doc fig

.github/workflows/build-linux.yml

@@ -1,9 +1,9 @@
 name: build-linux
 
 on:
-    # push:
-    #     branches:
-    #         - main
+    push:
+        branches:
+            - main
     release:
         types: 
             - published

.github/workflows/build-macos.yml

@@ -1,9 +1,9 @@
 name: build-macos
 
 on:
-    # push:
-    #     branches:
-    #         - main
+    push:
+        branches:
+            - main
     release:
         types: 
             - published

.github/workflows/build-windows.yml

@@ -1,9 +1,9 @@
 name: build-windows
 
 on:
-    # push:
-    #     branches:
-    #         - main
+    push:
+        branches:
+            - main
     release:
         types: 
             - published

.github/workflows/check-manifest.yml

@@ -1,9 +1,9 @@
 name: "Check Manifest"
 
 on:
-    # push:
-    #     branches:
-    #         - main
+    push:
+        branches:
+            - main
     release:
         types: [published]
 

.github/workflows/deploy-conda.yml

@@ -11,7 +11,7 @@ on:
 jobs:
     build_conda:
         # runs-on: ubuntu-latest
-        runs-on: ubuntu-20.04
+        runs-on: ubuntu-20.04  # github action's runner process fails with error: action is terminated. People suggested to not use ubuntu 22.04. But it also fails on 20.04 I tried.
         name: Publish to Conda
         strategy:
             fail-fast: false

.github/workflows/deploy-docs.yml

@@ -1,9 +1,9 @@
 name: deploy-docs
 
 on:
-    # push:
-    #     branches:
-    #         - main
+    push:
+        branches:
+            - main
     pull_request:
         branches:
             - main

.github/workflows/deploy-pypi.yml

@@ -1,9 +1,9 @@
 name: deploy-pypi
 
 on:
-    # push:
-    #     branches:
-    #         - main
+    push:
+        branches:
+            - main
     release:
         types:
             - published

docs/source/examples.rst

@@ -77,60 +77,60 @@ The above script generates the output file ``output.nc`` that contains all gener
    :figwidth: 100%
    :class: custom-dark
 
-   **Figure 2:** *Panels (a) and (b) show the east and north components of the ocean’s current velocity in the upper 0.3 m th to 2.5 m range, as measured by HF radars in Monterey Bay on January 25 , 2017, at 3:00 UTC. The data has been averaged hourly and mapped to a 2 km resolution Cartesian grid using unweighted least squares. The regions inside the solid black curves represent missing data that was filtered out due to high GDOP values from the original measurement. Panels (c) and (d) respectively show the east and north components of the velocity error computed for the locations where velocity data is available in Panels (a) and (b).*
+   **Figure 2:** *Panels (a) and (b) show the east and north components of the ocean's current velocity as measured by HF radars in Monterey Bay on January 25-th, 2017, at 3:00 UTC. The regions inside the solid black curves represent missing data that was filtered out due to high GDOP values from the original measurement. Panels (c) and (d) respectively show the east and north components of the velocity error computed for the locations where velocity data is available in Panels (a) and (b).*
 
 .. figure:: _static/images/plots/rbf_kernel_2d.png
    :align: left
    :figwidth: 100%
    :width: 90%
    :class: custom-dark
 
-   **Figure 3:** *The red fields represent the calculated spatial autocorrelation α for the east (a) and north (b) velocity data. The elliptical contour curves are the best fit of the exponential kernel ρ to the autocorrelation. The direction of the principal radii of ellipses is determined by the eigenvectors of M, representing the principal direction of correlation. The radii values are proportional to the eigenvalues of M, representing the correlation length scale. The axes are in the unit of data points spaced 2 km apart.*
+   **Figure 3:** *The red fields represent the calculated spatial autocorrelation α for the east (a) and north (b) velocity data. The elliptical contour curves are the best fit of the exponential kernel* :math:`\rho` *to the autocorrelation. The direction of the principal radii of ellipses is determined by the eigenvectors of* :math:`\boldsymbol{M}`, *representing the principal direction of correlation. The radii values are proportional to the eigenvalues of* :math:`\boldsymbol{M}`, *representing the correlation length scale. The axes are in the unit of data points spaced 2 km apart.*
 
 .. figure:: _static/images/plots/cor_cov.png
    :align: left
    :figwidth: 100%
    :width: 90%
    :class: custom-dark
 
-   **Figure 4:** *Correlation (first column) and covariance matrices (second column) of the east (first row) and north (second row) datasets are shown. The size of matrices are n = 485.*
+   **Figure 4:** *Correlation (first column) and covariance matrices (second column) of the east (first row) and north (second row) datasets are shown. The size of matrices are* :math:`n = 485`.
 
