BIOME4

Figure. Example BIOME4-famil| Step | Original BIOME4 | Dominance-mode (Biome.jl) | |–-|–-|–-| | 3. Competition | Rule-based comparisons of woody vs. grass PFTs using thresholds on LAI, NPP, fire/greendays, and hand-crafted switches (e.g., temperate evergreen vs. cool conifer; woody-desert fallback; tundra shrub vs. cold herb). | Continuous scored ranking: climate-space dominance (Gaussian proximity to each PFT's optima for clt/temp/prec) × NPP × (1 / dominance_factor). Highest score wins; subdominants retained for mixed states. |utput. The NetCDF typically contains biome (class id), optpft (dominant PFT id), and npp (per-PFT) on the model grid.

We offer two ways to run the BIOME4 logic:

1) Original BIOME4

This path follows an exact translation of the original BIOME core (Haxeltine & Prentice; Kaplan & Prentice). The original Fortran sources by Jed O. Kaplan are available at: https://github.com/jedokaplan/BIOME4

Inputs (climatologies; aligned grids)

• Temperature (temp): °C, monthly means (stacked in the 3rd dimension)
• Precipitation (prec): mm month⁻¹, monthly sums/means (consistent with your decision tree)
• Sun/Cloud (sun or clt): % (sunshine or cloudiness surrogate)
• Soils (Ksat, whc): saturated conductivity (mm h⁻¹) and water-holding capacity (mm cm⁻¹)
• CO₂: a single ppm value representative of the climatology period
• PFT list: default BIOME4 PFTs, or a customized list

Outputs

• biome (lon,lat): BIOME4 biome id (Kaplan & Prentice mapping)
• optpft (lon,lat): id of dominant PFT
• npp (lon,lat,pft): annual NPP per PFT (same order as pftlist)

Notes

• Ensure all rasters share the same grid/extent and monthly stacking.
• Units: temp = °C; prec = mm month⁻¹; sun/clt = %; Ksat = mm h⁻¹; WHC = mm cm⁻¹.
• CO₂ should match the climatology-period mean.

2) Dominance-mode (modified competition)

This mode keeps the BIOME4 physiology (environmental sieve → LAI & NPP per PFT) but modifies the competition step to rank PFTs by climate-space dominance plus productivity.

Rationale (dominance envelope)

We define a fundamental niche for each PFT ($p$) as a set of admissible intervals $[l_{p,j}, u_{p,j}]$ for key environmental predictors $j$, such as absolute minimum temperature ($t_{min}$), growing degree days ($gdd5, gdd0$), mean warm-month temperature ($t_{wm}$), and maximum snow depth ($maxdepth$).

For a given grid cell with environmental vector $\mathbf{x}$, the distance $d_{p,j}$ of the current value $x_j$ from the PFT's interval is:

\[d_{p,j}(x_j) = \begin{cases} l_{p,j} - x_j & \text{if } x_j < l_{p,j} \\ 0 & \text{if } l_{p,j} \le x_j \le u_{p,j} \\ x_j - u_{p,j} & \text{if } x_j > u_{p,j} \end{cases}\]

This distance is normalized by a predictor-specific tolerance parameter $s_{p,j}$. We compute a total squared standardized distance $D_{p}^2(\mathbf{x})$ across all predictors:

\[D_p^2(\mathbf{x}) = \sum_j \left( \frac{ d_{p,j}(x_j) }{ s_{p,j} } \right)^2\]

A fitness index $S_p(\mathbf{x})$ is then derived using a Gaussian-like kernel:

\[S_p(\mathbf{x}) = \exp \left( -\frac{1}{2} D_p^2(\mathbf{x}) \right)\]

The final competitive ability $C(\mathbf{x})$ integrates this fitness with the PFT's potential Net Primary Production (NPP) and an inverse dominance class weight ($D$):

\[C(\mathbf{x}) = S_p(\mathbf{x}) \cdot \text{NPP} \cdot \frac{1}{D}\]

All PFTs present in the grid cell are ranked by this competitive score $C$, and the highest-scoring PFT is selected as the dominant type (optpft). Sub-dominant woody and grass types are also recorded to capture mixed ecosystem states.

Outputs

Same schema as the original mode (biome, optpft, npp). The biome is still assigned using the BIOME4 PFT→biome mapping after dominance is chosen.

What’s different in competition?

At a high level, both modes:

1. Sieve PFTs by physiological constraints (e.g., tmin, gdd5/gdd0, tcm, twm, snow max depth, site water balance).
2. Compute LAI and NPP for all surviving PFTs.

They diverge at step (3):

This preserves BIOME4’s physiology and the environmental sieve, while replacing the final “if/else” winner logic with a continuous dominance score—making it easier to extend to custom PFT sets without rewriting many rule branches.

Inputs — recap (mechanistic modes)

• Required rasters: temp, prec (monthly stacks); typically also clt/sun, Ksat, whc.
• CO₂: single ppm value representative of your climatology period.
• PFT list: defaults to BIOME4.PFTClassification(); parameters/constraints can be customized.

Grid alignment is critical: all rasters must share the same X/Y grid and time stacking. Missing values are normalized internally to -9999.0.

Output variables (NetCDF)

• Coordinates: lon, lat
• Mechanistic schema
  • biome :: Int16 — BIOME4 biome id
  • optpft :: Int16 — dominant PFT id
  • npp :: Float64 — shape (lon, lat, pft); includes num_pfts + 1 slots (default/mixed)

Keep a sidecar legend for PFT id → name and biome id → label used in your plots.

Minimal troubleshooting

• Flat/missing output: check raster alignment/units; inspect the min/max debug prints in the driver.
• Everything “NA”: verify units and that monthly stacks are on the 3rd dimension.
• Resume issues: the driver skips rows where the primary variable (biome) already holds non-fill values; remove/rename the outfile to force a clean run.

References

• Haxeltine, A., & Prentice, I. C. (1996). BIOME3: An equilibrium terrestrial biosphere model based on ecophysiological constraints, resource availability, and competition among plant functional types Global Biogeochemical Cycles, 10(4), 693–709.
• Kaplan, J. O. (2001). Geophysical Applications of Vegetation Modeling.* (Ph.D. thesis), Lund University, Lund, Sweden. doi:10.5281/zenodo.1492908
• Prentice, I. C., Cramer, W., Harrison, S. P., Leemans, R., Monserud, R. A., & Solomon, A. M. (1992). A Global Biome Model Based on Plant Physiology and Dominance, Soil Properties and Climate. Journal of Biogeography, 19(2), 117–134.
• BIOME4 Fortran sources: https://github.com/jedokaplan/BIOME4