NMDS — reference vs revegetation (draft)

Bray–Curtis ordination of Laws Type E and reference parcel composition

Modified

July 25, 2026

Draft / sandbox — what this means

What: This page is an open, versioned analysis artifact—methods and numbers stay visible while tools mature (Snow et al. 2026).

Why: Closed file-share workflows hide how compliance figures were produced. Publishing drafts lets partners and the public inspect assumptions early, without treating prototypes as adopted policy or Technical Group findings.¹

Companion pages

STM (draft) — ABAG → early seral → reference-like scrub
Reference · Revegetation
Methods paper: STM / NMDS for Type E (draft)

Purpose

What: Non-metric multidimensional scaling (NMDS) maps parcel–year samples so distances reflect community dissimilarity (McCune and Grace 2002). Species values are relative cover (row sums = 1); distance is Bray–Curtis; fitting uses vegan::metaMDS (Oksanen et al. 2024)—the same core pipeline as ICWD’s stm helpers.

Why: Compliance uses absolute (capped) cover. Relative composition asks a different question: do revegetation parcels (ABAG / early seral starts) sit apart from reference scrub in species space, and do later years move toward the reference cloud?²

Methods (Laws adaptation)

Reveg: LAW090/094/095 (2022–2025 workbook); LAW118/129 combined for 2025.
Reference: long-format reference CSVs (transect hits as cover).
Aggregate within parcel–year; row-normalize; drop species in <2 samples.
Bray–Curtis → metaMDS (k = 2); stress on the figure.
Color by role (reference vs reveg); shape by parcel family.

Run 0 stress 0.1877248 
Run 1 stress 0.1883394 
Run 2 stress 0.2076997 
Run 3 stress 0.1813941 
... New best solution
... Procrustes: rmse 0.05320044  max resid 0.3045848 
Run 4 stress 0.1877248 
Run 5 stress 0.2086552 
Run 6 stress 0.2098296 
Run 7 stress 0.1915483 
Run 8 stress 0.1915483 
Run 9 stress 0.2093531 
Run 10 stress 0.1832765 
Run 11 stress 0.1883394 
Run 12 stress 0.1813941 
... Procrustes: rmse 1.15249e-05  max resid 5.776634e-05 
... Similar to previous best
Run 13 stress 0.1915483 
Run 14 stress 0.2302263 
Run 15 stress 0.1813941 
... Procrustes: rmse 9.156305e-06  max resid 4.5309e-05 
... Similar to previous best
Run 16 stress 0.1998426 
Run 17 stress 0.2056783 
Run 18 stress 0.1813941 
... Procrustes: rmse 2.990254e-06  max resid 1.1739e-05 
... Similar to previous best
Run 19 stress 0.2032681 
Run 20 stress 0.1877248 
*** Best solution repeated 3 times

[1] TRUE

Figure 1. NMDS of parcel–year relative composition: reference vs Laws Type E revegetation samples.

Table 1. Parcel–year samples in the NMDS (download: docs/laws_nmds_scores.csv).

Reading the plot for the STM

Pattern	Interpretation
Reference cloud	Composition near ESD potential (Type A scrub phases)
Reveg apart / on the edge	Still in ABAG / early seral space (STM)
Same parcel moves toward the cloud across years	Trajectory hypothesis for assisted recovery³

Limitations

Relative cover omits bare as a taxon; ABAG barrenness is understated.
Reference parcels are not filtered to Gravelly Loam alone; reveg targets are (laws_type_e_esd.csv).
PERMANOVA / envfit live in stm; this page is the Laws-facing first cut.

Code

code/nmds_laws.R · upstream stm/code/nmds_functions.R

References

McCune, Bruce, and James B. Grace. 2002. Analysis of Ecological Communities. Gleneden Beach, Oregon: MjM Software Design.

Oksanen, Jari, Gavin L. Simpson, F. Guillaume Blanchet, Roeland Kindt, Pierre Legendre, Peter R. Minchin, R. B. O’Hara, et al. 2024. vegan: Community Ecology Package. https://CRAN.R-project.org/package=vegan.

Snow, Tasha, Christopher Holdgraf, Wilson Sauthoff, Jessica Scheick, Ellianna Abrahams, Joanna Millstein, Sanjay Bhangar, et al. 2026. “A Path to Better Science Through Co-Creation and Open Infrastructure.” Perspectives of Earth and Space Scientists 7 (1): e2025CN000295. https://doi.org/10.1029/2025CN000295.

