BiomeEP: make ready for calibration to DBH distributions by fabern · Pull Request #330 · geco-bern/rsofun

fabern · 2026-04-23T10:41:18Z

This PR results from changes in

BiomeE attenuate aggressive cohort fusion #327 attenuate aggressive cohort fusion
BiomeEP: expose additional parameters to R #328 cleanup of input parameters (this is breaking. Essentially it clarifies ignored parameters and attenuates hardcoded parameters. Allowing to specify the previously hardcoded params. If not specified these hardcoded parameters remain the same as previously. This is needed for the spin-up of the initial vegetation structure to achieve a specific DBH distribution.)
additional changes from this PR BiomeEP: make ready for calibration to DBH distributions #330: add further output needed to restart simulations. This is needed for the spin-up of the initial vegetation structure to achieve a specific DBH distribution.

All the changes that are user-facing are documented in NEWS.md.

Specifics for this PR:

This PR adds print statements for debugging the restarting.

Validation:

With a better diff tool than GitHub (e.g. such as GtiKraken), the change in file rsofun/tests/testthat/_snaps/model-runs-snapshot-biomee.md shows:

negligible changes in biomee_res$out_daily_tile results at year 1 and at year 251
negligible changes in biomee_res$out_annual_tile results
negligible changes in biomee_res$out_annual_cohort results at year 1 (initialization) but
some changes in the biomee_res$out_annual_cohort results at year
- due to a different demographic structure:
  - 7 cohorts instead of 4 in mod_BiomeE_Pmodel_oac_yr251 and mod_BiomeE_PLULUC_primary_oac_yr251,
  - and 11 cohorts instead of 10 in mod_BiomeE_gsLeun_oac_yr251

~~Current status: test fails. ~~

~~Judging the print output of the test, it appears that~~

~~cohorts are correctly initialized.~~
~~cohorts are (in printout) not yet summarised on tile level (e.g. pleaf_c=0)~~
~~dtemp_pvy and wscal_pvy are not initialized~~
~~wscal_alldays is different~~

… level

Previously, cohorts within 20% relative DBH difference were merged. This has been reduced to 2%, with the aim to enable comparison with observed size distributions.

Thereby removing the more complicated relative threshold applied previously.

…ng strategy

Colum order of these variables was wrong: "NSN","seedC","leafC","rootC","sapwoodC","heartwoodC"

Caused by warning: ! `case_match()` was deprecated in dplyr 1.2.0. i Please use `recode_values()` instead.

── Failure ('test-model-runs.R:64:3'): biomee output check (gs leuning) ──────── Expected `all.equal(...)` to be TRUE. Differences: `actual` is a character vector ('Mean relative difference: 0.0002914499') `expected` is a logical vector (TRUE)

Following deprecation warnings starting with devtools 2.5.0 and following the instructions on https://devtools.r-lib.org/reference/install-deprecated.html

This streamlines the input preparation of the BiomeEP simulation

Namely: 'phenotype','Vmax','alphaBM','leafLS','lAImax','CNleaf0'

I.e. store them as part of the params_species_biomee and params_tile_biomee. Currently still hardcoded, but this allows for exposure of these in a next step.

… API) This change is currently undocumented, since without specifying these parameters, BiomeEP continues to use the former hardcoded values as default. This concerns parameters: ("kphio","beta","rd_to_vcmax","kc_jmax","kphio_par_a","kphio_par_b") in params_species and ("tau_acclim") in params_tile. Small fix

Namely: 'gamma_L','Vannual','betaOFF','betaON', 'leaf_size' (all unused)

…th `mortrate_d_c`) Change value of mortrate_d_c in the example BiomeEP data set

This allows to use gdb instead of lldb. lldb has some shortcomings with Fortran, namely it can't explore variables at a breakpoint. The recommendation is therefore to use gdb. Since this is not available for arm-based "M4" Macs, we run an arm-based Linux in a Docker container. Use VSCode extension "Dev Containers" together with the specified Docker file. Step 1: "Dev Containers: Reopen in Container" Step 2: set a breakpoin in code. Step 3: go to the debug Tab in VSCode and click Play on the correct debug action "(gdb) Launch - MacBook Air M4".

…ation year on)

…ohort_age`) If not provided it recovers old default of 0 years old at initialization.

…rt-with-full-initialization

…tion of NSC pool Since the function to compute bl_max and br_max war recently moved from Fortran to R (for the restarting), there had been a small bug (used `max()` instead of `pmax()`) which changed the initialization of NSC pools in case there were multiple species and one of them had LAI_light larger than 0.5.

