The Applicability of the National Wetland Condition Assessment (NWCA) Data for Aboveground Biomass Estimations
Authors: Kendall Harris\(^1\), Gregg Serenbetz\(^2\), Andy Baldwin\(^3\), Annie Rossi\(^2\)
\(^1\) ORISE Research Participant; \(^2\) US Environmental Protection Agency; \(^3\) University of Maryland
library("tidyr")## Warning: package 'tidyr' was built under R version 3.4.4library("plyr")## Warning: package 'plyr' was built under R version 3.4.4library("dplyr")##
## Attaching package: 'dplyr'## The following objects are masked from 'package:plyr':
##
## arrange, count, desc, failwith, id, mutate, rename, summarise,
## summarize## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionlibrary("data.table")## Warning: package 'data.table' was built under R version 3.4.4##
## Attaching package: 'data.table'## The following objects are masked from 'package:dplyr':
##
## between, first, lastlibrary("stringr")## Warning: package 'stringr' was built under R version 3.4.4library("ggplot2")## Warning: package 'ggplot2' was built under R version 3.4.4library("ggiraph")## Warning: package 'ggiraph' was built under R version 3.4.4library("ggpmisc")## Warning: package 'ggpmisc' was built under R version 3.4.4library("ggpubr")## Warning: package 'ggpubr' was built under R version 3.4.4## Loading required package: magrittr## Warning: package 'magrittr' was built under R version 3.4.4##
## Attaching package: 'magrittr'## The following object is masked from 'package:tidyr':
##
## extractAbstract
Plant communities are a significant component of wetland ecosystems and estimations of their aboveground biomass, along with floristic quality measures, can help measure their impact on wetland health, processes and changes. In this study we first sought to evaluate the accuracy of biomass equations for the purposes of estimating aboveground biomass of live trees from the National Wetland Condition Assessment (NWCA), which records estimated size classes of the DBH of individual trees on a plot. By applying regression equations developed by Chojnacky et al. (2014) to data collected by the NWCA we were able to estimate the aboveground biomass of the majority of live trees on NWCA plots across the country using only their DBH. To verify that these biomass estimates were reliable we used data from Harvard Forest LTER plots to compare biomass estimates calculated from actual vs. size-class DBH measurements. We then looked at how aboveground biomass estimations varied between wetland types and regions, and whether vegetation cover could serve as a useful predictor of biomass at NWCA sites. This poster will demonstrate the applicability of NWCA data for the purposes of estimating aboveground biomass, and how those estimations can be applied to future research.
Introduction
The National Wetland Condition Assessment
The National Wetland Condition Assessment (NWCA) is part of the National Aquatic Resources Survey; a continental-scale effort by the EPA to assess the condition of our nations waters. It is designed as a statistical survey that begins to address some of the gaps in our understanding of wetlands and to answer basic questions about the extent to which our nation’s wetlands support healthy ecological conditions and the prevalence of key stressors at the national and regional scale. The 2011 NWCA collected data to characterize biological, chemical, and physical features of each site. Vegetation, soil, hydrology, water chemistry, algae, and buffer characteristics were chosen for evaluation based on their utility in reflecting ecological condition of wetlands or key indicators of stress that may influence condition across broad national and regional scales. Data for each of these indicator groups were obtained from field observations, field samples collected at wetland sites, and laboratory analyses of field samples.
Vegetation Sampling
NWCA collects data on plant species composition and abundance, on vegetation structural attributes, and on ground surface attributes within vegetation plots at each sample site. These vegetation data collected by field crews are later used during analysis to calculate numerous metrics in a variety of categories that inform the development of Vegetation Multimetric Indices that serve as indicators of wetland vegetation condition. Data are collected in five 100-m2 Vegetation Plots placed systematically in the Assessment Area (AA). Plot layouts are designed to provide and unbiased characterization of the vegetation of the AA in a manner that can consistently be applied accross the country.
Question
Our purpose for this study was to see whether the data collected during NWCA can be used to accurately estimate aboveground biomass of the trees present on plots, and if so, whether factors such as wetland type and region, and cover estimates of tree species can be used to predict biomass.
Aboveground Biomass of Tree Species
Calculating
The NWCA does not collect height data of individual trees, and dbh measurements of individual trees are recorded as estimated size classes. There are 7 possible size classes, which increase with tree size: 5-10cm, 11-25cm, 26-50cm, 51-75cm, 76-100cm, 101-200cm and 200+. With these restraints in mind, estimates of the aboveground biomass of trees on NWCA plots were calculated by using regression equations developed by Chojnacky et al. (2014), which use only DBH and the woods’ specific gravity (as opposed to other equations which require tree height). Regression equations were assigned to the species level where possible, and to higher taxonomic levels when a regression equation did not exist for a species. Some groups, such as mangroves, had no regression equations at all and had to be removed from analysis. To deal with the estimated size classes we assigned individual trees the median DBH for the size class they were assigned (a tree with a size class of 26-50cm would be assigned a DBH of 38cm).
