Previous topic

Preprocessing Class

Next topic

Source Code

This Page

Operational Model Class

AMSIMP Operational Model Class. For information about this class is described below.

Copyright (C) 2021 AMSIMP

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>.

class amsimp.forecasting.OperationalModel(forecast_length=120, amsimp_ic=True, initialisation_conditions=None)[source]

This is the operational model class for AMSIMP.

generate_forecast()[source]

Generates a forecast with the current AMSIMP Global Forecast Model architecture.

Returns

The forecast generated with the operational model

Return type

iris.cube.CubeList

Notes

This model has been pretrained on the dataset from the year 2009 to the year 2016. The architecture of the current operational model is currently based on the ConvLSTM layer. The prognostic variables are: air temperature at 2 metres above the surface, air temperature at a pressure surface of 850 hectopascals, and geopotential at a pressure surface of 500 hectopascals.

model_architecture()[source]

Generates the operational AMSIMP Global Forecast Model architecture.

Returns

Operational AMSIMP Global Forecast Model architecture.

Return type

tf.keras.Sequential

Notes

This architecture is currently based on the ConvLSTM layer, which has been pretrained on the dataset from the year 2009 to the year 2016. A major drawback of LSTMs in its handling of spatiotemporal data is due to its usage of full connections in input-to-state and state-to-state transitions in which no spatial information is encoded. To overcome this problem, a distinguishing feature of a ConvLSTM cell is that all the inputs and gates of the ConvLSTM layer are 3D tensors whose last two dimensions are spatial dimensions.