Module 3: Estimating Biomass From Tree Ring Data

Author

Andria Dawson and Kelly Heilman

Published

2025-08-14

Biomass estimation from tree ring and forest census data

How do we get estimates of aboveground biomass and forest carbon from tree diameters and tree ring data?

Allometric equations define the relationship between commonly measured aspects of a tree (usually tree species, diameter, and/or height) and the weight of the tree (aboveground biomass, AGB). These equations are developed through a small set of destructive samples– this involves sampling tree size, height, and species, measuring tree mass, and developing a quantitative relationship between the commonly measured metric and tree mass.

In this module we will:

  1. Apply species-specific allometric equations to the Harvard Forest Census dataset to estimate AGB at the time of census

  2. Back-calculate tree diameter from the tree ring data at Harvard Forest

  3. Apply allometric equations to the back-calculated tree diameters at Harvard Forest to estimate a timeseries of AGB.

1. Estimating Forest biomass for trees at the time of census

We will use the allometric equations in the allodb R package

This package is available on github, and we highlight some additional resources here:

Code 
devtools::install_github("ropensci/allodb",upgrade = "never")
library(allodb)
library(tidyverse)
library(dplR)
library(gt)
library(reshape2)

HARV <- readRDS("data/tree_data_HARVARD_STAN_v3.1_102020.RDS")


taxon.conversion <- readRDS("data/taxon_conversion.RDS")

Tree.ring <- HARV$Xobs %>% left_join(., HARV$Tr)

# get the calendar years
year.df <- data.frame(year = max(Tree.ring$year):1, 
                      cal.year = 2012:(2013 - max(Tree.ring$year)))

Tree.ring <- Tree.ring %>% left_join(., year.df)

year.df <- data.frame(year = max(Tree.ring$year):1, 
                      cal.year = 2012:(2013 - max(Tree.ring$year)))
# set up the lat-long information:
HARV.ll <- data.frame(site = c("LF1", "LF2", "LF3"), 
                      lat = c(42.53065, 42.53128, 42.53008), 
                      lon = c(-72.18346, -72.18271, -72.18246))

Census <- HARV$Dobs %>% left_join(., year.df) %>% 
  rename("taxon" = "species")%>%
  left_join(., taxon.conversion) %>%# combine with taxon-infomation
  left_join(., HARV.ll) # combine with lat-long information

Census <- Census %>% 
  mutate(INV.YEAR = ifelse(cal.year >= 1985 & cal.year <= 1991, 1990, cal.year))

Here, we will focus on species from Harvard Forest, so lets see which species we have in our census data. This information is stored in the taxon column in the Census object:

Code 
# lets visualize the number of trees sampled at Harvard Forest each census year:
Census %>% group_by(taxon, `Scientific Name`, COMMON_NAME, INV.YEAR) %>%
  summarise(n.trees = n()) %>% 
  group_by(taxon, `Scientific Name`, COMMON_NAME) %>% 
  spread(INV.YEAR, n.trees, fill = 0) %>%
  ungroup() %>% gt()
taxon Scientific Name COMMON_NAME 1960 1967 1973 1990 1999 2009
ACRU Acer rubrum red maple 41 211 188 141 111 90
BEAL Betula alleghaniensis yellow birch 8 17 18 16 24 29
BELE Betula lenta sweet birch 2 5 5 7 10 16
FAGR Fagus grandifolia American beech 2 4 4 12 16 26
HAVI Hamamelis virginiana American witchhazel 2 5 5 9 10 25
PIST Pinus strobus eastern white pine 4 29 24 6 8 8
QURU Quercus rubra northern red oak 27 129 114 79 73 67
QUVE Quercus velutina black oak 1 7 6 4 3 3
TSCA Tsuga canadensis eastern hemlock 3 4 7 8 13 25
Code 
# we can also plot this up visually:

ggplot(data = Census)+
geom_bar(aes(x = as.character(INV.YEAR), fill = `Scientific Name`), position = "dodge")+theme_bw()+
ylab("Number of Trees Observed")+
xlab("Inventory Year")+facet_wrap(~site, ncol = 1)

Based on this table and figure, which species do you think will have the largest aboveground biomass?