…ablish-restart-with-full-initialization

fabern

@fabern Please check the comments added within self code review.

fabern · 2026-09-24T09:11:44Z

 params_species <- tibble(
  lifeform          = c(0, 1, 1, 1, 1),    # 0: grass; 1 Woody
-  phenotype         = c(0, 0, 1, 1, 1),    # 0: Deciduous; 1 Evergreen
+  phenotype         = c(0, 1, 1, 1, 1),    # 0: Deciduous; 1 Evergreen


Why was this changed? Better to change it back so that we still have a deciduous tree species in the example driver.

Yes, do this in a next commit.

fabern · 2026-09-24T11:34:26Z

    do_relayer = .false.

-    ! update vegn GDD and tk_pheno
-    vegn%gdd      = vegn%gdd + max(0.0, vegn%tk_daily - 278.15)


TODO: check why this could be removed

This was removed because it was unused. The gdd (growing degrees days) we use is actually computed on the cohort level cc%gdd. It would actually not be needed to be computed on the cohort level since it anyway uses a hardcoded threshold of 5 degC. But technically that approach would enable a species-specific critical threshold for GDD accumulation.

fabern · 2026-09-24T11:40:13Z

+    self%out_annual_cohorts(:, ANNUAL_COHORTS_ID) = 0
+    self%out_annual_cohorts(:, ANNUAL_COHORTS_CID) = 0


Is this change needed/usesful? If I remember correctly this was added to diagnose during debugging.

No, it is not needed. It would give addtional safety if a row in the output corresponds to an existing cohort, but at slight additional code and computational complexity. => not needed.

We need a way to filter out on R side rows of out_annual_cohorts that were not filled by the output loop over all cohorts. These rows exist because there can be less cohorts simulated than NCohortMax. This filtering is done on column ANNUAL_COHORTS_YEAR. If this is zero then it indicates that the row was not overwritten. We don't need ANNUAL_COHORTS_ID or ANNUAL_COHORTS_CID for that.

fabern · 2026-09-24T11:42:30Z

 # BiomeE: test model outputs, repeatability and agreement with reference values from subfolder ./_snaps/
 set.seed(10)

+test_that("BiomeEP Tksoil initial days (rolling mean, smoothed)", {


Is it useful to have this test?

I think it does not hurt.

… default simulation by changing This also switches back the second species to be deciduous. And the third species to be evergreen. NOTE: tests only use evergreen species.

fabern · 2026-09-24T16:31:58Z

This is ready to merge.

fabern added 30 commits February 16, 2026 15:27

Mark alphaBM as ineffective parameter

29118f1

Mark BiomeE wrong seedC,leafC,rootC,sapwoodC,heartwoodC,NSN on cohort…

78a97a2

… level

Clarify non-imposed `NCohortMax

904c7eb

Add comments

93e7ef2

Reduce cohort merging (20% => 2%) relative threshold

b648271

Previously, cohorts within 20% relative DBH difference were merged. This has been reduced to 2%, with the aim to enable comparison with observed size distributions.

Simplify cohort merging considering only absolute DBH differences (2 cm)

fde85af

Thereby removing the more complicated relative threshold applied previously.

BiomeE: Update example output and snapshot tests for new cohort mergi…

f85017d

…ng strategy

BiomeE: fix some variables in annual cohort output

1e14408

Colum order of these variables was wrong: "NSN","seedC","leafC","rootC","sapwoodC","heartwoodC"

For dplyr 1.2.0: replace case_match() => recode_values()

d1cf47c

Caused by warning: ! `case_match()` was deprecated in dplyr 1.2.0. i Please use `recode_values()` instead.

Fix pkgdown devtools::install_github() => pak::pak()

b9efdc2

Following deprecation warnings starting with devtools 2.5.0 and following the instructions on https://devtools.r-lib.org/reference/install-deprecated.html

Update NEWS.md

adb62a5

Add build_site_info() to run_biomee_f_bysite()

28ed327

This streamlines the input preparation of the BiomeEP simulation

Update .vscode

e2fcf33

Mark currently unused species parameters

e650cc1

Warn if user tries to provide dummy 'params_species' parameters

2a5be9f

Namely: 'phenotype','Vmax','alphaBM','leafLS','lAImax','CNleaf0'

Remove dummy 'params_species' parameters from example data

40f8f0d

Namely: 'phenotype','Vmax','alphaBM','leafLS','lAImax','CNleaf0'

BiomeEP: define GPP p-model parameter on species/tile level

e229073

I.e. store them as part of the params_species_biomee and params_tile_biomee. Currently still hardcoded, but this allows for exposure of these in a next step.