 .. figure:: _static/images/plots/kl_eigenvectors.png
    :align: left
    :figwidth: 100%
    :class: custom-dark
 
-   **Figure 5:** *The first 12 spatial eigenfunctions φi for the east velocity dataset (first and second rows) and north velocity dataset (third and fourth rows) are shown in the domain Ω in the Monterey Bay. The black curves is indicate the boundary of the missing domain Ω◦. We note that the oblique pattern in the east eigenfunctions is attributed to the anisotropy of the east velocity data, as illustrated in Figure 3a.*
+   **Figure 5:** *The first 12 spatial eigenfunctions* :math:`\phi_i` *for the east velocity dataset (first and second rows) and north velocity dataset (third and fourth rows) are shown in the domain* :math:`\Omega` *in the Monterey Bay. The black curves is indicate the boundary of the missing domain* :math:`\Omega_{\circ}`. *We note that the oblique pattern in the east eigenfunctions is attributed to the anisotropy of the east velocity data, as illustrated in Figure 3a.*
 
 .. figure:: _static/images/plots/ensemble.png
    :align: left
    :figwidth: 100%
    :class: custom-dark
 
-   **Figure 6:** *The reconstructed central ensemble (first column), mean of reconstructed ensemble (second column), and the standard deviation of reconstructed ensemble (third column) are shown in both Ω and Ω◦. The boundary of Ω◦ is shown by the solid black curve. The first and second rows correspond to the east and north velocity data, respectively.*
+   **Figure 6:** *The reconstructed central ensemble (first column), mean of reconstructed ensemble (second column), and the standard deviation of reconstructed ensemble (third column) are shown in both* :math:`\Omega` *and* :math:`\Omega_{\circ}`. *The boundary of* :math:`\Omega_{\circ}` *is shown by the solid black curve. The first and second rows correspond to the east and north velocity data, respectively.*
 
 .. figure:: _static/images/plots/deviation.png
    :align: left
    :figwidth: 100%
    :class: custom-dark
 
-   **Figure 7:** *The left to right columns show the plots of deviations d1(x), d2(x), d3(x), and d4(x), displayed in both domains Ω and Ω◦ with the first and second rows representing the east and north datasets, respectively. The solid black curve shows the boundary of Ω◦. The absolute values smaller than 10−8 are rendered as transparent and expose the ocean background, which includes the domain Ω for the first three deviations.*
+   **Figure 7:** *The left to right columns show the plots of deviations* :math:`d_1(\boldsymbol{x})`, :math:`d_2(\boldsymbol{x})`, :math:`d_3(\boldsymbol{x})`, *and* :math:`d_4(\boldsymbol{x})`, *displayed in both domains* :math:`\Omega` *and* :math:`\Omega_{\circ}` *with the first and second rows representing the east and north datasets, respectively. The solid black curve shows the boundary of* :math:`\Omega_{\circ}`. *The absolute values smaller than* :math:`10^{-8}` *are rendered as transparent and expose the ocean background, which includes the domain* :math:`\Omega` *for the first three deviations.*
 
 .. figure:: _static/images/plots/ensemble_js_distance.png
    :align: left
    :figwidth: 100%
    :width: 90%
    :class: custom-dark
 
-   **Figure 8:** *The JS distance between the expected distribution q(x, ξ) and the observed distribution p(x, ξ) is shown. The absolute values smaller than 10−8 are rendered as transparent and expose the ocean background, which includes the domain Ω where the JS distance between p(x, ξ) and q(x, ξ) is zero.*
+   **Figure 8:** *The JS distance between the expected distribution* :math:`q(\boldsymbol{x}, \xi)` *and the observed distribution* :math:`p(\boldsymbol{x}, \xi)` *is shown. The absolute values smaller than* :math:`10^{-8}` *are rendered as transparent and expose the ocean background, which includes the domain* :math:`\Omega` *where the JS distance between* :math:`p(\boldsymbol{x}, \xi)` *and* :math:`q(\boldsymbol{x}, \xi)` *is zero.*
 
