Merge pull request #851 from PCMDI/741_ljw_cloud

Cloud feedback metric

pcmdi_metrics/cloud_feedback/README.md

@@ -0,0 +1,71 @@
+# Cloud Feedback Metrics 
+
+The clould feedback metrics implemented to PMP is originally from [`assessed-cloud-fbks`](https://github.com/mzelinka/assessed-cloud-fbks)[![DOI](https://zenodo.org/badge/353136800.svg)](https://zenodo.org/badge/latestdoi/353136800), developed by [@mzelinka](https://mzelinka.github.io/) at LLNL. This code performs the analysis of [Zelinka et al. (2022)](https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2021JD035198). It computes GCM cloud feedback components and compares them to the expert-assessed values from [Sherwood et al. (2020)](https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019RG000678).
+
+## Instructions
+To use, please consider follow these steps:
+
+#### 1. Install PMP
+Install PMP via conda following [these instructions](http://pcmdi.github.io/pcmdi_metrics/install.html). For example:
+```
+conda create -n [YOUR_CONDA_ENVIRONMENT] -c conda-forge pcmdi_metrics
+``` 
+
+#### 2. Activate PMP installed conda environment
+Activate your environment (if PMP installed env is differnt than your current one):
+```
+conda activate [YOUR_CONDA_ENVIRONMENT]
+```
+
+#### 3. Clone PMP repo to your local
+Clone PMP repo to your local for pre-calculated data:
+```
+git clone https://github.com/PCMDI/pcmdi_metrics
+```
+
+Once completed, go to `pcmdi_metrics/pcmdi_metrics/cloud_feedback` directory
+```
+cd [YOUR LOCAL CLONED PMP REPOSITORY]
+cd pcmdi_metrics/pcmdi_metrics/cloud_feedback
+```
+
+#### 4. Edit parameter files
+In [`param/my_param.py`](param/my_param.py), update the "User Input" section so it points to your model's amip and amip-p4K files.
+```python
+
+# User Input:
+# ================================================================================================
+model = "GFDL-CM4"
+variant = "r1i1p1f1"
+
+input_files_json = "./param/input_files.json"
+
+# Flag to compute ECS
+# True: compute ECS using abrupt-4xCO2 run
+# False: do not compute, instead rely on ECS value present in the json file (if it exists)
+# get_ecs = True
+get_ecs = False
+
+# Output directory path (directory will be generated if it does not exist yet.)
+xml_path = "./xmls/"
+figure_path = "./figures/"
+output_path = "./output"
+output_json_filename = "_".join(["cloud_feedback", model, variant]) + ".json"
+# ================================================================================================
+```
+
+You will need to update [`param/input_files.json`](param/input_files.json) file as well to provide data path for your input files. 
+
+#### 5. Run the code and inspect the generated output files 
+Run calculation:
+```
+python cloud_feedback_driver.py -p param/my_param.py
+```
+
+Once code is completed, check `output` directory (`output_path` from the above parameter file) for JSON and `figures` directory (`figure_path` from the above parameter file) for figures and text tables.
+
+
+## References
+- Zelinka et al. (2022): [Evaluating climate models’ cloud feedbacks against expert judgement](https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2021JD035198), <em>J. Geophys. Res.</em>, 127, e2021JD035198, doi:10.1029/2021JD035198.
+
+- Sherwood et al. (2020): [A combined assessment of Earth’s climate sensitivity](https://agupubs.onlinelibrary.wiley.com/doi/abs/10.1029/2019RG000678), <em>Rev. Geophys.</em>, 58, e2019RG000678, doi:10.1029/2019RG000678.

