> ## Documentation Index
> Fetch the complete documentation index at: https://pyfia.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# How FIA estimation works

> The sampling design, the EVALID system, and the expansion factors behind every pyFIA estimate.

pyFIA implements design-based estimation following [Bechtold & Patterson (2005)](https://doi.org/10.2737/SRS-GTR-80), the foundational methodology for the FIA program. This page explains the concepts that make the estimates valid — and the one rule you must follow to keep them valid.

## The sampling design

FIA uses a three-phase, nationally consistent design:

1. **Phase 1** — remote sensing stratifies the landscape and identifies forested land.
2. **Phase 2** — field crews measure one ground plot per \~6,000 acres of forest: species, size, condition, and site attributes.
3. **Phase 3** — a subset of Phase 2 plots gets expanded measurements (crown condition, down woody material, soils).

Each ground plot is a cluster of four 24-ft-radius **subplots**, each with a nested 6.8-ft **microplot** for small trees and an optional 58.9-ft **macroplot** for large trees. Trees are assigned to a plot component by diameter, which determines their per-acre expansion.

## The EVALID system

This is the most important concept for getting correct estimates.

FIA data is organized into **evaluations**, each identified by a 6-digit **EVALID** (`SSYYTT`: state, year, type). An evaluation is a complete, statistically valid set of plots and strata for one purpose — estimating area, volume, growth, and so on.

A single downloaded state contains **many overlapping evaluations** (different years and types). They share plots. If you estimate across more than one at a time, plots are counted repeatedly and your totals inflate — often by 10-60x.

<Warning>
  **Always clip to one evaluation before estimating.** In pyFIA:

  ```python theme={null}
  db.clip_most_recent()                  # most recent VOL evaluation
  db.clip_most_recent(eval_type="GRM")   # for growth / mortality / removals
  db.clip_by_evalid(442301)              # or pick an EVALID explicitly
  ```
</Warning>

### Evaluation types

| Type                              | Used for                             |
| --------------------------------- | ------------------------------------ |
| `EXPALL`                          | Area estimation (most comprehensive) |
| `EXPVOL`                          | Volume and biomass                   |
| `EXPGROW` / `EXPMORT` / `EXPREMV` | Growth / mortality / removals        |
| `EXPCHNG`                         | Change estimation                    |

The evaluation **year** is a reference year for a complete estimate — not the year each plot was measured. EVALID `442301` represents Rhode Island's 2023 conditions even though it includes plots measured across several prior years (`INVYR`).

## Expansion factors

Estimates scale plot measurements up to the full population through a hierarchy of factors:

* **`TPA_UNADJ`** — base trees-per-acre a tree represents, by plot component.
* **`ADJ_FACTOR_*`** — per-stratum adjustments for non-sampled area, selected by tree diameter (micro / subplot / macro).
* **`EXPNS`** — the stratum expansion factor (acres represented per plot), from `POP_STRATUM`.

The per-acre contribution of a tree is, conceptually:

```
value × TPA_UNADJ × ADJ_FACTOR × EXPNS × CONDPROP_UNADJ
```

pyFIA applies all of these automatically — you never compute them by hand.

## Land and tree classifications

**Forest land** (`COND_STATUS_CD = 1`): at least 10% stocked, ≥1 acre, ≥120 ft wide.

**Timberland**: forest land that is unreserved and productive (capable of ≥20 ft³/acre/year).

**Growing stock**: live trees ≥5" DBH meeting merchantability standards (`TREECLCD = 2`).

These map directly to the `land_type` and `tree_type` parameters — see [Domain filtering](/guides/filtering).

## Volume types

| Variable   | Description                                                          |
| ---------- | -------------------------------------------------------------------- |
| `VOLCFNET` | Net cubic-foot volume (defects deducted) — the growing-stock default |
| `VOLCFGRS` | Gross cubic-foot volume (no deductions)                              |
| `VOLCFSND` | Sound cubic-foot volume (excludes rotten/missing cull)               |
| `VOLBFNET` | Net board-foot sawlog volume                                         |

## Growth, removals, and mortality (GRM)

GRM estimates come from remeasurement, stored in dedicated tables:

* **`TREE_GRM_COMPONENT`** — each tree's fate component (SURVIVOR, MORTALITY, CUT, INGROWTH).
* **`TREE_GRM_MIDPT`** — measurements at the remeasurement midpoint, for annualization.

Because GRM uses different evaluations, remember `eval_type="GRM"` when clipping.

## Common grouping codes

<AccordionGroup>
  <Accordion title="Ownership (OWNGRPCD)">
    10 = National Forest · 20 = Other Federal · 30 = State/Local · 40 = Private
  </Accordion>

  <Accordion title="Site productivity (SITECLCD)">
    1 = 225+ · 2 = 165-224 · 3 = 120-164 · 4 = 85-119 · 5 = 50-84 · 6 = 20-49 · 7 = 0-19 (cu ft/ac/yr)
  </Accordion>

  <Accordion title="Tree status (STATUSCD)">
    1 = Live · 2 = Dead · 3 = Removed (cut)
  </Accordion>
</AccordionGroup>

## Further reading

* [Understanding variance](/concepts/variance) — how pyFIA computes standard errors
* **Bechtold, W\.A. & Patterson, P.L. (2005).** *The Enhanced FIA Program — National Sampling Design and Estimation Procedures.* Gen. Tech. Rep. SRS-80. [DOI: 10.2737/SRS-GTR-80](https://doi.org/10.2737/SRS-GTR-80)