diff --git a/opensensor/collection_apis.py b/opensensor/collection_apis.py
index 1c4c85e..5e56042 100644
--- a/opensensor/collection_apis.py
+++ b/opensensor/collection_apis.py
@@ -11,6 +11,7 @@
 from opensensor.collections import (
     CO2,
     PH,
+    VPD,
     DeviceMetadata,
     Environment,
     Humidity,
@@ -140,20 +141,6 @@ def get_collection_name(response_model: Type[T]):
     return response_model.__name__
 
 
-def _get_old_project_projection(response_model: Type[T]):
-    project_projection = {
-        "_id": False,
-    }
-    for field_name, _ in response_model.__fields__.items():
-        if field_name == "timestamp":
-            project_projection["timestamp"] = "$timestamp"
-        if field_name == "unit":
-            project_projection["unit"] = "$metadata.unit"
-        else:
-            project_projection[field_name] = f"${field_name}"
-    return project_projection
-
-
 def _get_project_projection(response_model: Type[T]):
     old_name = get_collection_name(response_model)
     new_collection_name = new_collections[old_name]
@@ -170,37 +157,121 @@ def _get_project_projection(response_model: Type[T]):
     return project_projection
 
 
-def get_uniform_sample_pipeline(
-    response_model: Type[T],
-    device_ids: List[str],  # Update the type of the device_id parameter to List[str]
+def get_initial_match_clause(
+    device_ids: List[str],
     device_name: str,
     start_date: datetime,
     end_date: datetime,
     resolution: int,
-    old_collection: bool,
 ):
-    sampling_interval = timedelta(minutes=resolution)
     if start_date is None:
         start_date = datetime.utcnow() - timedelta(days=100)
     if end_date is None:
         end_date = datetime.utcnow()
 
+    # Defining the match clause for the pipeline
     match_clause = {
         "timestamp": {"$gte": start_date, "$lte": end_date},
         "metadata.device_id": {
-            "$in": device_ids
-        },  # Use $in operator for matching any device_id in the list
+            "$in": device_ids  # Use $in operator for matching any device_id in the list
+        },
         "metadata.name": device_name,
     }
+    return match_clause
+
+
+def get_vpd_pipeline(
+    device_ids: List[str],
+    device_name: str,
+    start_date: datetime,
+    end_date: datetime,
+    resolution: int,
+):
+    sampling_interval = timedelta(minutes=resolution)
+    match_clause = get_initial_match_clause(
+        device_ids, device_name, start_date, end_date, resolution
+    )
+
+    # We ensure both temperature and humidity exist for the calculation of VPD
+    match_clause["temp"] = {"$exists": True}
+    match_clause["rh"] = {"$exists": True}
+
+    # The MongoDB aggregation pipeline for VPD calculation
+    pipeline = [
+        {"$match": match_clause},
+        {
+            "$addFields": {
+                "group": {
+                    "$floor": {
+                        "$divide": [
+                            {"$subtract": ["$timestamp", start_date]},
+                            sampling_interval.total_seconds() * 1000,
+                        ]
+                    }
+                }
+            }
+        },
+        {
+            "$group": {
+                "_id": "$group",
+                "temp": {"$avg": "$temp"},
+                "rh": {"$avg": "$rh"},
+                "timestamp": {"$first": "$timestamp"},
+            }
+        },
+        {
+            "$addFields": {
+                "satvp": {
+                    "$multiply": [
+                        0.61078,
+                        {
+                            "$exp": {
+                                "$multiply": [
+                                    {"$divide": [17.27, {"$add": ["$temp", 237.3]}]},
+                                    "$temp",
+                                ]
+                            }
+                        },
+                    ]
+                }
+            }
+        },
+        {
+            "$addFields": {
+                "vpd": {"$multiply": ["$satvp", {"$subtract": [1, {"$divide": ["$rh", 100]}]}]}
+            }
+        },
+        {
+            "$project": {
+                "_id": False,
+                "timestamp": 1,
+                "vpd": 1,
+                "unit": "kPa",  # assuming kPa as the unit for VPD
+            }
+        },
+        {"$sort": {"timestamp": 1}},
+    ]
+    return pipeline
+
+
+def get_uniform_sample_pipeline(
+    response_model: Type[T],
+    device_ids: List[str],  # Update the type of the device_id parameter to List[str]
+    device_name: str,
+    start_date: datetime,
+    end_date: datetime,
+    resolution: int,
+):
+    sampling_interval = timedelta(minutes=resolution)
+    match_clause = get_initial_match_clause(
+        device_ids, device_name, start_date, end_date, resolution
+    )
 