Lets use the get_biomass function from allodb to get estimates of biomass

This function requires the following arguments and provides and output of Aboveground biomass in kilograms (kg):

  • dbh: diameter at breast height (cm), numeric

  • genus: genus of the tree, character

  • species: the species name of the tree, character, can be NULL if only genus is known

  • coords: a vector (longitude, latitude) of the plot location, or a matrix with 2 columns if trees have individual locations

Code 
require(allodb)

# lets get the biomass for the first tree in the Census data (a 36.83 cm Northern Red Oak (Quercus rubra):

Census.1 <- Census[1,] # first tree

Census.1$AGB.1 <-
  get_biomass(
    dbh = Census.1$dbh ,
    genus = Census.1$GENUS,
    species = Census.1$SPECIES,
    coords = c(Census.1$lon, Census.1$lat)
  )

Census.1$AGB
[1] 979.5547

Now lets apply this function to the whole dataset

Code 
ll <- matrix(c(Census$lon, Census$lat), nrow = length(Census$lon), ncol = 2, byrow = FALSE, dimnames = NULL)

# here we split up the census by sites 
Census.LF1 <- Census %>% filter(site %in% "LF1")
LF1.ll <- HARV.ll %>% filter(site %in% "LF1")

Census.LF2 <- Census %>% filter(site %in% "LF2")
LF2.ll <- HARV.ll %>% filter(site %in% "LF2")

Census.LF3 <- Census %>% filter(site %in% "LF3")
LF3.ll <- HARV.ll %>% filter(site %in% "LF3")

# get the biomass for each site:
Census.LF1$AGB_kg <-
  get_biomass(
    dbh = Census.LF1$dbh ,
    genus = Census.LF1$GENUS,
    species = Census.LF1$SPECIES,
    coords = c(LF1.ll$lon, LF1.ll$lat))


Census.LF2$AGB_kg <-
  get_biomass(
    dbh = Census.LF2$dbh ,
    genus = Census.LF2$GENUS,
    species = Census.LF2$SPECIES,
    coords = c(LF2.ll$lon, LF2.ll$lat))

Census.LF3$AGB_kg <-
  get_biomass(
    dbh = Census.LF3$dbh ,
    genus = Census.LF3$GENUS,
    species = Census.LF3$SPECIES,
    coords = c(LF3.ll$lon, LF3.ll$lat))

plot_radius = 20


Census.AGB <- rbind(Census.LF1, Census.LF2, Census.LF3) %>% 
  mutate(tph_corr = (1/(pi*plot_radius^2)) * (1/0.0001) * (1/1000)) #%>%  # get tree per hectare correction factor to scale tree biomass up
  #mutate(AGB_Mg_ha = AGB_kg*) # convert the tree level AGB to MG/ha)



ggplot(data = Census.AGB)+
  geom_point(aes(x = INV.YEAR, y = AGB_kg, group = ID, color = `Scientific Name`))+
  geom_line(aes(x = INV.YEAR, y = AGB_kg, group = ID, color = `Scientific Name`))+
  facet_wrap(~site)+theme_bw()+ylab("Tree Aboveground Biomass (kg)")+xlab("Year")

Lets get the total species trajectories of biomass over the census periods:

Which species are increasing the most over time? Is this what you expected based on the number of trees tallied? Here, we have added lines to visualize the trajectory of taxon AGB over time, but we don’t have estimates of variability in annual AGB increment over time.

Code 
# ggplot(data = Census.AGB)+
#   geom_point(aes(x = INV.YEAR, y =AGB_Mg_ha, group = ID, color = `Scientific Name`))+
#   geom_line(aes(x = INV.YEAR, y = AGB_Mg_ha, group = ID, color = `Scientific Name`))+
#   facet_wrap(~site)+theme_bw()+ylab("Tree Aboveground Biomass (Mg/ha)")+xlab("Year")

Census.AGB.sums <- Census.AGB %>% group_by(taxon, `Scientific Name`, site, INV.YEAR) %>% 
  summarise(AGB_Mg_ha_taxon = sum(AGB_kg*tph_corr, na.rm =TRUE))

ggplot(data = Census.AGB.sums)+
  geom_point(aes(x = INV.YEAR, y = AGB_Mg_ha_taxon, color = `Scientific Name`))+
  geom_line(aes(x = INV.YEAR, y = AGB_Mg_ha_taxon, color = `Scientific Name`))+
  facet_wrap(~site)+theme_bw()+ylab("Taxon Aboveground Biomass (Mg/ha)")+xlab("Year")

What affects the trajectory of AGB over time?