 .. figure:: _static/images/plots/kl_eigenvalues.png
    :align: left
    :figwidth: 100%
    :width: 70%
    :class: custom-dark
 
-   **Figure 9:**  *The eigenvalues λi, i = 1, . . . , n (green curves using left ordinate) and the energy ratio γm, m = 1, . . . , n (blue curves using right ordinate) are shown for the east and north velocity data. The horizontal dashed lines correspond to the 60% and 90% energy ratio levels, respectively, which equate to utilizing nearly 10 and 100 eigenmodes.*
+   **Figure 9:**  *The eigenvalues* :math:`\lambda_i`, :math:`i = 1, \dots , n` *(green curves using left ordinate) and the energy ratio* :math:`\gamma_m`, :math:`m = 1, \dots , n` *(blue curves using right ordinate) are shown for the east and north velocity data. The horizontal dashed lines correspond to the 60% and 90% energy ratio levels, respectively, which equate to utilizing nearly 10 and 100 eigenmodes.*
 
 Reproducing Figure 10
 ---------------------
@@ -155,7 +155,26 @@ Reproducing Figure 10
  :width: 70%
  :class: custom-dark
 
- **Figure 10:** *The JS distance between the probability distributions pm(x, ξ) and pn(x, ξ) is shown as a function of m = 0, . . . , n. These two distributions correspond to the ensemble generated by the m-term (truncated) and n-term (complete) KL expansions, respectively. We note that the abscissa of the figure is displayed as the percentage of the ratio m/n where n = 485.*
+ **Figure 10:** *The JS distance between the probability distributions* :math:`p_m(\boldsymbol{x}, \xi)` *and* :math:`p_n(\boldsymbol{x}, \xi)` *is shown as a function of* :math:`m = 0, \dots , n`. *These two distributions correspond to the ensemble generated by the* :math:`m` *term (truncated) and* :math:`n` *term (complete) KL expansions, respectively. We note that the abscissa of the figure is displayed as the percentage of the ratio* :math:`m/n` *where* :math:`n = 485`.
+
+Reproducing Figure 11
+---------------------
+
+Run ``plot_vel_distribution`` script:
+
+.. prompt:: bash
+
+    python plot_vel_distribution.py
+
+The above script creates the following plot:
+
+.. figure:: _static/images/plots/vel_distribution.png
+ :align: left
+ :figwidth: 100%
+ :width: 100%
+ :class: custom-dark
+
+ **Figure 11:** *Probability density functions of the east (a) and north (b) components of velocity data for January 2017. Circle points denote the empirical PDF, the solid black curve indicates the standard normal distribution* :math:`\mathcal{N}(0, 1)`, *and the dashed black curve shows the best-fit generalized Gaussian distribution (GGD) with the density function* :math:`p(v) \propto \exp(-\vert v \vert^{\beta})`. *The abscissa in both figures represents the normalized velocity components, with* :math:`\sigma` *being the standard deviation of the respective data.*
 
 .. |script_dir| replace:: ``/examples/uncertainty_quant``
 .. _script_dir: https://github.com/ameli/restoreio/blob/main/examples/uncertainty_quant/

examples/uncertainty_quant/_utils/_load_variables/__init__.py

@@ -8,5 +8,7 @@
 
 
 from ._get_datetime_info import get_datetime_info
+from ._get_fill_value import get_fill_value
+from ._make_array_masked import make_array_masked
 
-__all__ = ['get_datetime_info']
+__all__ = ['get_datetime_info', 'get_fill_value', 'make_array_masked']

