diff --git a/trough/_arb.py b/trough/_arb.py
index 7553bb5..a0454d7 100644
--- a/trough/_arb.py
+++ b/trough/_arb.py
@@ -36,7 +36,7 @@ def get_arb_paths(start_date, end_date, hemisphere, processed_dir):
 def get_arb_data(start_date, end_date, hemisphere, processed_dir=None):
     if processed_dir is None:
         processed_dir = config.processed_arb_dir
-    data = xr.concat([xr.open_dataarray(file) for file in get_arb_paths(start_date, end_date, hemisphere, processed_dir)], 'time')
+    data = utils.read_netcdfs(get_arb_paths(start_date, end_date, hemisphere, processed_dir), 'time')
     return data.sel(time=slice(start_date, end_date))
 
 
diff --git a/trough/_tec.py b/trough/_tec.py
index b04134c..64c94b7 100644
--- a/trough/_tec.py
+++ b/trough/_tec.py
@@ -73,7 +73,7 @@ def get_tec_paths(start_date, end_date, hemisphere, processed_dir):
 def get_tec_data(start_date, end_date, hemisphere, processed_dir=None):
     if processed_dir is None:
         processed_dir = config.processed_tec_dir
-    data = xr.concat([xr.open_dataarray(file) for file in get_tec_paths(start_date, end_date, hemisphere, processed_dir)], 'time')
+    data = utils.read_netcdfs(get_tec_paths(start_date, end_date, hemisphere, processed_dir), 'time')
     return data.sel(time=slice(start_date, end_date))
 
 
diff --git a/trough/_trough.py b/trough/_trough.py
index f9530d5..71c0245 100644
--- a/trough/_trough.py
+++ b/trough/_trough.py
@@ -33,11 +33,11 @@ def get_model(tec_data, hemisphere, omni_file):
     logger.info(f"{kp.shape=}")
     apex = Apex(date=utils.datetime64_to_datetime(tec_data.time.values[0]))
     mlat = 65.5 * np.ones((tec_data.time.shape[0], tec_data.mlt.shape[0]))
-    if hemisphere == 'south':
-        mlat = mlat * -1
     for i in range(10):
         glat, glon = apex.convert(mlat, tec_data.mlt.values[None, :], 'mlt', 'geo', 350, tec_data.time.values[:, None])
         mlat = _model_subroutine_lat(tec_data.mlt.values[None, :], glon, kp[:, None], hemisphere)
+    if hemisphere == 'south':
+        mlat = mlat * -1
     tec_data['model'] = xr.DataArray(
         mlat,
         coords={'time': tec_data.time, 'mlt': tec_data.mlt},
@@ -311,7 +311,7 @@ def get_label_paths(start_date, end_date, hemisphere, processed_dir):
 def get_trough_labels(start_date, end_date, hemisphere, labels_dir=None):
     if labels_dir is None:
         labels_dir = config.processed_labels_dir
-    data = xr.concat([xr.open_dataarray(file) for file in get_label_paths(start_date, end_date, hemisphere, labels_dir)], 'time')
+    data = utils.read_netcdfs(get_label_paths(start_date, end_date, hemisphere, labels_dir), 'time')
     return data.sel(time=slice(start_date, end_date))
 
 
diff --git a/trough/utils.py b/trough/utils.py
index d9a5040..e43a035 100644
--- a/trough/utils.py
+++ b/trough/utils.py
@@ -2,6 +2,7 @@
 import datetime
 import warnings
 import logging
+import xarray as xr
 try:
     import h5py
     from skimage.util import view_as_windows
@@ -153,3 +154,20 @@ def check(start, end, dt, hemisphere, processed_file):
         return True
 
     return check
+
+
+def read_netcdfs(files, dim):
+    """https://xarray.pydata.org/en/stable/user-guide/io.html#reading-multi-file-datasets
+    """
+    def process_one_path(path):
+        # use a context manager, to ensure the file gets closed after use
+        with xr.open_dataarray(path) as ds:
+            # load all data from the transformed dataset, to ensure we can
+            # use it after closing each original file
+            ds.load()
+            return ds
+
+    paths = sorted(files)
+    datasets = [process_one_path(p) for p in paths]
+    combined = xr.concat(datasets, dim)
+    return combined