A Sequence-to-Sequence ConvLSTM Approach for Leaf Area Index Forecasting over the South-Central United States

Abstract

Leaf Area Index (LAI) is a fundamental biophysical variable governing land-atmosphere interactions; however, LAI forecasting at high spatial resolution remains an unsolved challenge. While recent machine learning approaches have demonstrated LAI estimation at point or regional scales, none provides a gridded, meteorology-driven prognostic forecast suitable for subseasonal land surface and climate modeling applications. Here we present a sequence-to-sequence Convolutional LSTM (ConvLSTM) framework that generates daily 1-km LAI forecasts up to 30 days ahead, driven by historical LAI sequences and daily meteorological forcing including temperature and precipitation. Trained and evaluated over the South-Central United States -- a region of strong climate gradients and diverse vegetation -- the model achieves a domain-averaged RMSE of 0.36 at a 30-day lead time, more than a third lower than the persistence baseline. Forecast skill remains robust across seasons, geographic distributions, and plant functional types, including forests, grasslands, shrublands, and croplands. To our knowledge, this is the first demonstration of skillful LAI forecasting at a 30-day horizon at 1-km resolution.

Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.