Footnotes

Sandbox means placeholders and sparse labels are allowed during prototyping; challenge numbers in the code and sources, not from a screenshot alone.↩︎
Row-normalization removes magnitude so sparse vs dense stands can still be compared by who is present; bare ground is not a species here, so ABAG barrenness is only partly visible.↩︎
LAW118/129 are mostly 2025-only so far; 090-group has 2022–2025.↩︎

--- title: "NMDS — reference vs revegetation (draft)" subtitle: "Bray–Curtis ordination of Laws Type E and reference parcel composition" affiliation: "Inyo County Water Department" date-modified: today format: html: toc: true toc-depth: 3 execute: echo: false warning: false message: false --- {{< include _includes/sandbox_disclaimer.qmd >}} ::: {.callout-note} ## Companion pages - [STM (draft)](stm.html) — ABAG → early seral → reference-like scrub - [Reference](reference.html) · [Revegetation](index.html) - Methods paper: [STM / NMDS for Type E (draft)](paper_stm_nmds.html) ::: ## Purpose **What:** **Non-metric multidimensional scaling (NMDS)** maps parcel–year samples so distances reflect community **dissimilarity** [@McCuneGrace2002]. Species values are **relative cover** (row sums = 1); distance is Bray–Curtis; fitting uses `vegan::metaMDS` [@Oksanen2024]—the same core pipeline as ICWD’s [`stm`](https://github.com/inyo-gov/stm) helpers. **Why:** Compliance uses **absolute** (capped) cover. Relative composition asks a different question: do **revegetation** parcels (ABAG / early seral starts) sit apart from **reference** scrub in species space, and do later years move toward the reference cloud?^[Row-normalization removes magnitude so sparse vs dense stands can still be compared by *who* is present; bare ground is not a species here, so ABAG barrenness is only partly visible.] ## Methods (Laws adaptation) 1. **Reveg:** LAW090/094/095 (2022–2025 workbook); LAW118/129 combined for 2025. 2. **Reference:** long-format reference CSVs (transect hits as cover). 3. Aggregate within parcel–year; row-normalize; drop species in <2 samples. 4. Bray–Curtis → `metaMDS` (*k* = 2); stress on the figure. 5. Color by **role** (reference vs reveg); shape by parcel family. ```{r} #| label: fig-nmds-laws #| fig-cap: "NMDS of parcel–year relative composition: reference vs Laws Type E revegetation samples." #| fig-height: 6 #| fig-width: 8 source("code/nmds_laws.R") long <- load_laws_composition_long(".") built <- build_relative_community_laws(long) comm <- filter_rare_species(built$comm, min_occ = 2) comm <- comm[, colSums(comm) > 0, drop = FALSE] meta <- built$meta %>% dplyr::filter(sample %in% rownames(comm)) %>% dplyr::arrange(match(sample, rownames(comm))) comm <- comm[meta$sample, , drop = FALSE] meta <- meta %>% dplyr::mutate( role = dplyr::coalesce(as.character(role), "reference"), family = dplyr::case_when( parcel %in% c("LAW090", "LAW094", "LAW095") ~ "reveg 090/094/095", parcel == "LAW118/129" ~ "reveg 118/129", TRUE ~ "reference" ), group_label = dplyr::coalesce(as.character(reference_group), family) ) nmds <- run_nmds(comm, trymax = 40) sc <- scores_with_meta(nmds, meta) readr::write_csv(sc, "data/processed/laws_nmds_scores.csv") file.copy("data/processed/laws_nmds_scores.csv", "docs/laws_nmds_scores.csv", overwrite = TRUE) stress <- round(nmds$stress, 3) ggplot(sc, aes(x = NMDS1, y = NMDS2, color = role, shape = family)) + geom_point(size = 3, alpha = 0.85) + ggrepel::geom_text_repel( aes(label = paste0(parcel, "\n", year)), size = 2.4, max.overlaps = 40, show.legend = FALSE ) + scale_color_manual(values = c(reference = "#1b9e77", reveg = "#d95f02")) + labs( title = "Laws Type E — NMDS (relative composition)", subtitle = paste0("Bray–Curtis · stress = ", stress, " · rare spp dropped (min 2 samples)"), color = "Role", shape = "Family" ) + theme_minimal(base_size = 12) + coord_equal() ``` ```{r} #| label: tbl-nmds-samples #| tbl-cap: "Parcel–year samples in the NMDS (download: docs/laws_nmds_scores.csv)." sc_tbl <- sc %>% dplyr::transmute( Parcel = parcel, Year = year, Role = role, Family = family, `Ref group` = dplyr::coalesce(as.character(reference_group), "—"), NMDS1 = round(NMDS1, 3), NMDS2 = round(NMDS2, 3) ) %>% dplyr::arrange(Role, Parcel, Year) DT::datatable(sc_tbl, rownames = FALSE, options = list(pageLength = 20, scrollX = TRUE)) ``` ## Reading the plot for the STM | Pattern | Interpretation | |---------|----------------| | Reference cloud | Composition near ESD potential (Type A scrub phases) | | Reveg apart / on the edge | Still in **ABAG / early seral** space ([STM](stm.html)) | | Same parcel moves toward the cloud across years | Trajectory hypothesis for assisted recovery^[LAW118/129 are mostly 2025-only so far; 090-group has 2022–2025.] | ## Limitations - Relative cover omits bare as a taxon; ABAG barrenness is understated. - Reference parcels are not filtered to Gravelly Loam alone; reveg targets are (`laws_type_e_esd.csv`). - PERMANOVA / envfit live in `stm`; this page is the Laws-facing first cut. ## Code `code/nmds_laws.R` · upstream `stm/code/nmds_functions.R`