Forest demographic processes affecting AGB:

  • Recruitment of new small trees into the plot
  • Mortality of trees (or conversely, survival)
  • Growth of existing trees in the plot

Census data provide an estimate of these processes over the remeasurement period.

Recruitment makes up a small amount of AGB at these plots:

  site     ID taxon variable    value recruit.code   Recruitment
1  LF1 LF1001  TSCA     1960       NA            0 Existing Tree
2  LF1 LF1001  TSCA     1967       NA            0 Existing Tree
3  LF1 LF1001  TSCA     1973       NA            0 Existing Tree
4  LF1 LF1001  TSCA     1990 19.36831            1   New Recruit
5  LF1 LF1001  TSCA     1999 45.02890            0 Existing Tree
6  LF1 LF1001  TSCA     2009 99.23069            0 Existing Tree

There are some mortality events that remove AGB

Tree and taxon average annual growth and average annual aboveground biomass increment (AGBI) appear to vary across remeasurement periods:

Code 
# Growth
Average.Growth.census <- mortality.code.census %>% 
  group_by(site, ID) %>%
  arrange(cal.year) %>% 
  mutate(REMPER = cal.year - lag(cal.year),
         prevDBH = lag(dbh), 
         prevAGB = lag(AGB_kg)) %>% 
  # if the tree died in the remeasurement period, assume it died halfway through 
  mutate(REMPER.mort.corrected = ifelse(mort.code == 1, REMPER/2, REMPER)) %>%
  ungroup() %>%
  mutate(tree.AGBI = (AGB_kg - prevAGB)/REMPER.mort.corrected, 
         tree.growth = (dbh - prevDBH)/REMPER.mort.corrected) 


ggplot(data = Average.Growth.census) + 
  geom_point(aes(x = INV.YEAR, y = tree.AGBI, color = taxon, group = ID))+
  geom_line(aes(x = INV.YEAR, y = tree.AGBI, color = taxon, group = ID))+ facet_wrap(~site)+ylab("Tree Average Aboveground Biomass Increment (kg/year)")

Code 
ggplot(data = Average.Growth.census) + 
  geom_point(aes(x = INV.YEAR, y = tree.growth, color = taxon, group = ID))+
  geom_line(aes(x = INV.YEAR, y = tree.growth, color = taxon, group = ID))+ facet_wrap(~site)+ylab("Tree Average Growth Increment (cm/year)")+xlab("Inventory Year")

Code 
site.avgs <- Average.Growth.census %>% group_by(site, taxon, INV.YEAR)%>%
  summarise(taxon.AGBI.avg = mean(tree.AGBI, na.rm =TRUE), 
            taxon.DI.avg = mean(tree.AGBI, na.rm =TRUE))
ggplot(data = site.avgs) + 
  geom_point(aes(x = INV.YEAR, y = taxon.DI.avg, color = taxon))+
  geom_line(aes(x = INV.YEAR, y = taxon.DI.avg, color = taxon))+ facet_wrap(~site)+ylab("Taxon Average Growth Increment (cm/year)")+xlab("Inventory Year")

Code 
ggplot(data = site.avgs) + 
  geom_point(aes(x = INV.YEAR, y = taxon.AGBI.avg, color = taxon))+
  geom_line(aes(x = INV.YEAR, y = taxon.AGBI.avg, color = taxon))+ facet_wrap(~site)+ylab("Taxon Average AGBI (kgyear)")+xlab("Inventory Year")

However, some remeasurement periods are quite long, and we miss inter-annual variation that could lead to differences in estimates of annual AGBI. Tree-ring data can help us fill in these gaps in our understanding of how AGB varies over time, since they are a direct measurement of annual growth.

