research Project (National Science Foundation Awards OCE-0620959 and OCE-1237140) and National Science Foundation Award DBI 2016741. This is contribution number 1107 from the University of Georgia Marine Institute. We thank Marc Garbey for initiating the project, multiple field assistants for contributing to the sampling and measurement efforts, and James Grace for advice on statistical analyses.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Jacob Simon.

Ethics declarations

Competing Interests

The authors declare no competing interests.

Additional information

Communicated by Dan Friess

Appendices

Appendix

Fig. 7
figure 7

Map of the study site (Dean Creek marsh on Sapelo Island, GA, USA) delineated in 2010. Circles indicate locations of perimeter poles. Transects (rows of photographs) ran diagonally from the lower left to the upper right of the image. The darker areas are vegetated, and the lighter coloration is a salt pan. Image was generated using Google Earth, with satellite photos collected on May of 2014

Fig. 8
figure 8

Field apparatus. The carriage was constructed from two bicycles with the camera suspended between them pointing directly down. The carriage was pulled across the study area using ropes that also served to steer it

Fig. 9
figure 9

Manual annotation. A typical low marsh image in the annotation user interface (UI), divided into 15 cells. Blue squares in the center of some cells indicate the presence of Spartina. Only the Spartina annotation layer is shown, but each image annotation included 7 layers (one for each species plus bare substrate)

Statistical Appendices

Relevant Range Standardization

The relevant range standardization procedure used here was developed and discussed by Grace and Bollen (2005) as an improved method of coefficient standardization in multiple regression and SEM. The goal in utilizing any standardized coefficient lies in ease of interpretation, as unstandardized regression results can consist of many different units, and the magnitudes of the effects can be difficult to interpret relative to one another. The typical coefficient standardization technique puts results in units of standard deviations (SDs), which can produce issues as well, namely, the assumption that an SD in one variable is equal to an SD of another and the inherent relation of SD and therefore standardized coefficient magnitude to the sample variances which are likely to be different across variables. Since our variables are non-normally distributed, we opted for the alternative range standardization method. Here, “relevant range” refers to the range of practical values for a variable (e.g., plant abundance cannot be negative) and is set to log(0 + 1) through log10(15 + 1) for our image data variables to represent the log transformed, constrained values of 0–15 that can be reached. This produces coefficients interpretable based on percentages of each variables’ relevant range, e.g., as salinity increases across its relative range (25.2 to 137.5 psu), Spartina experiences a −97.2% change across its range (log(1) through log(16)).

Spatial Autocorrelation

It is important to acknowledge the role of spatial autocorrelation in our model results, as this significant statistical issue often goes unmentioned in the published literature (Gaspard et al. 2019). Moran’s I estimates were high for each species, with values ranging from 0.4 to 0.75 (Table 2). Effective sample sizes were therefore diminished; however, models still had high numbers of effective observations relative to most ecological studies. Abundance models often had considerably lower effective sample sizes, an expected outcome since the samples for these models were based on photos in which a given species was observed.

Since the approach used here is merely a first-order correction meant to detect and mitigate the effects of spatial autocorrelation, it did not fully address the issue. Workers in remote sensing have demonstrated that with higher spatial resolution in pixels, spatial autocorrelation strength increases (Spiker and Warner 2007). While the very high (sub-centimeter) resolution of our data likely introduces a good deal of autocorrelation, it is possible to develop models to utilize spatial patterns in residuals to gain further insights (McIntire and Fajardo 2009; Dray et al. 2012). Therefore, an important line of future work in developing this methodology will be identifying effective statistical or methodological methods that more rigorously address the issue of spatial autocorrelation.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Simon, J., Hopkinson, B. & Pennings, S. Insights into Salt Marsh Plant Community Distributions Through Computer Vision and Structural Equation Modeling. Estuaries and Coasts 46, 431–449 (2023). https://doi.org/10.1007/s12237-022-01147-w

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s12237-022-01147-w

Keywords

Navigation