From e08fb668ea6232c842911fc4beaac84db026abdf Mon Sep 17 00:00:00 2001
From: "Daniel.Frisinghelli" <daniel.frisinghelli@eurac.edu>
Date: Wed, 18 Aug 2021 08:51:19 +0000
Subject: [PATCH] Added metrics on daily basis.
---
Notebooks/eval_temperature.ipynb | 33 ++++++++++++++++++++++++++++----
1 file changed, 29 insertions(+), 4 deletions(-)
diff --git a/Notebooks/eval_temperature.ipynb b/Notebooks/eval_temperature.ipynb
index 4a2e344..c319d39 100644
--- a/Notebooks/eval_temperature.ipynb
+++ b/Notebooks/eval_temperature.ipynb
@@ -35,8 +35,8 @@
"PREDICTAND = 'tasmin'\n",
"MODEL = 'USegNet'\n",
"PPREDICTORS = 'ztuvq'\n",
- "PLEVLES = ['500', '850']\n",
- "SPREDICTORS = ''\n",
+ "PLEVELS = ['500', '850']\n",
+ "SPREDICTORS = 'pt2'\n",
"DEM = 'dem'\n",
"DOY = 'doy'"
]
@@ -106,7 +106,7 @@
"outputs": [],
"source": [
"# model predictions and observations NetCDF\n",
- "y_pred = TARGET_PATH.joinpath(PREDICTAND, '_'.join([MODEL, PREDICTAND, PPREDICTORS, PLEVELS, SPREDICTORS, DEM, DOY]) + '.nc')\n",
+ "y_pred = TARGET_PATH.joinpath(PREDICTAND, '_'.join([MODEL, PREDICTAND, PPREDICTORS, *PLEVELS, SPREDICTORS, DEM, DOY]) + '.nc')\n",
"if PREDICTAND == 'tas':\n",
" # read both tasmax and tasmin\n",
" tasmax = xr.open_dataset(search_files(OBS_PATH.joinpath('tasmax'), '.nc$').pop())\n",
@@ -175,6 +175,7 @@
"y_pred_avg = y_pred.mean(dim='time')\n",
"y_true_avg = y_true.mean(dim='time')\n",
"bias = y_pred_avg - y_true_avg\n",
+ "# bias = (y_pred - y_true).mean()\n",
"for var in bias:\n",
" print('Overall average bias {}: {:.2f}'.format(var, bias[var].mean().item()))"
]
@@ -188,6 +189,7 @@
"source": [
"# mean absolute error over reference period\n",
"mae = np.abs(y_pred_avg - y_true_avg).mean()\n",
+ "# mae = np.abs(y_pred - y_true).mean()\n",
"for var in mae:\n",
" print('Mean absolute error {}: {:.2f}'.format(var, mae[var].item()))"
]
@@ -201,6 +203,7 @@
"source": [
"# root mean squared error over reference period\n",
"rmse = ((y_pred_avg - y_true_avg) ** 2).mean()\n",
+ "# rmse = ((y_pred - y_true) ** 2).mean()\n",
"for var in rmse:\n",
" print('Root mean squared error {}: {:.2f}'.format(var, rmse[var].item()))"
]
@@ -215,7 +218,9 @@
"# Pearson's correlation coefficient over reference period\n",
"for var in y_pred_avg:\n",
" y_p = y_pred_avg[var].values[~np.isnan(y_pred_avg[var])]\n",
+ " # y_p = y_pred[var].values[~np.isnan(y_pred[var])]\n",
" y_t = y_true_avg[var].values[~np.isnan(y_true_avg[var])]\n",
+ " # y_t = y_true[var].values[~np.isnan(y_true[var])]\n",
" r, _ = stats.pearsonr(y_p, y_t)\n",
" print('Pearson correlation for {}: {:.2f}'.format(var, r))"
]
@@ -314,6 +319,16 @@
"plt.tight_layout()"
]
},
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "ec2f6293-384e-41b6-9e79-75bd852df0c0",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# TODO: Annual cylce"
+ ]
+ },
{
"cell_type": "markdown",
"id": "c70b369d-2d16-42e3-9300-4a18757ad1b2",
@@ -322,6 +337,16 @@
"### Bias of extreme values"
]
},
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "4acfc3f2-20ed-498c-ab35-f392ae0e64f9",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# TODO: smooth quantiles"
+ ]
+ },
{
"cell_type": "code",
"execution_count": null,
@@ -330,7 +355,7 @@
"outputs": [],
"source": [
"# percentiles of interest\n",
- "percentiles = [0.01, 0.02, 0.5, 0.98, 0.99]"
+ "percentiles = [0.01, 0.02, 0.98, 0.99]"
]
},
{
--
GitLab