Step 2: Use tree ring data to back-calculate tree diameter at Harvard Forest:

In order to use tree ring data to estimate tree AGB, we first need to convert measurements of annual tree growth into estimates of tree diameter.

There are many different ways to convert tree ring radial increments to an estimate of circumference (i.e. diameter) and to an estimate of area (i.e. basal area). The dplR package from opendendro has some functions to calculate basal area increment–check out bai.in, bai.out, and bakker.

For simplicity here, we will we assume that each tree grows symmetrically and in a circle. So we convert tree ring growth data for each core from radial increment to a diameter increment by multiplying by 2. Because we have a diameter measurement from a recent census, we can back-calculate the diameter by subtracting the following year’s increment. To calculate the Diameter for core () in the year just before the census diameter was taken year (), we subtract the diameter increment growth () from the Census measured Diameter 

Back-calculating tree diameter for a single tree Red Maple tree at Harvard Forest:

Read in the rwl files for red maple at Harvard Forest (using read.tucson function in dplR) and link them up to the census dataset.

Code 
library(dplR)
# read in the rwl files using the helpful dplR functions:
all.HF.rwl <- list.files("data/rwl/", pattern = ".rwl", full.names = T)


read.rwl.reformat <- function(SPCD.rwl){
        inc.rwl <- read.tucson(SPCD.rwl)
        # here, I reformat the data as a dataframe, get a tree id (id), and keep a core.id.
        # the core.id is a separate core that denotes what direction on the tree the sample was taken from
        inc.rwl.df <- data.frame(inc.rwl) %>% mutate(cal.year = row.names(inc.rwl)) %>%
          reshape2::melt(., id.vars = "cal.year") %>% 
          rename("core.id" = "variable", 
                 "incr" = "value") %>%
          mutate(id = substr(core.id, 1, 6)) %>% 
          filter(!is.na(incr))%>%
          mutate(cal.year = as.numeric(cal.year))
        
        return(inc.rwl.df)

}

all.HF.increment <- do.call(rbind, lapply(all.HF.rwl, read.rwl.reformat))
# get the diameter measurements for the cored trees:
Cored.Diam <- HARV$Tr %>% left_join(., year.df) %>%
  left_join(., all.HF.increment ) %>% filter(!is.na(core.id)) %>% select(-incr)%>%
  rename("cored.year"="year", 
         "cored.cal.year"= "cal.year")

inc.diam <- all.HF.increment  %>% left_join(., Cored.Diam)

# example of back-calculating diameters for all the red maple cores
core.all <- inc.diam %>% 
  
  # filter so we only have increments from the last census diameter measurement or earlier (for now)
  filter(cal.year <= cored.cal.year) %>% 
  
  # get the cumulative diameter increment, or how much the tree has grown from the outside in since a given year:
  # The increments are in mm, so convert to cm by dividing by 10
  # assume the tree is a perfect circle so multiply by 2 to get the diameter increment

  #filter(id %in% "LF1035") %>% 
  group_by(core.id) %>% 
  arrange(desc(cal.year)) %>% 
  
  # calculate the cumulative sum of diameter increments, from the cored diameter to the earliest:
  mutate(cum.diainc = cumsum((incr/10)*2), 
         dia.inc = (incr/10)*2) %>% 
  # simplify this dataset for use:
  select(cal.year, cored.cal.year, core.id, incr, id, dbh, cum.diainc, dia.inc, plot) %>%
  ungroup()%>%
  
  # grouping by the core id after arranging by calendar year is important!!
  group_by(core.id) %>%
  # if the year is the year the cored tree diameter was measured, use that as the dbh
  mutate(dbh_cm_bc = ifelse(cal.year == cored.cal.year, dbh, 
  # if the year is before the tree was cored, use lag to subtract the cumulative increment
  # lag gives use the previous value (gives the cumulative diameter increment from year == 2012 for cal.year == 2011)
                            ifelse(cal.year < cored.cal.year, dbh - lag(cum.diainc), NA)))
  
