add Tech Note section 2.34 on parameter tuning

doc/source/tech_note/Introduction/CLM50_Tech_Note_Introduction.rst

@@ -297,6 +297,8 @@ Biogeophysical and biogeochemical processes are simulated for each subgrid land
 
 #. Carbon isotope fractionation (Chapter :numref:`rst_Carbon Isotopes`)
 
+#. Tuning (Chapter :numref:`rst_Tuning`)
+
 .. _Figure Land processes:
 
 .. figure:: image1.png

doc/source/tech_note/References/CLM50_Tech_Note_References.rst

@@ -239,6 +239,10 @@ Collatz, G.J., Ball, J.T., Grivet, C., and Berry, J.A. 1991. Physiological and e
 
 Collatz, G.J., Ribas-Carbo, M., and Berry, J.A. 1992. Coupled photosynthesis-stomatal conductance model for leaves of C\ :math:`{}_{4}` plants. Aust. J. Plant Physiol. 19:519-538.
 
+.. _Collieretal2018:
+
+Collier, N., Hoffman, F. M., Lawrence, D. M., Keppel-Aleks, G., Koven, C. D., Riley, W. J., Mu, M., Randerson, J. T. 2018. The International Land Model Benchmarking (ILAMB) system: Design, theory, and implementation. Journal of Advances in Modeling Earth Systems, 10, 2731–2754. https://doi.org/10.1029/2018MS001354
+
 .. _Colmer2003:
 
 Colmer, T.D., 2003. Long-distance transport of gases in plants: a perspective on internal aeration and radial oxygen loss from roots. Plant Cell and Environment 26:17-36.
@@ -259,6 +263,10 @@ Cosby, B.J., Hornberger, G.M., Clapp, R.B., and Ginn, T.R. 1984. A statistical e
 
 Crawford, T. W., Rendig, V. V., and Broadent, F. E. 1982. Sources, fluxes, and sinks of nitrogen during early reproductive growth of maize (Zea mays L.). Plant Physiol. 70:1645-1660.
 
+.. _Dagonetal2020:
+
+Dagon, K., Sanderson, B. M., Fisher, R. A., and Lawrence, D. M. 2020. A machine learning approach to emulation and biophysical parameter estimation with the Community Land Model, version 5. Advances in Statistical Climatology, Meteorology and Oceanography, 6(2), 223-244.
+
 .. _Dahlinetal2015:
 
 Dahlin, K., R. Fisher, and P. Lawrence, 2015: Environmental drivers of drought deciduous phenology in the Community Land Model. Biogeosciences, 12:5061-5074.
@@ -647,6 +655,10 @@ Keller, M., Palace, M., Asner, G.P., Pereira, R., Jr. and Silva, J.N.M., 2004. C
 
 Kellner, E., Baird, A.J., Oosterwoud, M., Harrison, K. and Waddington, J.M., 2006. Effect of temperature and atmospheric pressure on methane (CH4) ebullition from near-surface peats. Geophys. Res. Lett. 33. DOI:10.1029/2006GL027509.
 
+.. _Kennedyetal2025:
+
+Kennedy, D., Dagon, K., Lawrence, D. M., Fisher, R. A., Sanderson, B. M., Collier, N., Hoffman, F. M., Koven, C. D., Kluzek, E., Levis, S., Lu, X., Oleson, K. W., Zarakas, C. M., Cheng, Y., Foster, A. C., Fowler, M. D., Hawkins, L. R., Kavoo, T., Kumar, S., Newman, A. J., Lawrence, P. J., Li, F., Lombardozzi, D. L., Lou, Y., Shuman, J. K., Swann, A. L. S., Swenson, S. C., Tand, G., Wieder, W. R., Wood, A. W., 2025. One-at-a-Time Parameter Perturbation Ensemble of the Community Land Model, Version 5.1. Journal of Advances in Modeling Earth Systems, 17(8), e2024MS004715. https://doi.org/10.1029/2024MS004715
+
 .. _Kimballetal1997:
 
 Kimball, J.S., Thornton, P.E., White, M.A. and Running, S.W. 1997. Simulating forest productivity and surface-atmosphere exchange in the BOREAS study region. Tree Physiology 17:589-599.
@@ -1624,6 +1636,10 @@ Wieder, W. R., Cleveland, C. C., Lawrence, D. M., and Bonan, G. B. 2015. Effects
 
 Williams, M., Rastetter, E.B., Fernandes, D.N., Goulden, M.L., Wofsy, S.C., Shaver, G.R., Melillo, J.M., Munger, J.W., Fan, S.M. and Nadelhoffer, K.J. 1996. Modelling the soil-plant-atmosphere continuum in a Quercus–Acer stand at Harvard Forest: the regulation of stomatal conductance by light, nitrogen and soil/plant hydraulic properties. Plant, Cell & Environment, 19: 911–927. doi:10.1111/j.1365-3040.1996.tb00456.x
 
+.. _Williamsonetal2013:
+
+Williamson, D., Goldstein, M., Allison, L., Blaker, A., Challenor, P., Jackson, L., and Yamazaki, K. 2013. History matching for exploring and reducing climate model parameter space using observations and a large perturbed physics ensemble. Climate Dynamics, 41(7), 1703–1729. https://doi.org/10.1007/s00382-013-1896-4
+
 .. _WiscombeWarren1980:
 