Remove further dummy 'params_species' parameters from example data

8bd7219

Namely: 'gamma_L','Vannual','betaOFF','betaON', 'leaf_size' (all unused)

Streamline use of mortality parameters

2d57cbf

Remove par_mort_under (which defaulted anyway to 1)

bdb692e

Remove par_mort (which defaulted anyway to 1 and is now replaced wi…

2b3cc00

…th `mortrate_d_c`) Change value of mortrate_d_c in the example BiomeEP data set

Temporarily activate CI on this branch

d09e5d3

fix docs

f3f50cd

Initialize tree distributions at maturalage (reprod. from first simul…

aa6a3be

…ation year on)

Initialize tree distribution age with new column (`init_cohort$init_c…

3f9e9e0

…ohort_age`) If not provided it recovers old default of 0 years old at initialization.

Cosmetics

895c584

Cosmetics II

67774df

fabern added 6 commits September 18, 2026 13:02

Clarify DBH criteria in NEWS.md

bbeab7e

Update vignette on BiomeEP use

c692c5d

Merge branch 'biomee-redefine-parameters' into biomee-establish-resta…

312e677

…rt-with-full-initialization

feat: [BiomeEP] add validity check on init_cohort species

4f4b05d

refactor: [BiomeEP] clarify role of NCohortMax in comments

f1dc249

fabern force-pushed the biomee-redefine-parameters branch from f38c7c8 to 072fe00 Compare September 23, 2026 09:10

Merge branch 'biomee-attenuate-aggress-cohort-fusion' into biomee-est…

83dd023

…ablish-restart-with-full-initialization

fabern mentioned this pull request Sep 23, 2026

BiomeE attenuate aggressive cohort fusion #327

Closed

1 task done

fabern marked this pull request as ready for review September 23, 2026 09:37

fabern mentioned this pull request Sep 23, 2026

BiomeEP: expose additional parameters to R #328

Closed

fabern changed the title ~~BiomeEP: Add further output useful to restart~~ BiomeEP: make ready for calibration to DBH distributions Sep 23, 2026

fabern added 2 commits September 23, 2026 12:01

update biomee output due to merge for previous commit

f7ad64e

Mention init_cohort$species check in NEWS.md

85ad793

fabern changed the base branch from biomee-redefine-parameters to master September 23, 2026 10:08

fabern added 3 commits September 24, 2026 12:56

docs: update NEWS.md

edfda90

refactor: Undo some formatting changes

995e186

docs: [BiomeEP] Add (optional) P-model parameters to documentation

6cc4c10

fabern commented Sep 24, 2026

View reviewed changes

fabern added 4 commits September 24, 2026 13:45

[BiomeEP}: move debug code to debug_print.f90 (unused)

6ceab1b

docs: [BiomeEP] add some comments

6c79668

feat: [BiomeEP] include grass species in example data sets, keep same…

6ba635e

… default simulation by changing This also switches back the second species to be deciduous. And the third species to be evergreen. NOTE: tests only use evergreen species.

test: add a BiomeEP deciduous species to snapshot tests

89a30d0

fabern force-pushed the biomee-establish-restart-with-full-initialization branch from cf52b9d to 89a30d0 Compare September 24, 2026 15:52

test: activate test for restart of BiomeEP simulation

56e15b5

fabern force-pushed the biomee-establish-restart-with-full-initialization branch from 0232080 to 56e15b5 Compare September 24, 2026 16:30

fabern mentioned this pull request Sep 24, 2026

BiomeEP: Expose further hardcoded parameters #343

Open

2 tasks

fabern merged commit 0bf91fc into master Sep 25, 2026
7 checks passed

fabern deleted the biomee-establish-restart-with-full-initialization branch September 25, 2026 07:35

Navigation Menu

Repository navigation

BiomeEP: make ready for calibration to DBH distributions - #330

BiomeEP: make ready for calibration to DBH distributions#330
fabern merged 80 commits into
masterfrom
biomee-establish-restart-with-full-initialization

fabern commented Apr 23, 2026 •

edited

Loading

Uh oh!

fabern left a comment

Uh oh!

fabern Sep 24, 2026

Uh oh!

fabern Sep 24, 2026

Uh oh!

fabern Sep 24, 2026

Uh oh!

fabern Sep 24, 2026

Uh oh!

fabern Sep 24, 2026

Uh oh!

fabern Sep 24, 2026

Uh oh!

fabern Sep 24, 2026

Uh oh!

fabern Sep 24, 2026

Uh oh!

fabern commented Sep 24, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

		self%out_annual_cohorts(:, ANNUAL_COHORTS_ID) = 0
		self%out_annual_cohorts(:, ANNUAL_COHORTS_CID) = 0

Conversation

fabern commented Apr 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Specifics for this PR:

Uh oh!

fabern left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fabern commented Sep 24, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

fabern commented Apr 23, 2026 •

edited

Loading