Accuracy of results
To assess the validity of using estimated size classes for DBH as opposed to actual measurements we used data from Harvard Forest LTER plots to compare biomass estimates calculated from actual vs. size-class DBH measurements. The Harvard Forest LTER plots have accurate DBH measurements for all individual trees on multiple plots which contain a variety of tree species. We took species and DBH data from several Harvard forest plots and applied the Jenkins regression equations to calculate biomass. We then assigned to each tree the mean DBH from the NWCA DBH size classes that would have been assigned to the individual tree during the NWCA survey, and calculated individual tree aboveground biomass using the Jenkins regression equations and the estimated DBH. By plotting the resulting biomass estimates against eachother, we recieved R\(^2\) values ranging from 0.75-0.93. When the data from all plots was combined, we received an R\(^2\) value of 0.91. This high correlation suggests that using estimated DBH classes is an acceptable way of estimating biomass.
harv <- read.csv("biomassharv.csv")
my.formula <- y ~ x
p1 <- ggplot(data = harv, aes(x = NWCA, y = HARV)) +
geom_smooth(method = "lm", se=FALSE, color="black", formula = my.formula) +
stat_poly_eq(formula = my.formula,
aes(label = paste(..rr.label.., sep = "~~~")),
parse = TRUE) + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
panel.background = element_blank(), axis.line = element_line(colour = "black")) +
geom_point() #+ scale_x_continuous(trans='log2')
p1 + ggtitle("Two methods of estimating biomass applied to Harvard Forest LTER plots") + labs(x="Estimates made using NWCA Size Classes for DBH (kg/tree) ", y = "Estimates made using actual DBH measurements (kg/tree)")
p1
PLOT 1: Aboveground biomass of a Harvard forest plot calculated by using the NWCA size classes for DBH vs the actual DBH measurement from Harvard Forest
Applications to NWCA Data
Biomass across different wetland types/regions
We expected the aboveground biomass of trees to vary significantly between different wetland types and regions, and that the Forested, Shrub Scrub and Wooded wetlands would have higher aboveground biomass per site than herbaceous sites. An ANOVA test run on mean aboveground biomass and wetland class returned a p-value of 9.969e-07. Additionally , two wetland classes varied significantly from the others: PFO (Palustrine Forested and Estuarine Shrub Scrub). These results indicate that biomass is correlated to wetland type, especially sites classified as PFO and E2SS.
biomass <- read.csv("siteMetrics.csv")
biomass <- biomass[, c(2:14)]
siteInfo <- read.csv("nwca2011_siteinfo.csv")
siteInfo <- siteInfo[,c(2,7,19, 20, 35, 52)]
biomass <- merge(biomass, siteInfo, by = "UID")
mass.class = lm(MeanBiomass ~ CLASS_FIELD_FWSST, data = biomass)
summary(mass.class)##
## Call:
## lm(formula = MeanBiomass ~ CLASS_FIELD_FWSST, data = biomass)
##
## Residuals:
## Min 1Q Median 3Q Max
## -3627 -1645 -563 12 60865
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 122.94 825.24 0.149 0.881623
## CLASS_FIELD_FWSSTE2SS 3503.82 1051.22 3.333 0.000912 ***
## CLASS_FIELD_FWSSTPEM 732.73 949.40 0.772 0.440545
## CLASS_FIELD_FWSSTPF -26.91 2509.87 -0.011 0.991449
## CLASS_FIELD_FWSSTPFO 2692.39 868.88 3.099 0.002035 **
## CLASS_FIELD_FWSSTPSS 292.06 947.16 0.308 0.757919
## CLASS_FIELD_FWSSTPUBPAB -33.63 2275.03 -0.015 0.988212
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 4741 on 598 degrees of freedom
## Multiple R-squared: 0.06179, Adjusted R-squared: 0.05237
## F-statistic: 6.563 on 6 and 598 DF, p-value: 9.969e-07#biomassA <- subset(biomass, MeanBiomass > 0)
#df[df$depth<10]<-0
#biomassA$CLASS_FIELD_FWSST[biomassA$CLASS_FIELD_FWSST == "E2EM"] <- "E2EM (n=28)"
biomassA <- read.csv("biomassA.csv")
p2 <- ggboxplot(biomassA, x = "CLASS_FIELD_FWSST", y = "MeanBiomass",
color = "CLASS_FIELD_FWSST",
yscale = "log2",
legend = "none",
add = "mean",
ylab = "Tree Biomass(kg/100m^2)", xlab = "Wetland Class", title = "Aboveground Biomass of Trees By Wetland Class", xtickslab.rt = 45, ytickslab.rt = 45)
p2
PLOT 2: Aboveground tree biomass of NWCA sites (where aboveground biomass of trees > 0), grouped by wetland type. Palustrine forested sites had the highest average aboveground biomass per site. The stars indicate mean values. Wetland classes that had the highest average aboveground biomass were Pulustrine Forested, Palustrine Shrub Scrub and Estuarine Shrub Scrub Sites, while the lowest average aboveground biomass by class were Palustrine Farmed, Palustrine Aquatic Bottom and Estuarine Emergent Sites.
