Add max, min and std operators to multimodel

doc/recipe/preprocessor.rst

@@ -635,8 +635,8 @@ to observational data, these biases have a significantly lower statistical
 impact when using a multi-model ensemble. ESMValTool has the capability of
 computing a number of multi-model statistical measures: using the preprocessor
 module ``multi_model_statistics`` will enable the user to ask for either a
-multi-model ``mean`` and/or ``median`` with a set of argument parameters passed
-to ``multi_model_statistics``.
+multi-model ``mean``, ``median``, ```max``, ``min`` and / or ``std`` with a set
+of argument parameters passed to ``multi_model_statistics``.
 
 Multimodel statistics in ESMValTool are computed along the time axis, and as
 such, can be computed across a common overlap in time (by specifying ``span:

esmvalcore/preprocessor/_multimodel.py

@@ -62,6 +62,12 @@ def _compute_statistic(data, statistic_name):
         statistic_function = np.ma.median
     elif statistic_name == 'mean':
         statistic_function = np.ma.mean
+    elif statistic_name == 'std':
+        statistic_function = np.ma.std
+    elif statistic_name == 'max':
+        statistic_function = np.ma.max
+    elif statistic_name == 'min':
+        statistic_function = np.ma.min
     else:
         raise NotImplementedError
 
@@ -326,7 +332,8 @@ def multi_model_statistics(products, span, output_products, statistics):
     output_products: dict
         dictionary of output products.
     statistics: str
-        statistical measure to be computed (mean or median).
+        statistical measure to be computed. Available options: mean, median,
+        max, min, std
     Returns
     -------
     list

tests/unit/preprocessor/_multimodel/test_multimodel.py

@@ -90,6 +90,29 @@ def test_compute_statistic(self):
         self.assert_array_equal(stat_mean, expected_mean)
         self.assert_array_equal(stat_median, expected_median)
 
+    def test_compute_std(self):
+        """Test statistic."""
+        data = [self.cube1.data[0], self.cube2.data[0] * 2]
+        stat = _compute_statistic(data, "std")
+        expected = np.ma.ones((3, 2, 2)) * 0.5
+        expected[0, 0, 0] = 0
+        self.assert_array_equal(stat, expected)
+
+    def test_compute_max(self):
+        """Test statistic."""
+        data = [self.cube1.data[0] * 0.5, self.cube2.data[0] * 2]
+        stat = _compute_statistic(data, "max")
+        expected = np.ma.ones((3, 2, 2)) * 2
+        expected[0, 0, 0] = 0.5
+        self.assert_array_equal(stat, expected)
+
+    def test_compute_min(self):
+        """Test statistic."""
+        data = [self.cube1.data[0] * 0.5, self.cube2.data[0] * 2]
+        stat = _compute_statistic(data, "min")
+        expected = np.ma.ones((3, 2, 2)) * 0.5
+        self.assert_array_equal(stat, expected)
+
     def test_put_in_cube(self):
         """Test put in cube."""
         cube_data = np.ma.ones((2, 3, 2, 2))