diff --git a/src/scores/probability/crps_impl.py b/src/scores/probability/crps_impl.py
index b6309fc6..24aa450e 100644
--- a/src/scores/probability/crps_impl.py
+++ b/src/scores/probability/crps_impl.py
@@ -50,7 +50,7 @@ def check_crps_cdf_inputs(
         raise ValueError("Dimensions of `threshold_weight` must be a subset of dimensions of `fcst`")
 
     if dims is not None and not set(dims).issubset(fcst.dims):
-        raise ValueError("`dims` must be a subset of `fcst` dimensions")
+        raise ValueError("`dims` must be a subset of `fcst` dimensions")  # pragma: no cover
 
     if fcst_fill_method not in ["linear", "step", "forward", "backward"]:
         raise ValueError("`fcst_fill_method` must be 'linear', 'step', 'forward' or 'backward'")
@@ -289,9 +289,6 @@ def crps_cdf(
             Journal of Business & Economic Statistics, 29(3), 411–422. http://www.jstor.org/stable/23243806
     """
 
-    if weights is not None:
-        raise NotImplementedError("Weights handling is coming soon")
-
     dims = scores.utils.gather_dimensions(
         fcst.dims,
         obs.dims,
@@ -344,8 +341,10 @@ def crps_cdf(
             include_components=include_components,
         )
 
+    weighted = scores.functions.apply_weights(result, weights)        
+
     dims_to_collapse = scores.utils.dims_complement(result, dims=dims)
-    result = result.mean(dim=dims_to_collapse)
+    result = weighted.mean(dim=dims_to_collapse)
 
     return result
 
@@ -540,7 +539,7 @@ def check_crps_cdf_brier_inputs(fcst, obs, threshold_dim, fcst_fill_method, dims
         raise ValueError("Dimensions of `obs` must be a subset of dimensions of `fcst`")
 
     if dims is not None and not set(dims).issubset(fcst.dims):
-        raise ValueError("`dims` must be a subset of `fcst` dimensions")
+        raise ValueError("`dims` must be a subset of `fcst` dimensions")  # pragma: no cover
 
     if fcst_fill_method not in ["linear", "step", "forward", "backward"]:
         raise ValueError("`fcst_fill_method` must be 'linear', 'step', 'forward' or 'backward'")
diff --git a/src/scores/sample_data.py b/src/scores/sample_data.py
index 6b3e40bf..9e56adfd 100644
--- a/src/scores/sample_data.py
+++ b/src/scores/sample_data.py
@@ -34,9 +34,9 @@ def continuous_observations(large_size: bool = False) -> xr.DataArray:
     periods = 10
 
     if large_size:  # pragma: no cover
-        num_lats = 364  # pragma: no cover
-        num_lons = 720  # pragma: no cover
-        periods = 240  # pragma: no cover
+        num_lats = 364  # pragma: no cover - used in notebooks and tested manually
+        num_lons = 720  # pragma: no cover - used in notebooks and tested manually
+        periods = 240  # pragma: no cover - used in notebooks and tested manually
 
     lat = np.linspace(-90, 90, num_lats)
     lon = np.linspace(0, 360, num_lons)
diff --git a/src/scores/utils.py b/src/scores/utils.py
index d28a54d5..eb05c0f5 100644
--- a/src/scores/utils.py
+++ b/src/scores/utils.py
@@ -208,12 +208,3 @@ def check_dims(xr_data, expected_dims, mode=None):
                 )
 
 
-def create_latitude_weights(latitudes):
-    '''
-    A common way of weighting errors is to make them proportional to the amount of area
-    which is contained in a particular region. This is approximated by the cosine
-    of the latitude on an LLXY grid. Nuances not accounted for include the variation in
-    latitude across the region, or the irregularity of the surface of the earth.
-    '''
-    weights = np.cos(np.deg2rad(latitudes))
-    return weights
\ No newline at end of file