p3 <- ggboxplot(biomassA, x = "ECO_X_WETGRP", y = "MeanBiomass",
color = "ECO_X_WETGRP",
yscale = "log2",
legend="none",
add = "mean",
ylab = "Tree Biomass (kg/100m^2)", xlab = "EPA Aggregated Wetland Group", title = "Aboveground Biomass of Trees By EPA Aggregated Wetland Group", xtickslab.rt = 45, ytickslab.rt = 45) +
font("title", size = 14, color = "black", face = "bold")+
font("subtitle", size = 10, color = "black")+
font("caption", size = 10, color = "black")+
font("xlab", size = 12, color = "black")+
font("ylab", size = 12, color = "black")+
font("xy.text", size = 7, color = "black", face = "bold")
p3
PLOT 3: Aboveground tree biomass of NWCA sites (where aboveground biomass of trees > 0)grouped by wetland type and ecoregion. The stars indicate mean values. The NWCA assigned wetland groups are formatted as ECOREGION-CLASS, with a Herbaceous and Wooded option for each group. (All-EH = All Estuarine Herbaceous, All-EW = All Estuarine Wooded, CPL-PRLH = Coastal Plains Palustrine Riverine Lacustrine Herbaceous, CPL-PRLW = Coastal Plains Palustrine Riverine Lacustrine Wooded, EMU-PRLH = Eastern Mountain Palustrine Riverine Lacustrine Herbaceous, EMU-PRLW = Eastern Mountain Palustrine Riverine Lacustrine Wooded, IPL-PRLH = Interior Plains Palustrine Riverine Lacustrine Herbaceous , IPL-PRLW = Interior Plains Palustrine Riverine Lacustrine Wooded, W-PRLH = Western Palustrine Riverine Lacustrine Herbaceous, W-PRLW = Western Palustrine Riverine Lacustrine Wooded) The average aboveground biomass of trees in the wooded groups was consistantly higher than their herbacaeous counterparts.
Cover as a predictor of biomass
To look at whether tree cover could help predict the aboveground biomass of trees at a site, we plotted the average cover of trees at a site against the average aboveground biomass. To focus our results we also took a closer look at palustrine forested sites, the class of wetland with the highest average aboveground tree biomass during the 2011 NWCA survey.
biomassB <- subset(biomass, MeanBiomass > 0)
biomassB <- subset(biomassB, XABCOV_TREE_COMB > 0)
my.formula <- y ~ x
p4 <- ggplot(data = biomassA, aes(x = XABCOV_TREE_COMB, y = MeanBiomass)) +
geom_smooth(method = "lm", se=FALSE, color="black", formula = my.formula) +
geom_point() + facet_wrap(~CLASS_FIELD_FWSST) + scale_y_continuous(trans='log2')
p4 + theme_bw() + theme(panel.border = element_blank(), panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), axis.line = element_line(colour = "black")) + labs(y="Mean site aboveground biomass (kg/500m^2)", x = "Mean cover of trees per site (%) ")
biomassPFO <- subset(biomassB, CLASS_FIELD_FWSST == "PFO")
my.formula <- y ~ x
p4 <- ggplot(data = biomassPFO, aes(x = XABCOV_TREE_COMB, y = MeanBiomass)) +
geom_smooth(method = "lm", se=FALSE, color="black", formula = my.formula) +
stat_poly_eq(formula = my.formula,
aes(label = paste(..rr.label.., sep = "~~~")),
parse = TRUE) + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
panel.background = element_blank(), axis.line = element_line(colour = "black")) +
geom_point() + facet_wrap(~ECOREGION) + scale_y_continuous(trans='log2')
p4
biomassPFO <- subset(biomassB, CLASS_FIELD_FWSST == "PFO")
my.formula <- y ~ x
p4 <- ggplot(data = biomassPFO, aes(x = XABCOV_TREE_COMB, y = MeanBiomass)) +
geom_smooth(method = "lm", se=FALSE, color="black", formula = my.formula) + geom_point() + facet_wrap(~ECOREGION) + scale_y_continuous(trans='log2')
p4 + theme_bw() + theme(panel.border = element_blank(), panel.grid.major = element_blank(),
panel.grid.minor = element_blank(), axis.line = element_line(colour = "black")) + labs(y="Mean site aboveground biomass (kg/500m^2)", x = "Mean cover of trees per site (%) ")
Discussion
Our results show that the aboveground biomass of trees can be accurately calculated using the estimated DBH size classes provided by the NWCA survey. The R\(^2\) value of 0.91 indicates a high degree of correlation between biomass estimates derived from actual DBH measurements and estimated DBH.
We also confirmed our prediction that aboveground biomass of trees is related to classes and groups of wetlands. The overlapping error bars on these box plots show that aboveground biomass of trees was not the only defining factor when it came to site classification, though some of the extreme points may indicate misclassification (Palustrine Forested Site with biomass = 20, Estuarine Emergent sites with biomass approaching 2000).