diff --git a/ods_tools/odtf/mapping/base.py b/ods_tools/odtf/mapping/base.py
index a680304..6d7ab97 100644
--- a/ods_tools/odtf/mapping/base.py
+++ b/ods_tools/odtf/mapping/base.py
@@ -278,7 +278,7 @@ def get_transformations(self, available_columns: List[str]) -> List[DirectionalM
                 for transform in transform_list:
                     transform.parse()
                     if ((transform.transformation_tree is None or
-                        not self.has_missing_columns(transform.transformation_tree, missing_columns)) and
+                         not self.has_missing_columns(transform.transformation_tree, missing_columns)) and
                         (transform.when_tree is None or
                          not self.has_missing_columns(transform.when_tree, missing_columns))):
                         valid_transforms.append(transform)
diff --git a/ods_tools/oed/source.py b/ods_tools/oed/source.py
index ac69329..d63db4d 100644
--- a/ods_tools/oed/source.py
+++ b/ods_tools/oed/source.py
@@ -283,8 +283,8 @@ def as_oed_type(cls, oed_df, column_to_field):
                     oed_df[column] = oed_df[column].cat.add_categories('')
                 oed_df[column] = oed_df[column]  # make a copy f the col in case it is read_only
                 oed_df.loc[is_empty(oed_df, column), column] = ''
-            elif pd_dtype[column].startswith('Int'):
-                to_tmp_dtype[column] = 'float'
+            if pd.api.types.is_numeric_dtype(pd_dtype[column]):  # make sure empty string are converted to nan
+                oed_df[column] = pd.to_numeric(oed_df[column], errors='coerce')
 
         return oed_df.astype(to_tmp_dtype).astype(pd_dtype)
 
@@ -302,7 +302,7 @@ def prepare_df(cls, df, column_to_field, ods_fields):
         """
         # set default values
         for col, field_info in column_to_field.items():
-            fill_empty(df, col, OedSchema.get_default_from_ods_fields(ods_fields, col))
+            fill_empty(df, col, OedSchema.get_default_from_ods_fields(ods_fields, field_info['Input Field Name']))
 
         # add required columns that allow blank values if missing
         present_field = set(field_info['Input Field Name'] for field_info in column_to_field.values())
diff --git a/tests/test_ods_package.py b/tests/test_ods_package.py
index d0988c1..c56b2dd 100644
--- a/tests/test_ods_package.py
+++ b/tests/test_ods_package.py
@@ -430,6 +430,59 @@ def test_field_required_allow_blank_are_set_to_default(self):
             assert (modified_exposure.location.dataframe['BITIV'] == 0).all()  # check default is applied
             assert (modified_exposure.ri_info.dataframe['RiskLevel'] == '').all()  # check it works for string
 
+    def test_fill_empty(self):
+        oed_schema = OedSchema.from_oed_schema_info(None)
+        test_fields = {
+            'intvalue': {
+                'Input Field Name': 'IntValue',
+                'Type & Description': 'a single int column with default',
+                'Required Field': 'O',
+                'Data Type': 'int',
+                'Allow blanks?': 'YES',
+                'Default': '0',
+                'Valid value range': 'n/a',
+                'pd_dtype': 'Int32'},
+            'IntValueMultipleXX': {
+                'Input Field Name': 'IntValueMultipleXX',
+                'Type & Description': '',
+                'Required Field': 'O',
+                'Data Type': 'int',
+                'Allow blanks?': 'YES',
+                'Default': '0',
+                'Valid value range': 'n/a',
+                'pd_dtype': 'Int32'},
+            'StringValueMultipleXX': {
+                'Input Field Name': 'StringValueMultipleXX',
+                'Type & Description': '',
+                'Required Field': 'O',
+                'Data Type': 'nvarchar(50)',
+                'Allow blanks?': 'YES',
+                'Default': 'foobar',
+                'Valid value range': 'n/a',
+                'pd_dtype': 'category'}
+        }
+
+        for field_name, field_info in test_fields.items():
+            oed_schema.schema['input_fields']['Loc'][field_name.lower()] = field_info
+
+        loc_df = pd.DataFrame({
+            'PortNumber': [1, 1, 1, 1],
+            'AccNumber': [1, 1, 1, 1],
+            'LocNumber': [1, 2, 3, 4],
+            'CountryCode': ['UK', 'UK', 'UK', 'UK', ],
+            'LocPerilsCovered': ['WW2', 'WTC;WSS', 'QQ1;WW2', 'WTC'],
+            'BuildingTIV': ['1', '1', '1', '1'],
+            'ContentsTIV': ['1', '1', '1', '1'],
+            'LocCurrency': ['GBP', 'GBP', 'GBP', 'GBP'],
+            'intvalue': [1, '2', '', ''],
+            'IntValueMultiple1': [1, '2', '', ''],
+            'StringValueMultiple01': [1, '2', '', ''],
+        })
+        oed = OedExposure(**{'location': loc_df, 'use_field': True, 'oed_schema_info': oed_schema})
+        assert oed.location.dataframe['IntValue'].to_list() == [1, 2, 0, 0]
+        assert oed.location.dataframe['IntValueMultiple1'].to_list() == [1, 2, 0, 0]
+        assert oed.location.dataframe['StringValueMultiple01'].to_list() == ['1', '2', 'foobar', 'foobar']
+
     def test_relative_and_absolute_path(self):
         original_cwd = os.getcwd()
         try:
@@ -629,7 +682,7 @@ def test_peril_filtering(self):
             'PortNumber': [1, 1, 1, 1],
             'AccNumber': [1, 1, 1, 1],
             'LocNumber': [1, 2, 3, 4],
-            'CountryCode': ['UK', 'UK', 'UK', 'UK',],
+            'CountryCode': ['UK', 'UK', 'UK', 'UK', ],
             'LocPerilsCovered': ['WW2', 'WTC;WSS', 'QQ1;WW2', 'WTC'],
             'BuildingTIV': ['1', '1', '1', '1'],
             'ContentsTIV': ['1', '1', '1', '1'],