examples/uncertainty_quant/_utils/_load_variables/_get_fill_value.py

@@ -0,0 +1,57 @@
+# SPDX-FileCopyrightText: Copyright 2016, Siavash Ameli <sameli@berkeley.edu>
+# SPDX-License-Identifier: BSD-3-Clause
+# SPDX-FileType: SOURCE
+#
+# This program is free software: you can redistribute it and/or modify it under
+# the terms of the license found in the LICENSE.txt file in the root directory
+# of this source tree.
+
+
+# =======
+# Imports
+# =======
+
+import numpy
+
+__all__ = ['get_fill_value']
+
+
+# ==============
+# get fill value
+# ==============
+
+def get_fill_value(east_vel_obj, north_vel_obj):
+    """
+    Finds missing value (or fill value) from wither of east of north velocity
+    objects.
+    """
+
+    # Missing Value
+    if hasattr(east_vel_obj, '_FillValue') and \
+            (not numpy.isnan(float(east_vel_obj._FillValue))):
+        fill_value = numpy.fabs(float(east_vel_obj._FillValue))
+
+    elif hasattr(north_vel_obj, '_FillValue') and \
+            (not numpy.isnan(float(north_vel_obj._FillValue))):
+        fill_value = numpy.fabs(float(north_vel_obj._FillValue))
+
+    elif hasattr(east_vel_obj, 'missing_value') and \
+            (not numpy.isnan(float(east_vel_obj.missing_value))):
+        fill_value = numpy.fabs(float(east_vel_obj.missing_value))
+
+    elif hasattr(north_vel_obj, 'missing_value') and \
+            (not numpy.isnan(float(north_vel_obj.missing_value))):
+        fill_value = numpy.fabs(float(north_vel_obj.missing_value))
+
+    elif hasattr(east_vel_obj, 'fill_value') and \
+            (not numpy.isnan(float(east_vel_obj.fill_value))):
+        fill_value = numpy.fabs(float(east_vel_obj.fill_value))
+
+    elif hasattr(north_vel_obj, 'fill_value') and \
+            (not numpy.isnan(float(north_vel_obj.fill_value))):
+        fill_value = numpy.fabs(float(north_vel_obj.fill_value))
+
+    else:
+        fill_value = 999.0
+
+    return fill_value

examples/uncertainty_quant/_utils/_load_variables/_make_array_masked.py

@@ -0,0 +1,65 @@
+# SPDX-FileCopyrightText: Copyright 2016, Siavash Ameli <sameli@berkeley.edu>
+# SPDX-License-Identifier: BSD-3-Clause
+# SPDX-FileType: SOURCE
+#
+# This program is free software: you can redistribute it and/or modify it under
+# the terms of the license found in the LICENSE.txt file in the root directory
+# of this source tree.
+
+
+# =======
+# Imports
+# =======
+
+import numpy
+
+__all__ = ['make_array_masked']
+
+
+# =================
+# Make Array masked
+# =================
+
+def make_array_masked(array, fill_value):
+    """
+    Often the array is not masked, but has nan or inf values. This function
+    creates a masked array and mask nan and inf.
+
+    Input:
+        - array: is a 2D numpy array.
+    Output:
+        - array: is a 2D numpy.ma array.
+
+    Note: array should be numpy object not netCDF object. So if you have a
+          netCDF object, pass its numpy array with array[:] into this function.
+    """
+
+    if (not hasattr(array, 'mask')) or (numpy.isscalar(array.mask)):
+
+        # Mask based on the fill values
+        mask = (array >= fill_value - 1.0)
+
+        # Mask based on nan values
+        mask_nan = numpy.isnan(array)
+        if mask_nan.any():
+            mask = numpy.logical_or(mask, mask_nan)
+
+        # Mask based on inf values
+        mask_inf = numpy.isinf(array)
+        if mask_inf.any():
+            mask = numpy.logical_or(mask, mask_inf)
+
+        array = numpy.ma.masked_array(array, mask=mask)
+
+    else:
+        # This array is masked. But check if any non-masked value is nan or inf
+        for i in range(array.shape[0]):
+            for j in range(array.shape[1]):
+                for k in range(array.shape[2]):
+                    if bool(array.mask[i, j, k]) is False:
+                        if numpy.isnan(array[i, j, k]) or \
+                                numpy.isinf(array[i, j, k]):
+                            array.mask[i, j, k] = True
+                            array.data[i, j, k] = numpy.nan
+
+    return array

examples/uncertainty_quant/_utils/_plot_utils/__init__.py

@@ -9,7 +9,7 @@
 
 from ._draw_map import draw_map
 from ._plot_utilities import plt, make_axes_locatable, save_plot, \
-        load_plot_settings
+        load_plot_settings, PercentFormatter
 
 __all__ = ['draw_map', 'plt', 'make_axes_locatable', 'save_plot',
-           'load_plot_settings']
+           'load_plot_settings', 'PercentFormatter']