--- title: Do starting points matter? subtitle: Sensitivity test of pre-fire stand composition date: today categories: [] date-modified: last-modified draft: true format: html: embed-resources: true --- To compare my model output to exclosure data, I'd like to find data on the relationship between age and growth in stands that burned at high-severity. Using iLand output like *fraction of crownkill* and *basal area consumed* didn't work well for this question, since they both aggregate to the scale of the resource unit, and I kept ending up with trees that slipped through and weren't killed in the fire, complicating the age-growth relationships. Instead, I've figured out how to take the stand output and look at the years before and after fire to find the resource unit IDs, and then I pull out the stand and sapling data for just those resource units. I'm currently filtering by the following conditions: - experienced 1 fire (and only 1 fire) in the 100 year simulation - contained stands and saplings in the resource unit in the year before the fire (TSF = -1) - had no trees or saplings in the year of the fire (TSF = 0) - had no trees or saplings in the year following the fire (TSF = 1) In practice it looks something like this: ::: column-margin Here's part\* of the code I'm using to filter the output: ::: ```{r} #| eval: false prefireStand <- stand %>% filter(TSF == -1 & # only TSF from year b4 fire count_ha > 0 & NPPabove_kg > 0 & # only rids with biomass b4 fire cohort_count_ha > 0 & cohort_basal_area > 0) %>% distinct(rid, .keep_all = TRUE) # find unique rids postfireStand <- stand %>% filter(TSF == 1 & count_ha == 0 & height_avg_m == 0 & NPPabove_kg == 0) %>% distinct(rid, .keep_all = TRUE) fireStand = stand %>% filter(TSF == 0 & count_ha == 0 & height_avg_m == 0 & NPPabove_kg == 0) %>% distinct(rid, .keep_all = TRUE) RID = postfireStand %>% filter(rid %in% prefireStand$rid, rid %in% fireStand$rid) %>% pull(rid) %>% unique() stand = stand %>% filter(rid %in% RID, cohort_id %in% postfireStand$cohort_id, TSF > 0, count_ha > 0) %>% select(c(year, ru, rid, species, count_ha, height_avg_m, dbh_avg_cm, basal_area_m2, TSF)) ``` ::: column-margin \*Not included is the code that finds stands that burned once and only once, that's a bit lengthy and I'm confident it's working the way I want ::: Following that, I wanted to test out how much the prefire stand composition matters for the age growth relationships of different species. ```{r} #| label: setup #| warning: false #| echo: false library(tidyverse) library(terra) library(cowplot) theme_set(theme_cowplot()) ``` ```{r} #| label: load data #| warning: false #| echo: false data = read.csv("data/ageGrowth_sensitivity.csv") ``` # Sensitivity test I'm using 3 approaches: 1. Not filtering by pre-fire composition at all 2. Selecting just the stands with a spruce importance value of 0.5 or more 3. Selecting just the stands with a spruce importance value of 0.9 or more ::: column-margin I tried a spruce importance value of 1, but in many of the replicates there weren't any stands that met that criteria at all. ::: # How much does prefire stand composition matter? All of the following is from runs of moose, hare, moose + hare and no moose/hare under historic climate, using the POLE-FM parameters and replicated 10x. ```{r} #| label: plot birch #| warning: false #| echo: false data %>% filter(species == "Bene") %>% filter(Scenario != "Hare") %>% filter(Scenario != "Moose") %>% ggplot() + geom_line(aes(x = TSF, y = height, color = Scenario)) + xlim(c(0,50)) + geom_ribbon(aes(x = TSF, ymin = low, ymax = high, fill = Scenario), alpha = 0.3) + labs(x = " ", y = "Av. Height (m)", title ="Birch") + scale_color_manual(values = c("#a6611a", "#018571")) + scale_fill_manual(values = c("#a6611a", "#018571")) + facet_wrap(~filter) ``` Looks like birch doesn't persist in the later decades after fire in the stands that were mature black spruce