Copy template and rename for threshold identification

enstools/feature/identification/threshold/__init__.py

+from .identification import  MultiThresholdIdentification

enstools/feature/identification/threshold/identification.py

+from ..identification import IdentificationTechnique
+
+import xarray as xr
+
+class MultiThresholdIdentification(IdentificationTechnique):
+
+    def __init__(self, some_parameter='', **kwargs):
+        # Constructor. Called from example_template.py, parameters can be passed and set here.
+        self.param = some_parameter
+
+        # Specify processing mode 2d or 3d: In 2d, identification will be performed on 2d levels,
+        #   in 3d the identification will be performed per 3d block
+        self.processing_mode = '2d'
+
+        pass
+
+
+    def precompute(self, dataset: xr.Dataset, **kwargs):
+        # Precomputation, done once for the entire dataset.
+        # All work which requires or takes the entire input can be performed here, identify() later is called in
+        #   parallel for all timesteps, ensemble members, ...
+
+        # Also data fields need to be added here in case they are needed and populated later on.
+        # e.g. dataset['identified_areas'] = xr.zeros_like(dataset['existing_field'], dtype=bool)
+        return dataset
+
+    def identify(self, dataset: xr.Dataset, **kwargs):
+        # Identification, here is the place for your identification algorithm.
+        # This is called in parallel for each spatial subset of the data, e.g., for each member and for each timestep
+        # Therefore, dataset contains only spatial dimensions (lat,lon) if self.processing_mode = '2d' is set,
+        #   (see __init__), or (lat,lon,level) if '3d' is set.
+
+        # here your code ... identify objects for this subset.
+
+        # Let's say you detected 5 objects:
+        obj_list = []
+
+        for i in range(5):
+            # get an instance of a new object, can pass an ID or set in manually afterwards
+            obj = self.get_new_object()
+            # set some ID to it
+            obj.id = i + 1
+
+            # get properties of object and populate them (like defined in template.proto)
+            properties = obj.properties
+
+            properties.size = 42
+            properties.centroid.x = i
+            properties.centroid.y = 2*i
+            properties.centroid.z = 3*i
+
+            properties.list_of_something.append("type x")
+            properties.list_of_something.append("type z")
+
+            obj_list.append(obj)
+
+        # return the dataset (can be changed here), and the list of objects
+        return dataset, obj_list
+
+    def postprocess(self, dataset: xr.Dataset, pb2_desc, **kwargs):
+        # Here post process: called once after processing the whole dataset.
+        # If needed, the dataset or objects can be changed here.
+
+        # objects are already in finished protobuf description, cf. the JSON output.
+        # pb2_desc.sets is list of trackable sets, each one contains the timesteps list, where each entry is a valid time and contains a list of objects.
+        return dataset, pb2_desc

enstools/feature/identification/threshold/threshold.proto

+syntax = "proto2";
+
+/* This file contains the feature description for an identified object.
+ * The "Properties" message is required and contains all the properties of an object.
+ */
+
+message MaskArray {
+    repeated int32 shape = 1;
+    required bytes data = 2;
+}
+
+message Properties {
+    required MaskArray masks = 1;
+    repeated float thresholds = 2;
+}
+