pcmdi_metrics/cloud_feedback/cloud_feedback_driver.py

@@ -0,0 +1,197 @@
+#!/usr/bin/env python
+
+import json
+import os
+from collections import OrderedDict
+
+import pcmdi_metrics.cloud_feedback.lib.cld_fbks_ecs_assessment_v3 as dataviz
+from pcmdi_metrics.cloud_feedback.lib import (
+    AddParserArgument,
+    CloudRadKernel,
+    cloud_feedback_metrics_to_json,
+    compute_ECS,
+    organize_ecs_jsons,
+    organize_err_jsons,
+    organize_fbk_jsons,
+)
+
+# collect user input
+param = AddParserArgument()
+
+model = param.model
+institution = param.institution
+variant = param.variant
+grid_label = param.grid_label
+version = param.version
+input_files_json = param.input_files_json
+path = param.path
+xml_path = param.xml_path
+data_path = param.data_path
+figure_path = param.figure_path
+output_path = param.output_path
+output_json_filename = param.output_json_filename
+get_ecs = param.get_ecs
+debug = param.debug
+
+cmec = False
+if hasattr(param, "cmec"):
+    cmec = param.cmec  # Generate CMEC compliant json
+print("CMEC:" + str(cmec))
+
+print("model:", model)
+print("institution:", institution)
+print("variant:", variant)
+print("grid_label:", grid_label)
+print("version:", version)
+print("path:", path)
+print("input_files_json:", input_files_json)
+print("xml_path:", xml_path)
+print("figure_path:", figure_path)
+print("output_path:", output_path)
+print("output_json_filename:", output_json_filename)
+print("get_ecs:", get_ecs)
+print("debug:", debug)
+
+if get_ecs:
+    exps = ["amip", "amip-p4K", "piControl", "abrupt-4xCO2"]
+else:
+    exps = ["amip", "amip-p4K"]
+
+# generate xmls pointing to the cmorized netcdf files
+os.makedirs(xml_path, exist_ok=True)
+
+filenames = dict()
+
+if input_files_json is not None:
+    with open(input_files_json) as f:
+        ncfiles = json.load(f)
+else:
+    print('Warning: input files were not explicitly given. They will be searched from ', path)
+
+for exp in exps:
+    filenames[exp] = dict()
+    if exp == "amip-p4K":
+        activity = "CFMIP"
+    else:
+        activity = "CMIP"
+    if "amip" in exp:
+        fields = [
+            "tas",
+            "rsdscs",
+            "rsuscs",
+            "wap",
+            "clisccp",
+        ]  # necessary for cloud feedback calcs
+    else:
+        fields = ["tas", "rlut", "rsut", "rsdt"]  # needed for ECS calc
+    for field in fields:
+        if field == "clisccp":
+            table = "CFmon"
+        else:
+            table = "Amon"
+
+        if input_files_json is None:  # PCMDI internal setup
+            searchstring = os.path.join(
+                path,
+                activity,
+                institution,
+                model,
+                exp,
+                variant,
+                table,
+                field,
+                grid_label,
+                version,
+                "*.nc",
+            )
+        else:
+            searchstring = os.path.join(
+                ncfiles[exp][field]["path"], ncfiles[exp][field]["file"]
+            )
+        xmlname = os.path.join(xml_path, ".".join([exp, model, variant, field, "xml"]))
+        os.system("cdscan -x " + xmlname + " " + searchstring)
+        filenames[exp][field] = xmlname
+
+if debug:
+    with open(os.path.join(output_path, "filenames.json"), "w") as f:
+        json.dump(filenames, f, sort_keys=True, indent=4)
+
+# calculate all feedback components and Klein et al (2013) error metrics:
+fbk_dict, obsc_fbk_dict, err_dict = CloudRadKernel(filenames)
+
+print("calc done")
+
+# add this model's results to the pre-existing json file containing other models' results:
+updated_fbk_dict, updated_obsc_fbk_dict = organize_fbk_jsons(
+    fbk_dict, obsc_fbk_dict, model, variant, datadir=data_path
+)
+updated_err_dict = organize_err_jsons(err_dict, model, variant, datadir=data_path)
+
+ecs = None
+if get_ecs:
+    # calculate ECS and add it to the pre-existing json file containing other models' results:
+    ecs = compute_ECS(filenames)
+    print("calc ECS done")
+    print("ecs: ", ecs)
+updated_ecs_dict = organize_ecs_jsons(ecs, model, variant, datadir=data_path)
+
+os.makedirs(output_path, exist_ok=True)
+if debug:
+    with open(os.path.join(output_path, "fbk_dict.json"), "w") as f:
+        json.dump(fbk_dict, f, sort_keys=True, indent=4)
+    with open(os.path.join(output_path, "err_dict.json"), "w") as f:
+        json.dump(err_dict, f, sort_keys=True, indent=4)
+    with open(os.path.join(output_path, "updated_err_dict.json"), "w") as f:
+        json.dump(updated_err_dict, f, sort_keys=True, indent=4)
+    with open(os.path.join(output_path, "updated_fbk_dict.json"), "w") as f:
+        json.dump(updated_fbk_dict, f, sort_keys=True, indent=4)
+    with open(os.path.join(output_path, "updated_ecs_dict.json"), "w") as f:
+        json.dump(updated_ecs_dict, f, sort_keys=True, indent=4)
+
+# generate plots and extract metrics from the plotting routines
+os.makedirs(figure_path, exist_ok=True)
+climo_cld_rmse, cld_fbk_rmse, assessed_cld_fbk, ecs = dataviz.make_all_figs(
+    updated_fbk_dict,
+    updated_obsc_fbk_dict,
+    updated_err_dict,
+    updated_ecs_dict,
+    model,
+    figdir=figure_path,
+    datadir=data_path,
+    debug=debug,
+)
+print("get metrics done")
+
+# save final metrics and accompanying important statistics in JSON format
+print("-- Metric result --")
+print("model:", model)
+print("variant:", variant)
+print("clim_cloud_rmse:", climo_cld_rmse)
+print("cloud_feedback_rmse:", cld_fbk_rmse)
+print("assessed_cloud_feedback:", assessed_cld_fbk)
+print("ecs:", ecs)
+
+output_dict = OrderedDict()
+output_dict["RESULTS"] = OrderedDict()
+output_dict["RESULTS"][model] = OrderedDict()
+output_dict["RESULTS"][model][variant] = OrderedDict()
+output_dict["RESULTS"][model][variant]["assessed_cloud_feedback"] = OrderedDict()
+output_dict["RESULTS"][model][variant]["assessed_cloud_feedback"]["high_cloud_altitude"] = assessed_cld_fbk[0]
+output_dict["RESULTS"][model][variant]["assessed_cloud_feedback"]["tropical_marine_low_cloud"] = assessed_cld_fbk[1]
+output_dict["RESULTS"][model][variant]["assessed_cloud_feedback"]["tropical_anvil_cloud_area"] = assessed_cld_fbk[2]
+output_dict["RESULTS"][model][variant]["assessed_cloud_feedback"]["land_cloud_amount"] = assessed_cld_fbk[3]
+output_dict["RESULTS"][model][variant]["assessed_cloud_feedback"]["middle_latitude_marine_low_cloud_amount"] = assessed_cld_fbk[4]
+output_dict["RESULTS"][model][variant]["assessed_cloud_feedback"]["high_latitude_low_cloud_optical_depth"] = assessed_cld_fbk[5]
+output_dict["RESULTS"][model][variant]["assessed_cloud_feedback"]["implied_unassessed"] = assessed_cld_fbk[6]
+output_dict["RESULTS"][model][variant]["assessed_cloud_feedback"]["sum_of_assessed"] = assessed_cld_fbk[7]
+output_dict["RESULTS"][model][variant]["assessed_cloud_feedback"]["total_cloud_feedback"] = assessed_cld_fbk[8]
+output_dict["RESULTS"][model][variant]["clim_cloud_rmse"] = climo_cld_rmse
+output_dict["RESULTS"][model][variant]["cloud_feedback_rmse"] = cld_fbk_rmse
+output_dict["RESULTS"][model][variant]["equilibrium_climate_sensitivity"] = ecs
+
+
+cloud_feedback_metrics_to_json(
+    output_path, output_json_filename, output_dict, cmec_flag=cmec
+)
+
+print("Done!")