Key figure

A small appendix for the key figure of the companion paper.

GPP

Code

g_gpp <- read_tsv("outputs/evaluation_fluxes.tsv") %>% 
  filter(variable == "gpp") %>% 
  left_join(read_tsv("outputs/fluxnet_fluxes.tsv") %>% 
              mutate(observed = value*10^2*365/10^3) %>% 
              select(-value)) %>% 
  left_join(read_tsv("data/fluxes/rtsif.tsv") %>% 
              mutate(satellite = rtsif*15.343*365/10^3) %>% 
              select(-rtsif)) %>%
  gather(type, value, -site, -date, -variable) %>% 
  filter(site == "Paracou") %>% 
  group_by(site, variable, type, date = floor_date(date, "15 days")) %>% 
  summarise_all(mean, na.rm = T) %>%
  mutate(week = week(date)) %>%  
  group_by(site, variable, type, week) %>% 
  summarise(l = quantile(value, 0.025, na.rm = TRUE), 
            m = mean(value, na.rm = TRUE), 
            h = quantile(value, 0.975, na.rm = TRUE)) %>% 
  mutate(title = "Mean seasonal variation of GPP") %>% 
  ggplot(aes(week, m)) +
  geom_rect(aes(xmin = week, xmax = end), fill  = "#fff4e0",
           ymin = -Inf, ymax = Inf, alpha = 0.5,
           data = data.frame(week = week(as_date(c("2000-8-1"))),
                             m = -Inf,
                             end = week(as_date(c("2000-12-1"))),
                             site = c("Paracou"))) +
  geom_ribbon(aes(ymin = l, ymax = h, fill = type), col = NA, alpha = 0.2) +
  geom_smooth(aes(col = type), se = FALSE) +
  theme_bw() +
  scale_fill_discrete("") +
  scale_color_discrete("") +
    scale_x_continuous(breaks = week(as_date(paste("2000-", 1:12, "-01"))),
                     labels = c("J", "F", "M", "A", "M", "J", "J", "A", "S", "O", "N", "D")) +
  ylab(expression("Gross Primary Productivity [GPP, "~kgC~m^{-~2}~yr^{-~1}~"]")) +
  xlab("") +
  scale_color_manual("", values = as.vector(cols[c("obs", "sat", "sim")]), 
                     labels = c("Tower", "Satellite", "TROLL")) +
  scale_fill_manual("", values = as.vector(cols[c("obs", "sat", "sim")]), 
                     labels = c("Tower", "Satellite", "TROLL")) +
  geom_text(aes(col = "simulated", label = lab), show.legend = FALSE,
            data = data.frame(week = 8, 
                              m = 1,
                              lab = c("RMSEP=0.75"))) +
  theme(legend.position = c(0.7, 0.2), 
        legend.background = element_rect(colour = NA, fill = NA)) +
  facet_wrap(~ title)
g_gpp

Mean annual cycle from fortnightly means of growth primary productivity for Paracou and Tapajos. Bands are the intervals of means across years, and rectangles in the background correspond to the site’s climatological dry season.

ET

Code

g_et <- read_tsv("outputs/evaluation_fluxes.tsv") %>% 
  filter(variable == "et") %>% 
  left_join(read_tsv("outputs/fluxnet_fluxes.tsv") %>% 
              mutate(observed = value*20) %>% 
              select(-value)) %>% 
  gather(type, value, -site, -date, -variable) %>% 
  filter(site == "Paracou") %>% 
  group_by(site, variable, type, date = floor_date(date, "15 days")) %>% 
  summarise_all(mean, na.rm = T) %>%
  mutate(week = week(date)) %>%  
  group_by(site, variable, type, week) %>% 
  summarise(l = quantile(value, 0.025, na.rm = TRUE), 
            m = mean(value, na.rm = TRUE), 
            h = quantile(value, 0.975, na.rm = TRUE)) %>% 
  mutate(title = "Mean seasonal variation of ET") %>% 
  ggplot(aes(week, m)) +
  geom_rect(aes(xmin = week, xmax = end), fill  = "#fff4e0",
           ymin = -Inf, ymax = Inf, alpha = 0.5,
           data = data.frame(week = week(as_date(c("2000-8-1"))),
                             m = -Inf,
                             end = week(as_date(c("2000-12-1"))),
                             site = c("Paracou"))) +
  geom_ribbon(aes(ymin = l, ymax = h, fill = type), col = NA, alpha = 0.2) +
  geom_smooth(aes(col = type), se = FALSE) +
  theme_bw() +
  theme(legend.position = "bottom") +
  scale_fill_discrete("") +
  scale_color_discrete("") +
    scale_x_continuous(breaks = week(as_date(paste("2000-", 1:12, "-01"))),
                     labels = c("J", "F", "M", "A", "M", "J", "J", "A", "S", "O", "N", "D")) + 
  ylab(expression("Evapotranspiration [ET, "~mm~day^{-~1}~"]")) + xlab("") +
    scale_color_manual("", values = as.vector(cols[c("obs", "sim")]), 
                     labels = c("Tower", "TROLL")) +
  scale_fill_manual("", values = as.vector(cols[c("obs", "sim")]), 
                     labels = c("Tower", "TROLL")) +
  geom_text(aes(col = "simulated", label = lab), show.legend = FALSE,
            data = data.frame(week = 8, 
                              m = 1,
                              site = c("Paracou"),
                              lab = c("RMSEP=0.60"))) +
  theme(legend.position = c(0.7, 0.2), 
        legend.background = element_rect(colour = NA, fill = NA)) +
  facet_wrap(~ title)
g_et

Mean annual cycle from fortnightly means of evapotranspiration for Paracou and Tapajos. Bands are the intervals of means across years, and rectangles in the background correspond to the site’s climatological dry season.

