Deep learning enables ocean chlorophyll front detection using geostationary weather satellite
[Chl-a] maps showing the Gulf Stream off the Mid-Atlantic Bight on August 6, 2018, derived from (A) ABI onboard weather satellite GOES-16 (daily composite) using deep learning, (B) VIIRS ocean color data, (C) NOAA gap-filled, and (D) multi-sensor Ocean Colour Climate Change Initiative. For comparison, geo-polar blended (E) and (F) VIIRS-derived SST are also shown.
Credit: Guangming Zheng, Chris W. Brown
The conceptual design of the experimental SEM for extreme precipitation
The center oval represents the predictand (EP), while each outer oval defines one latent variable (LV) with a list in a brace of the designated manifest variables (MV) as the predictor can- didates. There are four LVs, hypothetically representing energy sup- ply (ES), water supply (WS), surface forcing (SF), and cloud forcing (CF). Each effect from one LV to another or to EP is expressed by an arrow for its direction and a coefficient along the line for its strength.
Credit: Chao Sun
Outcome of the generalized Spatio-Temporal Threshold Clustering method for three clusters by seasons for frequency of extreme daily precipitation field
Credit: Vitaly Kholodovsky and Xin-Zhong Liang
Annual linear trend for connectivity index for frequency of extreme precipitation field
Arrows show selected years, where n is the number of isolated structures, and m is the number of non-zero grid cells.
Credit: Vitaly Kholodovsky and Xi. Brown
Clusters of Monsoon Dry and Wet Years
Forecasting spatial distribution of rainfall remains a challenge. Clustering with AI offer ways forward. Producing local forecasts at S2S timescales for Food, Water, Energy and Health by applying AI/ML to data and forecasts will advance climate services.
Credit: Sahastrabuddhe et al. 2019
Deep Learning for Efficient Radiative Transfer Modeling
The radiative transfer model (RTM) is one of the most critical yet time-consuming elements of the inverse and assimilation problems. Deep Learning (DL) shows promising results towards efficient and accurate RTM simulations.
Xingming Liang, Kayo Ide (ESSIC), Kevin Garrett (NOAA)
The rapid growth of Artificial Intelligence (AI) and Machine Learning (ML) has brought tremendous opportunities to advance our knowledge in earth science. This forum provides a platform for ESSIC scientists to foster novel research while connecting to inter- and cross-disciplinary AI/ML communities on campus and beyond.
The expertise of ESSIC-AI Forum broad and deep, inter-disciplinary research on earth science encompassing theory, modeling, data analysis, and prediction, as well as cross-disciplinary research such as impact of climate change on our society.