LAI-DA for eCLM-PDAF

[jjokella]

First implementation of LAI-DA by @lstrebel.

Testing

• by @zhao7777
• goal: use LAI-DA in ESMDA in https://github.com/HPSCTerrSys/TSMP_DA
• OL parameter perturbations
  • soil hydraulic parameters perturbation + forcing perturbation (influence small)
    • spread too small: simulated LAI ensemble does not cover MODIS LAI observations (500m resolution)
  • currently: Try out perturbing Medlyn-parameters + Kmax + leaf carbon nitrogen in addition to SHP/forcing perturbation

Summary of PR: LAI Data Assimilation

This branch implements assimilation of Leaf Area Index (LAI) observations into eCLM (with BGC active).

Background

In eCLM with BGC, LAI is a diagnostic quantity derived from leaf carbon (leafc) and plant-functional-type constants for specific leaf area (slatop, dsladlai) via the Thornton & Zimmermann (2007) formula. Because LAI is not a direct model state variable, assimilation must operate on leafc (and the consistent leafn) rather than on LAI directly.

New enkfpf.par parameters

• Parameter: CLM:update_lai
• Section: [CLM]
• Default: 0
• Description: Selects the LAI DA mode (0 = off, 1 = gridcell state vector, 2 = patch-level state vector)

---

• Parameter: CLM:update_lai_params
• Section: [CLM]
• Default: 0
• Description: Enables joint state–parameter estimation of slatop/dsladlai (0 = off, 1 = with mode 1, 2 = with mode 2)

---

Mode 1 (update_lai=1): the state vector holds one gridcell-averaged LAI per grid cell. A set function aggregates per-patch LAI from leafc; after the PDAF update an inverse transformation distributes the updated LAI back to per-patch leafc/leafn.

Mode 2 (update_lai=2): the state vector holds leafc, slatop, and dsladlai for every patch (size: 3 × #patches). The mapping to observed gridcell LAI is performed inside obs_op_pdaf, using a nonlinear ensemble observation operator.

Documentation

• docs/users_guide/running_tsmp_pdaf/input_enkfpf.md
• docs/users_guide/running_tsmp_pdaf/lai_da.md (new)