  # to convince yourself that this is working right, check that the estimated increments == the tree diameter increment:
  # core.all %>% mutate(check.dia.inc = dbh_cm_bc - lead(dbh_cm_bc)) %>% 
  #   filter(!round(dia.inc, 3) == round(check.dia.inc, 3) )
Code 
east.core <- core.all %>% filter(core.id %in% "LF1005e")
 ggplot()+geom_line(data = east.core, aes(x = cal.year, y = dbh_cm_bc, color = core.id))+
   geom_point(data = east.core %>% filter(cal.year == cored.cal.year), aes(x = cal.year, y = dbh))+ylab("Tree diameter (cm)")+xlab("Year")

Multiple cores: Within-tree variation in increment and diameter observations

Using our simple method get a back-calculated tree diameter estimate based on each core. When using tree-ring data to estimate trends in tree size (or Basal Area, Aboveground Biomass, etc), the variation in tree ring widths from different cores in the same tree and year represents observation uncertainty. Here, a single red maple tree (LF1005) has three cores, with variability in tree ring increment and the back-calculated tree size.

Code 
LF2012 <- core.all %>% filter(id %in% "LF1005")

 DBH.bc.p <- ggplot()+geom_line(data = LF2012, aes(x = cal.year, y = dbh_cm_bc, color = core.id))+
   geom_point(data = LF2012 %>% filter(cal.year == cored.cal.year), aes(x = cal.year, y = dbh))+ylab("Tree diameter (cm)")+xlab("Year")
 
 inc.p <-  ggplot()+geom_line(data = LF2012, aes(x = cal.year, y = dia.inc, color = core.id))+
   ylab("Tree Diameter Increment (cm)")+xlab("Year")
 
 cowplot::plot_grid(inc.p, DBH.bc.p, ncol = 1)

We note that the approach in this module has some issues, including the possibility of estimating negative tree diameters near the beginning of the time series, ignoring bark thickness, and assuming perfectly circular tree trunks.

Later in our tutorial, we will demonstrate how developing a “data model” can allow us to estimate uncertainty due to information from different cores and to include this uncertainty in our estimates of tree diameter and aboveground biomass. For this module, we will average the diameter increments for each core before back-calculating the diameters to demonstrate how to estimate biomass over time.

Step 3: visualizing biomass over time (with no uncertainty)

Code 
tree.averages <- inc.diam %>% 
  group_by(id, cal.year, dbh, cored.cal.year, taxon, plot) %>%
  summarise(incr.avg = mean(incr, na.rm = TRUE)) %>% 
  
  # filter so we only have increments from the last census diameter measurement or earlier (for now)
  filter(cal.year <= cored.cal.year) %>% 
  
  # get the cumulative diameter increment, or how much the tree has grown from the outside in since a given year:
  # The increments are in mm, so convert to cm by dividing by 10
  # assume the tree is a perfect circle so multiply by 2 to get the diameter increment
  group_by(id) %>% 
  arrange(desc(cal.year)) %>% 
  
  # calculate the cumulative sum of diameter increments, from the cored diameter to the earliest:
  mutate(cum.diainc = cumsum((incr.avg/10)*2), 
         dia.inc = (incr.avg/10)*2) %>% 
  # simplify this dataset for use:
  select(cal.year, cored.cal.year, incr.avg, id, dbh, cum.diainc, dia.inc, plot, taxon) %>%
  ungroup()%>%
  
  # grouping by the core id after arranging by calendar year is important!!
  group_by(id) %>%
  # if the year is the year the cored tree diameter was measured, use that as the dbh
  mutate(dbh_cm_bc = ifelse(cal.year == cored.cal.year, dbh, 
  # if the year is before the tree was cored, use lag to subtract the cumulative increment
  # lag gives use the previous value (gives the cumulative diameter increment from year == 2012 for cal.year == 2011)
                            ifelse(cal.year < cored.cal.year, dbh - lag(cum.diainc), NA)))
`summarise()` has grouped output by 'id', 'cal.year', 'dbh', 'cored.cal.year',
'taxon'. You can override using the `.groups` argument.
Code 
 ggplot()+geom_line(data = tree.averages, aes(x = cal.year, y = dbh_cm_bc, group = id, color = taxon))+
   geom_point(data = tree.averages %>% filter(cal.year == cored.cal.year), aes(x = cal.year, y = dbh))+ylab("Tree diameter (cm)")+xlab("Year")

Converting forest biomass to carbon using carbon fractions

We convert back-calculated tree diameters to aboveground biomass, again using the allometric equations in the allodb package.

