Add DLWP (Deep Learning Weather Prediction) for Weather Forecasting

docs/source/modules/models_dlwp.rst

@@ -0,0 +1,35 @@
+DLWP
+========
+
+Description
+-----------
+
+``dlwp`` (Deep Learning Weather Prediction) is a U-Net architecture for 
+global weather forecasting on cubed-sphere grids. It predicts future atmospheric 
+states from current observations using a data-driven approach.
+
+Modular
+-------
+
+``dlwp``
+
+Example of how to use it
+------------------------
+
+.. code-block:: python
+    from pyhazards.models import build_model
+    import torch
+
+    # Build model
+    model = build_model(
+        name="dlwp",
+        task="regression",
+        in_channels=7,      # 7 atmospheric variables
+        num_faces=6,        # Cubed-sphere grid
+        face_size=64,       # 64×64 per face
+    )
+
+    # Single-step forecast
+    current_state = torch.randn(1, 7, 6, 64, 64)  # Current atmospheric state
+    future_state = model(current_state)            # State 6 hours later
+    print(future_state.shape)  # torch.Size([1, 7, 6, 64, 64])

docs/source/modules/models_wavecastnet.rst

@@ -1,13 +1,13 @@
 WaveCastNet
 ===========
 
-WaveCastNet is a deep learning model for earthquake wavefield forecasting using 
-sequence-to-sequence learning with Convolutional Long Expressive Memory (ConvLEM) cells.
-
-Overview
+Description
 --------
 
-WaveCastNet predicts the future evolution of seismic wavefields for earthquake early 
+``wavecastnet`` is a deep learning model for earthquake wavefield forecasting using 
+sequence-to-sequence learning with Convolutional Long Expressive Memory (ConvLEM) cells.
+
+It predicts the future evolution of seismic wavefields for earthquake early 
 warning systems. It uses a sequence-to-sequence architecture with ConvLEM cells to:
 
 - Process past wavefield observations (e.g., 60 timesteps of 3-component particle velocity)
@@ -17,6 +17,11 @@ warning systems. It uses a sequence-to-sequence architecture with ConvLEM cells
 The model operates on dense spatial grids and is designed for real-time forecasting 
 of ground motions without requiring explicit magnitude or epicenter estimation.
 
+Modular
+-------
+
+``wavecastnet``
+
 Example of how to use it
 ------------------------
 

docs/source/references.rst

@@ -24,4 +24,6 @@ Model Publications
 
 - Marjani et al. (2024). *Application of Explainable Artificial Intelligence in Predicting Wildfire Spread: An ASPP-Enabled CNN Approach*. `[link] <https://ieeexplore.ieee.org/document/10568207>`__.
 - Taghizadeh et al. (2025). *Interpretable physics-informed graph neural networks for flood forecasting*. `[link] <https://onlinelibrary.wiley.com/doi/10.1111/mice.13484>`__.
-- Lyu et al. (2025). *Rapid wavefield forecasting for earthquake early warning via deep sequence to sequence learning*. `[link] <https://doi.org/10.1038/s41467-025-65435-2>`__.
+- Lyu et al. (2025). *Rapid wavefield forecasting for earthquake early warning via deep sequence to sequence learning*. `[link] <https://doi.org/10.1038/s41467-025-65435-2>`__.
+- Wyn et al. (2020). *Improving Data-Driven Global Weather Prediction Using Deep Convolutional Neural Networks on a Cubed Sphere*. `[link] https://doi.org/10.1029/2020MS002109`__.
+- Wyn et al. (2021). *Sub-Seasonal Forecasting With a Large Ensemble of Deep-Learning Weather Prediction Models*. `[link] https://doi.org/10.1029/2021MS002502`__.

pyhazards/models/__init__.py

@@ -10,6 +10,7 @@
 from .cnn_aspp import WildfireCNNASPP, cnn_aspp_builder
 from .hydrographnet import HydroGraphNet, HydroGraphNetLoss, hydrographnet_builder
 from .wavecastnet import WaveCastNet, ConvLEMCell, wavecastnet_builder
+from .dlwp import DLWP, dlwp_builder, DoubleConv, Down, Up
 
 __all__ = [
     # Core API
@@ -43,6 +44,13 @@
     "WaveCastNet", 
     "ConvLEMCell",
     "wavecastnet_builder",
+
+    # Weather models
+    "DLWP",
+    "dlwp_builder",
+    "DoubleConv",
+    "Down",
+    "Up",
 ]
 
 # -------------------------------------------------
@@ -142,3 +150,15 @@
               "dropout": 0.1, 
     }, 
 )
+
+register_model(
+    "dlwp",
+    dlwp_builder,
+    defaults={
+        "num_levels": 4,
+        "base_channels": 64,
+        "kernel_size": 3,
+        "activation": "relu",
+        "dropout": 0.1,
+    },
+)