---
title: "Analysing green manure decomposition, nutrient release and crop response"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Analysing green manure decomposition, nutrient release and crop response}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r setup, include = FALSE}
knitr::opts_chunk$set(collapse = TRUE, comment = "#>", fig.width = 6.5, fig.height = 4)
has_gg <- requireNamespace("ggplot2", quietly = TRUE)
library(GreenManureR)
```

## Overview

**GreenManureR** supports experiments in which a green manure or cover crop
(Mucuna, sunhemp, dhaincha and others) is incorporated, its residues decompose
in litterbags, nutrients are released, and one or more succeeding crops
(for example kharif and rabi) respond. The package links these three stages:

1. `decomp_fit()` and `decomp_t50()`: residue decomposition;
2. `nutrient_release()`: N, P, K and other nutrient release;
3. `greenmanure_n_balance()`: N inputs, crop uptake and apparent balance;
4. `carryover_effect()`: residual effects on succeeding crops;
5. `gm_joint_analysis()` and `greenmanure_report()`: integration and reporting.

The examples use simulated data (`gm_litterbag`, `gm_agronomy`) that mimic the
structure of a real experiment.

```{r data}
head(gm_litterbag, 3)
head(gm_agronomy, 3)
```

## Decomposition models

With $M$ the percentage of initial mass remaining, three models are fitted per
group and ranked by AICc:

* single pool: $M = 100\,e^{-kt}$;
* double pool: $M = 100\,[A e^{-k_f t} + (1-A) e^{-k_s t}]$ with $k_s \le k_f$;
* asymptotic: $M = C + (100 - C) e^{-kt}$, where $C$ is the recalcitrant residue.

```{r decomp}
fit <- decomp_fit(gm_litterbag, time = "time", mass = "mass", group = "treatment")
fit$table[fit$table$selected, c("treatment", "model", "k", "k_slow", "pool_A", "R2", "AICc")]
```

The time to 50% mass loss uses the selected model, with a Monte-Carlo interval
based on the covariance matrix of the nonlinear least squares fit:

```{r t50}
decomp_t50(fit)
```

```{r decomp-plot, eval = has_gg}
plot(fit)
```

## Nutrient release

Nutrient remaining is `mass x concentration / initial concentration`, so values
above 100% indicate net immobilization. Supplying the dry matter input
(`biomass`, t/ha) also yields absolute release in kg/ha.

```{r nutrient}
nr <- nutrient_release(gm_litterbag, nutrients = c("N", "P", "K"),
                       group = "treatment", biomass = "biomass_tha",
                       at = c(14, 30, 60))
nr$t50[, c("treatment", "nutrient", "model", "t50")]
subset(nr$release_at, nutrient == "N" & time == 30)
```

```{r nutrient-plot, eval = has_gg}
plot(nr)
```

## Nitrogen balance

```{r nbal}
nb <- greenmanure_n_balance(gm_agronomy, by = "season")
s <- nb$summary
s[s$fert_N == 0 & s$treatment != "Control",
  c("season", "treatment", "N_gm_input_mean", "N_uptake_mean", "ANR_mean", "NFRV_mean")]
```

`ANR` is the apparent N recovery from the green manure relative to the control
at the same fertilizer N level. `NFRV` is the fertilizer N replacement value,
obtained by inverting a quadratic response fitted to the control plots; it is
`NA` when the yield lies outside the fitted range.

## Residual effects on succeeding crops

```{r carry}
co <- carryover_effect(gm_agronomy, response = "grain_yield", strata = "fert_N",
                       season_levels = c("Kharif", "Rabi"))
co$effects[co$effects$fert_N == 0,
           c("season", "treatment", "diff", "pct_change", "p_adj", "retention_vs_first")]
co$pooled
```

The `treatment x season` term in the pooled analysis tests whether the green
manure effect changes between crops; `retention_vs_first` gives the second-crop
effect as a percentage of the first-crop effect.

```{r carry-plot, eval = has_gg}
plot(co)
```

## Joint analysis

```{r joint}
ja <- suppressWarnings(
  gm_joint_analysis(fit, nr, nb, co, nutrient_name = "N", at_day = 60, fert_N_level = 0))
ja$data
```

With only three species the rank correlations in `ja$cor` are descriptive; to
draw inference, pool several sites or years.

## Reporting

`greenmanure_report()` writes CSV tables, PNG figures and a Word report to a
directory of your choice:

```{r report, eval = FALSE}
greenmanure_report(fit, nr, nb, co, ja, outdir = "gm_report")
```