Perserve dtype of source cube with area weighted regridder

lib/iris/experimental/regrid.py

@@ -473,12 +473,17 @@ def _regrid_area_weighted_array(src_data, x_dim, y_dim,
     if y_dim is not None:
         new_shape[y_dim] = grid_y_bounds.shape[0]
 
+    # Use input cube dtype or convert values to the smallest possible float
+    # dtype when necessary.
+    dtype = np.promote_types(src_data.dtype, np.float16)
+
     # Flag to indicate whether the original data was a masked array.
     src_masked = ma.isMaskedArray(src_data)
     if src_masked:
-        new_data = ma.zeros(new_shape, fill_value=src_data.fill_value)
+        new_data = ma.zeros(new_shape, fill_value=src_data.fill_value,
+                            dtype=dtype)
     else:
-        new_data = ma.zeros(new_shape)
+        new_data = ma.zeros(new_shape, dtype=dtype)
     # Assign to mask to explode it, allowing indexed assignment.
     new_data.mask = False
 

lib/iris/tests/experimental/regrid/test_regrid_area_weighted_rectilinear_src_and_grid.py

@@ -490,7 +490,7 @@ def test_global_data_same_res(self):
         src.coord('latitude').guess_bounds()
         src.coord('longitude').guess_bounds()
         res = regrid_area_weighted(src, src)
-        self.assertArrayShapeStats(res, (73, 96), 279.945167, 16.345842)
+        self.assertArrayShapeStats(res, (73, 96), 279.945160, 16.345842)
 
     @tests.skip_data
     def test_global_data_subset(self):
@@ -510,7 +510,7 @@ def test_global_data_subset(self):
         dest.add_dim_coord(dest_lon, 1)
 
         res = regrid_area_weighted(src, dest)
-        self.assertArrayShapeStats(res, (40, 30), 280.979336, 16.640421)
+        self.assertArrayShapeStats(res, (40, 30), 280.979310, 16.640421)
 
     @tests.skip_data
     def test_circular_subset(self):
@@ -530,7 +530,7 @@ def test_circular_subset(self):
         dest.add_dim_coord(dest_lon, 1)
 
         res = regrid_area_weighted(src, dest)
-        self.assertArrayShapeStats(res, (40, 7), 285.653967, 15.212710)
+        self.assertArrayShapeStats(res, (40, 7), 285.653960, 15.212710)
 
     @tests.skip_data
     def test_non_circular_subset(self):
@@ -551,7 +551,7 @@ def test_non_circular_subset(self):
         dest.add_dim_coord(dest_lon, 1)
 
         res = regrid_area_weighted(src, dest)
-        self.assertArrayShapeStats(res, (40, 7), 285.550800, 15.190245)
+        self.assertArrayShapeStats(res, (40, 7), 285.550814, 15.190245)
 
 
 if __name__ == "__main__":

lib/iris/tests/results/experimental/regrid/regrid_area_weighted_rectilinear_src_and_grid/const_lat_cross_section.cml

@@ -1,6 +1,6 @@
 <?xml version="1.0" ?>
 <cubes xmlns="urn:x-iris:cubeml-0.2">
-  <cube dtype="float64" standard_name="air_potential_temperature" units="K">
+  <cube dtype="float32" standard_name="air_potential_temperature" units="K">
     <attributes>
       <attribute name="source" value="Iris test case"/>
     </attributes>
@@ -70,6 +70,6 @@
       </coord>
     </coords>
     <cellMethods/>
-    <data dtype="float64" shape="(5, 18)" state="loaded"/>
+    <data dtype="float32" shape="(5, 18)" state="loaded"/>
   </cube>
 </cubes>

lib/iris/tests/results/experimental/regrid/regrid_area_weighted_rectilinear_src_and_grid/const_lat_cross_section.data.0.json

@@ -1 +1 @@
-{"std": 0.049194271888948204, "min": 287.83400456072997, "max": 288.09036649547346, "shape": [5, 18], "masked": false, "mean": 288.00649837715025}
+{"std": 0.04919414967298508, "min": 287.8340148925781, "max": 288.0903625488281, "shape": [5, 18], "masked": false, "mean": 288.0065002441406}

lib/iris/tests/results/experimental/regrid/regrid_area_weighted_rectilinear_src_and_grid/const_lon_cross_section.cml

@@ -1,6 +1,6 @@
 <?xml version="1.0" ?>
 <cubes xmlns="urn:x-iris:cubeml-0.2">
-  <cube dtype="float64" standard_name="air_potential_temperature" units="K">
+  <cube dtype="float32" standard_name="air_potential_temperature" units="K">
     <attributes>
       <attribute name="source" value="Iris test case"/>
     </attributes>
@@ -64,6 +64,6 @@
       </coord>
     </coords>
     <cellMethods/>
-    <data dtype="float64" shape="(5, 12)" state="loaded"/>
+    <data dtype="float32" shape="(5, 12)" state="loaded"/>
   </cube>
 </cubes>

lib/iris/tests/results/experimental/regrid/regrid_area_weighted_rectilinear_src_and_grid/const_lon_cross_section.data.0.json

@@ -1 +1 @@
-{"std": 0.053120598663116025, "min": 287.90810573168818, "max": 288.14679568802029, "shape": [5, 12], "masked": false, "mean": 288.02641852126851}
+{"std": 0.05311942473053932, "min": 287.9081115722656, "max": 288.14678955078125, "shape": [5, 12], "masked": false, "mean": 288.02642822265625}

lib/iris/tests/results/experimental/regrid/regrid_area_weighted_rectilinear_src_and_grid/hybridheight.cml

@@ -1,6 +1,6 @@
 <?xml version="1.0" ?>
 <cubes xmlns="urn:x-iris:cubeml-0.2">
-  <cube dtype="float64" standard_name="air_potential_temperature" units="K">
+  <cube dtype="float32" standard_name="air_potential_temperature" units="K">
     <attributes>
       <attribute name="source" value="Iris test case"/>
     </attributes>
@@ -433,6 +433,6 @@
       </coord>
     </coords>
     <cellMethods/>
-    <data dtype="float64" shape="(2, 5, 16, 21)" state="loaded"/>
+    <data dtype="float32" shape="(2, 5, 16, 21)" state="loaded"/>
   </cube>
 </cubes>

lib/iris/tests/results/experimental/regrid/regrid_area_weighted_rectilinear_src_and_grid/hybridheight.data.0.json

@@ -1 +1 @@
-{"std": 0.10611990310218537, "min": 287.59569757163996, "max": 288.16508077725081, "shape": [2, 5, 16, 21], "masked": false, "mean": 287.91804930402083}
+{"std": 0.10611984878778458, "min": 287.595703125, "max": 288.1650695800781, "shape": [2, 5, 16, 21], "masked": false, "mean": 287.91802978515625}