9 Abstract Effective tidal marsh restoration requires
10 predictive models that can serve as planning and
11 design tools to answer basic questions such as which,
12 if any, plant species will colonize a proposed resto-
13 ration site. To develop such a tool, a predictive model
14 of oligohaline tidal marsh vegetation was developed
15 from reference marshes in the Skagit River Delta
16 (Washington, USA) and applied to a 1.1-ha restoration
17 treatment site. Probability curves for the elevational
18 distributions of common marsh species were gener-
19 ated from RTK-GPS point samples of reference tidal
20 marshes. The probability curves were applied to a
21 LIDAR-derived digital elevation model to generate
22 maps predicting the occurrence probability of each
23 species within treatment and control sites. The treat-
24 ment and control sites, located within a recently
25 restored area that had been diked but never completely
26 drained, were covered by a mono-culture of non-
27 native Typha angustifolia L. (narrow-leaf cattail)
28 growing 40–60 cm lower in elevation than in the
29 reference marsh. The T. angustifolia was mowed
30 repeatedly in the treatment site to allow colonization
31 by predicted native marsh species. Four years after
32 mowing, T. angustifolia was replaced on 60 % of the
33 treatment site by native sedges (Carex lyngbyei,
34 Eleocharis palustris), consistent with the predictive
vegetation model; the control site remained covered 35
by T. angustifolia. The mowing experiment confirmed 36
that pre-emptive competition from T. angustifolia was 37
preventing vegetation recovery in the restoration site 38
following dike removal, and implied that some 39
vegetation species may be refractory to environmental 40
change, such as dike removal or sea-level rise, because 41
of differences in recruitment and adult niches.