use new apip for adaptive localization

src/ert/analysis/_es_update.py

@@ -550,11 +550,11 @@ def analysis_ES(
                 for param_batch_idx in TimedIterator(batches, progress_callback):
                     X_local = temp_storage[param_group.name][param_batch_idx, :]
                     D = smoother_adaptive_es.perturb_observations(
-                        size=S.shape, alpha=1.0
+                        ensemble_size=ensemble_size, alpha=1.0
                     )
                     temp_storage[param_group.name][
                         param_batch_idx, :
-                    ] = smoother_adaptive_es.adaptive_assimilate(
+                    ] = smoother_adaptive_es.assimilate(
                         X=X_local,
                         Y=S,
                         D=D,
@@ -612,10 +612,14 @@ def analysis_IES(
     ens_mask: npt.NDArray[np.bool_],
     source_fs: EnsembleReader,
     target_fs: EnsembleAccessor,
-    iterative_ensemble_smoother: ies.SIES,
+    iterative_ensemble_smoother: Optional[ies.SIES],
     progress_callback: Callable[[AnalysisEvent], None],
     misfit_process: bool,
 ) -> None:
+    # perhaps we should return iterative_ensemble_smoother: ies.SIES object?
+    # if it is not passed, create it and return
+    # if it is passed, update with new iteration and return
+
     iens_active_index = np.flatnonzero(ens_mask)
 
     tot_num_params = sum(
@@ -624,6 +628,8 @@ def analysis_IES(
     )
     param_groups = list(source_fs.experiment.parameter_configuration.keys())
     ensemble_size = ens_mask.sum()
+
+    # modify this to always return all parameters
     param_ensemble = _param_ensemble_for_projection(
         source_fs, iens_active_index, ensemble_size, param_groups, tot_num_params
     )
@@ -655,6 +661,20 @@ def analysis_IES(
             )
 
         noise = rng.standard_normal(size=(len(observation_values), S.shape[1]))
+
+        if iterative_ensemble_smoother is not None:
+            iterative_ensemble_smoother = ies.SIES(
+                parameters=param_ensemble,
+                covariance=observation_errors**2,
+                observations=observation_values,
+                seed=rng,
+            )
+
+        iterative_ensemble_smoother.sies_iteration(
+            S, step_length,
+        )
+
+        # where is step-length?? perhaps on init outside?
         iterative_ensemble_smoother.fit(
             S,
             observation_errors,