 Wiscombe, W.J., and Warren, S.G. 1980. A model for the spectral albedo of snow. I. Pure snow. J. Atmos. Sci. 37:2712-2733.

doc/source/tech_note/Tuning/CLM50_Tuning.rst

@@ -0,0 +1,27 @@
+.. _rst_Tuning :
+
+Parameter Tuning
+=================================
+
+CLM has a lot of parameters (over 200!). Biological processes are not governed by fundamental laws of physics and are often represented empirically. Although some parameters represent physically meaningful and measurable quantities, others are constants fit to data and are not strongly constrained. Parameter values are uncertain, they can vary across plant functional types (PFTs), and they interact through the coupled biogeophysical and biogeochemical processes that govern simulated carbon, water, and energy fluxes. 
+
+The parameter values reported in the tables throughout the CLM Tech Note are often derived from studies of a single process in isolation, or from a particular ecosystem or species, or at a spatial and temporal scale different from those at which CLM is applied. Such studies are invaluable for establishing process understanding and providing initial estimates and plausible ranges for parameters, but it’s reasonable to expect that the values need to be adjusted to transfer to a global model that aggregates diverse species into a handful of PFTs and operates at coarse resolution. 
+
+CLM is primarily developed as the land component of the Community Earth System Model (CESM), and a central goal of model development is a configuration that produces a credible simulation of the global carbon, water, and energy cycles when coupled. Achieving this requires calibrating CLM as a whole, which has typically been done through hand tuning by expert model developers. Here we describe an automated tuning methodology introduced in CLM6 by Hawkins et al. (in prep). As a result, the parameter values distributed with the CLM6 release may differ from the values listed elsewhere in this tech note. 
+
+Calibration approach
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Building on the framework established by :ref:`Dagon et al. (2020)<Dagonetal2020>`, we use an emulator-based calibration approach. We first generate a perturbed parameter ensemble (PPE) and use it to train statistical emulators that serve as computationally efficient surrogates for CLM. The emulators are then used to calibrate selected parameters against observations, and benchmarking is used to select a final calibrated parameter set from the resulting candidates. The method is summarized here in general terms; full methodological detail, parameter ranges, observational products, and diagnostic results are documented separately (Wieder et al. in prep.; Hawkins et al. in prep.).
+
+**Parameter selection and ranges.**  Parameters relevant to the calibration objectives are selected; for CLM6, this included PFT-specific parameters governing photosynthesis, stomatal conductance, respiration, allocation, and turnover. Plausible perturbation ranges for each parameter are defined following the one-at-a-time (OAAT) sensitivity framework established for CTSM by :ref:`Kennedy et al. (2025)<Kennedyetal2025>`. All other parameters are held fixed to limit the influence of calibration on processes and variables that are not explicitly targeted (e.g., hydrology). 
+
+**Ensemble generation.**  A large PPE is generated using the history matching technique :ref:`Williamson et al. (2013)<Williamsonetal2013>`. Ensemble members are run on a representative subset of gridcells (“sparse grid;” :ref:`Kennedy et al. (2025)<Kennedyetal2025>` and spun up to quasi-equilibrium before transient historical simulations are performed (1850-2023), driven with CRU-JRA meteorological forcing. The result is a set of simulations that span the relevant parameter space (see Hawkins et al. (in prep) for details on the experimental design).
+
+**Emulation.**  Because running the full model is computationally expensive, statistical emulators (e.g., Gaussian processes) are trained on the PPE output to predict simulated variables (e.g., GPP, LAI, vegetation carbon) as a continuous function of the input parameters. The emulators provide a fast and differentiable surrogate for CLM within the sampled parameter space.
+
+**Optimization.**  Parameter values are optimized by minimizing the discrepancy between emulated and observed values of the targeted variables; for CLM6, we used the biome-mean LAI, GPP, and vegetation carbon, weighting each biome equally after standardization. We employ gradient-based optimization from multiple initialization points, enabled by the differentiability of emulators. A penalty in the loss discourages parameters from moving away from their default values unless doing so meaningfully reduces model bias, so parameters that exert little influence on the targeted variables tend to remain near their documented defaults.
+
+**Evaluation and selection.**  The optimization yields an ensemble of candidate parameter sets that satisfy the observational constraints. These candidates are examined for their behavior in processes that were not directly targeted during calibration, and a small number are run and evaluated using standard land-model benchmarking tools (e.g., International Land Model Benchmarking, :ref:`Collier et al. (2018)<Collieretal2018>` to confirm that improvements in the calibrated variables are not accompanied by unacceptable degradation elsewhere. A single preferred parameter set is selected from these candidates for the model release.
+
+**Adjustment for coupled-model biases and feedbacks.**  Calibration is performed in land-only configurations. When the calibrated parameters are used in the fully coupled model, biases in other model components and land-atmosphere feedbacks can alter the simulated response. Selected parameters may therefore be adjusted further to account for biases that emerge only in the coupled system. As these adjustments depend on the configuration and on the evolving state of other model components, the released parameter values reflect this additional tuning and may not correspond exactly to the values obtained from the land-only calibration.

doc/source/tech_note/index.rst

@@ -49,5 +49,6 @@ CLM Technical Note
    Dust/CLM50_Tech_Note_Dust.rst
    Isotopes/CLM50_Tech_Note_Isotopes.rst
    Land-Only_Mode/CLM50_Tech_Note_Land-Only_Mode.rst
+   Tuning/CLM50_Tuning.rst
    References/CLM50_Tech_Note_References.rst