     # Determine the $project
-    if old_collection:
-        project_projection = _get_old_project_projection(response_model)
-    else:
-        old_name = get_collection_name(response_model)
-        new_collection_name = new_collections[old_name]
-        project_projection = _get_project_projection(response_model)
-        match_clause[new_collection_name] = {"$exists": True}
+    old_name = get_collection_name(response_model)
+    new_collection_name = new_collections[old_name]
+    project_projection = _get_project_projection(response_model)
+    match_clause[new_collection_name] = {"$exists": True}
 
     # Query a uniform sample of documents within the timestamp range
     pipeline = [
@@ -234,6 +305,7 @@ def get_uniform_sample_pipeline(
     "co2": CO2,
     "readings": Moisture,
     "pH": PH,
+    "VPD": VPD,
 }
 model_class_attributes = {v: k for k, v in model_classes.items()}
 
@@ -248,38 +320,57 @@ def sample_and_paginate_collection(
     page: int,
     size: int,
     unit: str,
-    old_collection: bool,
 ):
     api_keys, _ = get_api_keys_by_device_id(device_id)
     device_ids, target_device_name = reduce_api_keys_to_device_ids(api_keys, device_id)
     offset = (page - 1) * size
-    pipeline = get_uniform_sample_pipeline(
-        response_model,
-        device_ids,
-        target_device_name,
-        start_date,
-        end_date,
-        resolution,
-        old_collection,
-    )
+
+    # Determine the right pipeline to use based on the response model
+    if response_model is VPD:
+        pipeline = get_vpd_pipeline(
+            device_ids,
+            target_device_name,
+            start_date,
+            end_date,
+            resolution,
+        )
+    else:
+        pipeline = get_uniform_sample_pipeline(
+            response_model,
+            device_ids,
+            target_device_name,
+            start_date,
+            end_date,
+            resolution,
+        )
+
     pipeline.extend([{"$skip": offset}, {"$limit": size}])
 
     db = get_open_sensor_db()
     collection = db[collection_name]
     raw_data = list(collection.aggregate(pipeline))
+
     # Add UTC offset to timestamp field
     for item in raw_data:
         item["timestamp"] = item["timestamp"].replace(tzinfo=timezone.utc).isoformat()
-    data_field = model_class_attributes[response_model]
-    model_class = model_classes[data_field]
-    data = [model_class(**item) for item in raw_data]
-    if data_field == "temp" and unit:
-        for item in data:
-            convert_temperature(item, unit)
+
+    if response_model is VPD:
+        # If the response model is VPD, you already have VPD-related data from the pipeline.
+        # So, you can directly use it to create the response model instances.
+        data = [VPD(**item) for item in raw_data]
+    else:
+        data_field = model_class_attributes[response_model]
+        model_class = model_classes[data_field]
+        data = [model_class(**item) for item in raw_data]
+        if data_field == "temp" and unit:
+            for item in data:
+                convert_temperature(item, unit)
+
     # Re-run for total page count
     pipeline.append({"$count": "total"})
     data_count = list(collection.aggregate(pipeline))
     total_count = data_count[0]["total"] if data else 0
+
     return Page(items=data, total=total_count, page=page, size=size)
 
 
@@ -308,9 +399,6 @@ async def historical_data_route(
             collection_name = user.collection_name
         else:
             collection_name = "FreeTier"
-        migration_finished = migration_complete(collection_name)
-        if not migration_finished:
-            collection_name = get_collection_name(entity)
 
         return sample_and_paginate_collection(
             entity,
@@ -322,7 +410,6 @@ async def historical_data_route(
             page=page,
             size=size,
             unit=unit,
-            old_collection=not migration_finished,
         )
 
     return historical_data_route
@@ -355,6 +442,12 @@ async def historical_data_route(
     response_model=Page[PH],
     methods=["GET"],
 )
+router.add_api_route(
+    "/VPD/{device_id}",
+    create_historical_data_route(VPD),
+    response_model=Page[VPD],
+    methods=["GET"],
+)
 
 
 @router.post("/environment/")
diff --git a/opensensor/collections.py b/opensensor/collections.py
index a2d16ec..0fad60d 100644
--- a/opensensor/collections.py
+++ b/opensensor/collections.py
@@ -66,3 +66,13 @@ class Environment(BaseModel):
     co2: CO2 | None = None
     moisture: Moisture | None = None
     pH: PH | None = None
+
+
+class VPD(BaseModel):
+    """
+    VPD is a Computed Projection from other data points.
+    """
+
+    timestamp: datetime
+    vpd: float
+    unit: str