prior to burning without herbivores. that seems counter-intuitive to me? ```{r} #| label: plot aspen #| warning: false #| echo: false data %>% filter(species == "Potr") %>% filter(Scenario != "Hare") %>% filter(Scenario != "Moose") %>% ggplot() + geom_line(aes(x = TSF, y = height, col = Scenario)) + xlim(c(0,50)) + geom_ribbon(aes(x = TSF, ymin = low, ymax = high, fill = Scenario), alpha = 0.3) + scale_color_manual(values = c("#a6611a", "#018571")) + scale_fill_manual(values = c("#a6611a", "#018571")) + geom_hline(yintercept = 0, linetype = "dashed") + labs(x = " ", y = "Av. Height (m)", title ="Aspen") + facet_wrap(~filter) ``` Aspen ends up much taller following fire in spruce stands, and it's taller in the scenarios with browsing, which again, seems like the opposite of what I'd expect ```{r} #| label: plot spruce #| warning: false #| echo: false data %>% filter(species == "Pima") %>% filter(Scenario != "Hare") %>% filter(Scenario != "Moose") %>% ggplot() + geom_line(aes(x = TSF, y = height, col = Scenario)) + xlim(c(0,50)) + geom_ribbon(aes(x = TSF, ymin = low, ymax = high, fill = Scenario), alpha = 0.3) + scale_color_manual(values = c("#a6611a", "#018571")) + geom_hline(yintercept = 0, linetype = "dashed") + labs(x = "Time since fire (years)", y = "Av. Height (m)", title ="Spruce") + scale_fill_manual(values = c("#a6611a", "#018571")) + facet_wrap(~filter) ``` This looks like there's some spruce that slipped through the filters - I don't fully understand how spruce could have a height of 0 in TSF = 0 and TSF = 1, but appear at 4 meters in TSF = 2. I'll clip for now: ```{r} #| label: plot spruce TSF2 #| warning: false data %>% filter(species == "Pima") %>% filter(Scenario != "Hare") %>% filter(Scenario != "Moose") %>% filter(TSF > 2) %>% ggplot() + geom_line(aes(x = TSF, y = height, col = Scenario)) + xlim(c(0,50)) + geom_ribbon(aes(x = TSF, ymin = low, ymax = high, fill = Scenario), alpha = 0.3) + scale_color_manual(values = c("#a6611a", "#018571")) + geom_hline(yintercept = 0, linetype = "dashed") + labs(x = "Time since fire (years)", y = "Av. Height (m)", title ="Spruce") + scale_fill_manual(values = c("#a6611a", "#018571")) + facet_wrap(~filter) ``` Okay, so pre-fire stand conditions don't matter much for spruce growth - makes sense, since it's set up to grow steady/incrementally. But somehow it's slightly taller in the scenario without browsing, even though neither herbivore actually browse black spruce. This is making me think I need to chat with Juha/Kati. # Does the herbivore matter? Here's the same data but now broken out by herbivore: ```{r} #| label: plot all birch #| warning: false #| echo: false data %>% filter(species == "Bene") %>% ggplot() + geom_line(aes(x = TSF, y = height, color = Scenario)) + xlim(c(0,50)) + geom_ribbon(aes(x = TSF, ymin = low, ymax = high, fill = Scenario), alpha = 0.3) + labs(x = "Time since fire (years)", y = "Av. Height (m)", title ="Birch") + facet_wrap(~filter) ``` ```{r} #| label: plot all aspen #| warning: false #| echo: false data %>% filter(species == "Potr") %>% ggplot() + geom_line(aes(x = TSF, y = height, color = Scenario)) + xlim(c(0,50)) + geom_ribbon(aes(x = TSF, ymin = low, ymax = high, fill = Scenario), alpha = 0.3) + labs(x = "Time since fire (years)", y = "Av. Height (m)", title ="Aspen") + facet_wrap(~filter) ``` ```{r} #| label: plot all spruce #| warning: false #| echo: false data %>% filter(species == "Pima") %>% filter(TSF >2) %>% ggplot() + geom_line(aes(x = TSF, y = height, color = Scenario)) + xlim(c(0,50)) + geom_ribbon(aes(x = TSF, ymin = low, ymax = high, fill = Scenario), alpha = 0.3) + labs(x = "Time since fire (years)", y = "Av. Height (m)", title ="Black Spruce") + facet_wrap(~filter) ``` Overall, similar broad trends. Herbivore guild doesn't differ too much in their impacts - the biggest thing I can pick out is that aspen is taller earlier when browsed by hare and not moose. Makes sense.