LAI

Code

g_lai <- read_tsv("outputs/evaluation_fluxes.tsv") %>% 
  filter(variable == "lai") %>% 
  select(-variable) %>% 
  mutate(type = "simulated") %>% 
  rename(value = simulated) %>% 
  bind_rows(read_tsv("outputs/lai.tsv") %>% 
              rename(type = variable)) %>% 
    mutate(type = recode(type, 
                       "LAD" = "TLS (Smith et al., 2019)",
                       "LAI" = "Modis",
                       "PAI" = "UAV",
                       "simulated" = "TROLL")) %>% 
  filter(site == "Paracou") %>% 
  group_by(site, type, date = floor_date(date, "15 days")) %>% 
  summarise_all(mean, na.rm = T) %>%
  mutate(week = week(date)) %>%  
  group_by(site, type, week) %>% 
  summarise(l = quantile(value, 0.025, na.rm = TRUE), 
            m = mean(value, na.rm = TRUE), 
            h = quantile(value, 0.975, na.rm = TRUE)) %>% 
  mutate(title = "Mean seasonal variation of LAI") %>% 
  ggplot(aes(week, m)) +
  geom_rect(aes(xmin = week, xmax = end), fill  = "#fff4e0",
           ymin = -Inf, ymax = Inf, alpha = 0.5,
           data = data.frame(week = week(as_date(c("2000-8-1"))),
                             m = -Inf,
                             end = week(as_date(c("2000-12-1"))),
                             site = c("Paracou"))) +
  geom_ribbon(aes(ymin = l, ymax = h, fill = type), col = NA, alpha = 0.2) +
  geom_smooth(aes(col = type), se = FALSE) +
  theme_bw() +
  scale_fill_manual("", values = c("#ffa006", "#0da1f8", "#3f3f3f")) +
  scale_color_manual("", values = c("#ffa006", "#0da1f8", "#3f3f3f")) +
    scale_x_continuous(breaks = week(as_date(paste("2000-", 1:12, "-01"))),
                     labels = c("J", "F", "M", "A", "M", "J", "J", "A", "S", "O", "N", "D")) +
  ylab(expression("Leaf Area Index [LAI, "~m^2~m^{-~2}~"]")) + xlab("") +
  geom_text(aes(col = "TROLL", label = lab), show.legend = FALSE,
            data = data.frame(week = 8, 
                              m = 5.2,
                              site = c("Paracou"),
                              lab = c("RMSEP=0.11"))) +
  theme(legend.position = c(0.7, 0.2), 
        legend.background = element_rect(colour = NA, fill = NA)) +
  facet_wrap(~ title)
g_lai

Mean annual cycle from fortnightly means of leaf area index for Paracou and Tapajos. Bands are the intervals of means across years, and rectangles in the background correspond to the site’s climatological dry season.

FT

Code

troll_traits <- list(
    read_tsv("outputs/Paracou_species_troll.tsv"),
    read_tsv("outputs/Tapajos_species_troll.tsv")
) %>% bind_rows() %>% 
  unique() %>% 
  rename_all(~ gsub("s_", "", .)) %>% 
  select(-seedmass, -regionalfreq) %>% 
  rename(species = name) %>% 
  mutate(species = gsub("_", " ", species)) %>% 
  gather(trait, trait_value, -species) %>% 
  filter(trait == "tlp")
troll_traitdist <- read_tsv("outputs/evaluation_forest2.tsv") %>% 
  filter(dbh >= 10) %>% 
  mutate(plot = repetition) %>% 
  select(site, plot, species) %>%
  filter(site == "Paracou") %>% 
  left_join(troll_traits)
g_tlp <- read_tsv("outputs/functional_composition.tsv") %>%
  filter(trait == "tlp") %>% 
  mutate(title = "Individual tree trait distribution") %>% 
  ggplot(aes(trait_value, group = plot)) +
  stat_density(aes(col = "Observed"), geom="line",position="identity") +
  stat_density(aes(col = "Simulated"), geom="line",position="identity",
               data = troll_traitdist) +
  theme_bw() +
  xlab(expression("Leaf water potential at turgor loss point"~"["~pi[~" TLP"]~","~MPa~"]")) +
  facet_wrap(~ title) +
  scale_color_discrete("") +
  scale_color_manual("", values = as.vector(cols[c("obs", "sim")]), 
                     labels = c("Field", "TROLL")) +
  theme(legend.position = c(0.55, 0.2), 
        legend.background = element_rect(colour = NA, fill = NA)) 
g_tlp

Functional composition estimated by TROLL at the Paracou and Tapajos site expressed in terms of distribution per trait and site.

Assembly

Code

g <- cowplot::plot_grid(g_gpp, g_et, g_lai, g_tlp)

`geom_smooth()` using method = 'loess' and formula = 'y ~ x'

Warning: Removed 1 row containing non-finite outside the scale range
(`stat_smooth()`).

`geom_smooth()` using method = 'loess' and formula = 'y ~ x'
`geom_smooth()` using method = 'loess' and formula = 'y ~ x'

Code

ggsave("figures/key_figure.png", g, dpi = 300, width = 9, height = 9, bg = "white")