Code 
tree.averages <- tree.averages %>% left_join(.,taxon.conversion) %>% filter(dbh_cm_bc > 0)
Joining with `by = join_by(taxon)`
Code 
TR.LF1 <- tree.averages %>% filter(plot %in% "1")
LF1.ll <- HARV.ll %>% filter(site %in% "LF1")

TR.LF2 <- tree.averages %>% filter(plot %in% "2")
LF2.ll <- HARV.ll %>% filter(site %in% "LF2")

TR.LF3 <- tree.averages %>% filter(plot %in% "3")
LF3.ll <- HARV.ll %>% filter(site %in% "LF3")

# get the biomass for each site:
TR.LF1$AGB_kg <-
  get_biomass(
    dbh = TR.LF1$dbh_cm_bc ,
    genus = TR.LF1$GENUS,
    species = TR.LF1$SPECIES,
    coords = c(LF1.ll$lon, LF1.ll$lat))


TR.LF2$AGB_kg <-
  get_biomass(
    dbh = TR.LF2$dbh_cm_bc,
    genus = TR.LF2$GENUS,
    species = TR.LF2$SPECIES,
    coords = c(LF2.ll$lon, LF2.ll$lat))

TR.LF3$AGB_kg <-
  get_biomass(
    dbh = TR.LF3$dbh_cm_bc,
    genus = TR.LF3$GENUS,
    species = TR.LF3$SPECIES,
    coords = c(LF3.ll$lon, LF3.ll$lat))

TR.AGB <- rbind(TR.LF1, TR.LF2, TR.LF3)


ggplot()+
  geom_line(data = TR.AGB, aes(x = cal.year, y = AGB_kg, group = id, color = COMMON_NAME))+ylab("Tree Aboveground Biomass (kg)")+xlab("Year")

Scaling Tree Ring-based tree-level biomass using the sample design:

For the tree ring sample, a double-nested plot sampling design was deployed at harvard forest, where the smaller diameter trees (<20cm) were sampled in the inner plot and large trees were sampled in the inner plot (>=20cm). This means that each cored tree represents a different sub-sample of the population. We use unit conversions to convert tree biomass units from kg/plot to Mg/ha

Code 
 # we need to adjust the biomass units from kg/plot to Mg/ha
    inner_factor = (1 / (pi*13^2)) * (1/0.0001) * (1/1000)
    outer_factor = (1 / (pi*20^2)) * (1/0.0001) * (1/1000)
    
    
   TR.AGB <-  TR.AGB %>% mutate(tph_corr = ifelse(dbh >= 20, outer_factor, inner_factor)) #%>% mutate(AGB_Mg_ha= AGB_kg*TPA_factor)
   
   saveRDS(TR.AGB, "data/TR_AGB_ests_allodb.rds")

Let’s calculate the total taxon AGB over time & Aboveground biomass increment

Code 
TAXA.AGB <- TR.AGB %>% group_by(COMMON_NAME, GENUS, SPECIES, cal.year, plot) %>% 
  summarise(AGB_Mg_ha_taxa = sum(AGB_kg*tph_corr)) %>% 
  filter(cal.year <= 2010) %>% 
  ungroup() %>% 
  group_by(COMMON_NAME, GENUS, SPECIES, plot) %>%
  arrange(cal.year) %>% 
  mutate(AGBI.taxa = AGB_Mg_ha_taxa - lag(AGB_Mg_ha_taxa)) 
`summarise()` has grouped output by 'COMMON_NAME', 'GENUS', 'SPECIES',
'cal.year'. You can override using the `.groups` argument.
Code 
ggplot()+
  geom_line(data = TAXA.AGB, aes(x = cal.year, y = AGB_Mg_ha_taxa, color = COMMON_NAME), size = 2)+ylab("Taxa Aboveground Biomass (Mg/ha)")+xlab("Year")+facet_wrap(~plot)
Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
ℹ Please use `linewidth` instead.

