AllomVignette.Rmd

---
title: "PEcAn.allometry Vignette"
author: "Mike Dietze"
date: "April 23, 2015"
output: html_vignette
vignette: >
  %\VignetteIndexEntry{PEcAn.allometry Vignette}
  %\VignetteEngine{knitr::rmarkdown}
---

## install package from Github

Only needs to be done the first time
```{r, eval=FALSE}
library(devtools)
install_github("PecanProject/pecan",subdir="modules/allometry")
```


## Define species groups

```{r}
library(PEcAn.allometry)

## define the Plant Functional Types you want to fit based on USDA acronyms and USFS species codes (spcd)
## can involve one or many species
## multiple PFTs can be fit at once by putting them in a list
## Note that the name used in the PFT list defines the name that the code will match against when making predictions for new individuals
pfts = list(FAGR = data.frame(spcd=531,acronym="FAGR"))

## example of a PFT with multiple species (LH = late hardwood)
## note that if you're just using Jenkins the acronym column is optional
pfts = list(LH = data.frame(spcd = c(531,318),acronym=c("FAGR","ACSA3")))

```

## Run the Bayesian allometry model

Note the side effects of this function are that it will create two files in your working directory for every PFT and component pair, a pdf file of diagnostics and a RData file saving the full output. These files will be named based on the PFT and the component.

The return from this function will be a summary table of statistics. The function will actually run two variants of the model, a "global" modelthat fits a single equation to all equations and a 'hierarchical' model that accounts for the variability among equations. This function also print out DIC statistics on which fit was better (lowest score wins): DIC is the hierarchical model, DICg is the global model.

```{r}
allom.stats = AllomAve(pfts,ngibbs=500)
```


If you want to run with a response variable other than the default (e.g. components = 6; stem biomass), look up the relevant component IDs in data(allom.components). The default component is 3 (total aboveground biomass). Note that if you specify multiple PFTs (as a list) and multiple components (as a vector) then AllomAve will generate allometries for all PFT x component combinations

```{r}
allom.stats = AllomAve(pfts,ngibbs=500,components=c(3,6))
```

## Predict for individual trees

```{r}
allom.fit = load.allom(getwd())
dbh = 1:50
pred = allom.predict(allom.fit,dbh = dbh,pft = "LH",component = 3,use = "Bg",interval = "prediction")
conf = allom.predict(allom.fit,dbh = dbh,pft = "LH",component = 3,use = "Bg",interval = "confidence")
PI = apply(pred,2,quantile,c(0.025,0.5,0.975),na.rm=TRUE)
CI = apply(conf,2,quantile,c(0.025,0.5,0.975),na.rm=TRUE)
plot(dbh,CI[2,],type='l',lwd=3,ylim=range(PI),ylab="Biomass (kg)")
lines(dbh,CI[1,],lty=2,col="blue")
lines(dbh,CI[3,],lty=2,col="blue")
lines(dbh,PI[1,],lty=3,col="red")
lines(dbh,PI[3,],lty=3,col="red")
```

## Predict for a stand
```{r}
## simulated DBH's
dbh = rpois(100,20)
hist(dbh)

stand = allom.predict(allom.fit,dbh = dbh,pft = "LH",component = 3,use = "Bg",interval = "prediction")
AGB = apply(stand,1,sum)
hist(AGB)
```