diff --git a/src/xray/backends.py b/src/xray/backends.py
index d6594e5462e..4b849403c9f 100644
--- a/src/xray/backends.py
+++ b/src/xray/backends.py
@@ -151,7 +151,23 @@ def convert_variable(var):
                 data = np.asscalar(var[...])
             attributes = OrderedDict((k, var.getncattr(k))
                                      for k in var.ncattrs())
-            return xarray.XArray(dimensions, data, attributes,
+            # netCDF4 specific encoding; save _FillValue for later
+            encoding = {}
+            filters = var.filters()
+            if filters is not None:
+                encoding.update(filters)
+            chunking = var.chunking()
+            if chunking is not None:
+                if chunking == 'contiguous':
+                    encoding['contiguous'] = True
+                    encoding['chunksizes'] = None
+                else:
+                    encoding['contiguous'] = False
+                    encoding['chunksizes'] = tuple(chunking)
+            encoding['endian'] = var.endian()
+            encoding['least_significant_digit'] = \
+                attributes.pop('least_significant_digit', None)
+            return xarray.XArray(dimensions, data, attributes, encoding,
                                  indexing_mode='orthogonal')
         return FrozenOrderedDict((k, convert_variable(v))
                                  for k, v in self.ds.variables.iteritems())
@@ -180,10 +196,20 @@ def set_variable(self, name, variable):
         # we let the package handle the _FillValue attribute
         # instead of setting it ourselves.
         fill_value = variable.attributes.pop('_FillValue', None)
-        self.ds.createVariable(varname=name,
-                               datatype=variable.dtype,
-                               dimensions=variable.dimensions,
-                               fill_value=fill_value)
+        encoding = variable.encoding
+        self.ds.createVariable(
+            varname=name,
+            datatype=variable.dtype,
+            dimensions=variable.dimensions,
+            zlib=encoding.get('zlib', False),
+            complevel=encoding.get('complevel', 4),
+            shuffle=encoding.get('shuffle', True),
+            fletcher32=encoding.get('fletcher32', False),
+            contiguous=encoding.get('contiguous', False),
+            chunksizes=encoding.get('chunksizes'),
+            endian=encoding.get('endian', 'native'),
+            least_significant_digit=encoding.get('least_significant_digit'),
+            fill_value=fill_value)
         nc4_var = self.ds.variables[name]
         nc4_var.set_auto_maskandscale(False)
         nc4_var[:] = variable.data[:]
diff --git a/test/test_dataset.py b/test/test_dataset.py
index 0ba6dcf6e87..1dde96cab0e 100644
--- a/test/test_dataset.py
+++ b/test/test_dataset.py
@@ -508,7 +508,9 @@ def test_open_encodings(self):
         actual = open_dataset(tmp_file)
 
         self.assertXArrayEqual(actual['time'], expected['time'])
-        self.assertDictEqual(actual['time'].encoding, expected['time'].encoding)
+        actual_encoding = {k: v for k, v in actual['time'].encoding.iteritems()
+                           if k in expected['time'].encoding}
+        self.assertDictEqual(actual_encoding, expected['time'].encoding)
 
         os.remove(tmp_file)
 
@@ -539,6 +541,15 @@ def test_dump_and_open_encodings(self):
         ds.close()
         os.remove(tmp_file)
 
+    def test_compression_encoding(self):
+        data = create_test_data()
+        data['var2'].encoding.update({'zlib': True,
+                                      'chunksizes': (10, 10),
+                                      'least_significant_digit': 2})
+        actual = self.roundtrip(data)
+        for k, v in data['var2'].encoding.iteritems():
+            self.assertEqual(v, actual['var2'].encoding[k])
+
     def test_mask_and_scale(self):
         f, tmp_file = tempfile.mkstemp(suffix='.nc')
         os.close(f)