Code 
ggplot()+
  geom_line(data = TAXA.AGB, aes(x = cal.year, y = AGBI.taxa, color = COMMON_NAME), size = 2)+ylab("Taxa Aboveground Biomass Increment (Mg/ha)")+xlab("Year")+facet_wrap(~plot)
Warning: Removed 9 rows containing missing values or values outside the scale range
(`geom_line()`).

The simplest way of estimating Carbon from tree-level biomass is to assume that the “carbon fraction,” that is the fraction of the tree’s weight that is made up of Carbon, is 50%, so multiplying your biomass value by 0.5 will give you a mass of Carbon per unit area.

Other carbon fractions are available, for example for live trees vs. dead trees, and for different forest types/species, but for simplicity we will use 0.5 here.

Code 
TAXA.C <- TR.AGB %>% group_by(COMMON_NAME, GENUS, SPECIES, cal.year, plot) %>% 
  summarise(Carbon_Mg_ha_taxa = sum(AGB_kg*tph_corr*0.5)) #%>% 
`summarise()` has grouped output by 'COMMON_NAME', 'GENUS', 'SPECIES',
'cal.year'. You can override using the `.groups` argument.
Code 
  # ungroup()%>%
  # # average across sites
  # summarise(Carbon_Mg_ha_taxa_avg = mean(Carbon_Mg_ha_taxa))%>% filter(cal.year <= 2010)

ggplot()+
  geom_line(data = TAXA.C, aes(x = cal.year, y = Carbon_Mg_ha_taxa, color = COMMON_NAME), line_width = 2)+ylab("Taxa Aboveground Carbon (Mg C/ha)")+xlab("Year")+facet_wrap(~plot)
Warning in geom_line(data = TAXA.C, aes(x = cal.year, y = Carbon_Mg_ha_taxa, :
Ignoring unknown parameters: `line_width`

Comparing Census and tree-ring derived Aboveground biomass and Carbon

Code 
Census.totals <- Census.AGB.sums %>% left_join(.,taxon.conversion) %>% 
  group_by(COMMON_NAME, INV.YEAR) %>% 
  summarise(AGB_Mg_ha_Site = mean(AGB_Mg_ha_taxon))
Joining with `by = join_by(taxon, `Scientific Name`)`
`summarise()` has grouped output by 'COMMON_NAME'. You can override using the
`.groups` argument.
Code 
TR.totals <- TAXA.AGB %>% 
  group_by(COMMON_NAME, cal.year) %>% 
  summarise(AGB_Mg_ha_Site = mean(AGB_Mg_ha_taxa))
`summarise()` has grouped output by 'COMMON_NAME'. You can override using the
`.groups` argument.
Code 
ggplot()+
  geom_line(data = TR.totals , aes(x = cal.year, y = AGB_Mg_ha_Site, color = "Tree Ring"), size = 2)+
  geom_point(data = Census.totals, aes(x = INV.YEAR, y = AGB_Mg_ha_Site, color = "Census"), size= 2) + ylab("Taxa Aboveground Biomass (Mg/ha)")+xlab("Year")+facet_wrap(~COMMON_NAME, scales = "free_y")+
  scale_color_manual(values = c("Tree Ring" = "brown", 
                                "Census" = "forestgreen"), 
                     name = "Data Source")

Issues and strengths of estimating biomass from tree ring data alone:

Strengths:

  • Annual estimates of tree growth allow us to estimate interannual variabliity in AGB and AGBI (which we may be interested in correlating to climate, disturbances, etc)

  • tree ring time-series may cover time periods prior to traditional census or forest inventory data

Issues:

  • Approach in this module ignore uncertainty in our estimates due to information from different cores

  • We assume that the census diameter closest to the time of coring is “truth,” but diameter measurements also have measurement uncertainty

  • Lack of demographic data prior to coring leads contributes to an underestimation of AGB in the past; i.e. the “Fading-record” problem, where we miss AGB from trees that have died and prior to coring (and were not sampled).

  • Often, we may not have a dendroecological sample with a core from all the trees.

In the next module, we will introduce a modelling approach that leverages the strengths of both forest census and tree ring data to address some of the issues.