Revisit the notation

The subscripts are awkward, but I think this is consistent.

docs/reference/configuration/keywords.rst

@@ -1378,14 +1378,14 @@ Keywords controlling the ES algorithm
         Including outliers in the Smoother algorithm can dramatically increase the
         coupling between the ensemble members. It is therefore important to filter out
         these outliers prior to data assimilation. An observation,
-        :math:`\mathbf{d}^o_i`, will be classified as an outlier if
+        :math:`d^o_i`, will be classified as an outlier if
 
-        :math:`|\mathbf{d}^o_i - \bar{\mathbf{d}}_i| > \mathrm{ENKF\_ALPHA} \left(\mathbf{s}_{\mathbf{d}_i} + \mathbf{s}^o_{\mathbf{d}_i}\right)`
+        :math:`|d^o_i - \mathbf{\bar{d}}| > \mathrm{ENKF\_ALPHA} \left( s_{d_i} + s^o_{d_i} \right)`
 
         where :math:`\mathbf{d}^o` is the vector of observed data,
         :math:`\mathbf{\bar{d}}` is the average of the forecasted data ensemble,
         :math:`\mathbf{s}_\mathbf{d}` is the vector of estimated standard deviations
-        for the forecasted data ensemble, and :math:`\mathbf{s}^o_{\mathbf{d}_i}` is the
+        for the forecasted data ensemble, and :math:`\mathbf{s}^o_\mathbf{d}` is the
         vector of standard deviations for the observation error (specified a priori).
 
         Observe that for the updates many settings should be applied on the analysis