Raymond Schmitt

(Woods Hole Oceanographic Institution)

"Ocean salinity and machine learning for improved S2S forecasts of precipitation on land"

When Feb 26, 2020
from 03:30 pm to 04:30 pm
Where 112 Walker Building, John Cahir Auditorium
Contact Name Laifang Li
Contact email
Add event to calendar vCal

Ray Schmitt Woods Hole

Ocean salinity and machine learning for improved S2S forecasts of precipitation on land 

Ray Schmitt
Woods Hole Oceanographic Institution 

Most water on Earth is in the ocean and it is the ultimate source of terrestrial rain. Latent heat flux also dominates the transfer of energy from ocean to the atmosphere and land. While about 50% of surface evaporation from the subtropical ocean falls back as local precipitation, the rest is exported from the evaporation-dominated high pressure systems. This generates high sea surface salinity (SSS) in the subtropical gyres and low SSS in the high and low latitude oceans and coastal regions that receive runoff from land.  About 30% of the water exported from the subtropical oceans rains out on land, as expected from its fraction of global area. Anomalously large water export from the ocean leads to higher SSS, guaranteeing that some part of the climate system will experience more rain; lower SSS indicates less export and less rain elsewhere. We have found that seasonal anomalies in SSS in particular areas of the ocean have remarkable skill for predicting future rainfall in certain regions on land. That is, springtime SSS anomalies in different regions of the North Atlantic are good predictors of summer rain in the Sahel of Africa and the US Midwest. Similarly, spring South China Sea SSS can be used to predict summer rain in the Yangtze River valley. The delay mechanism involves soil moisture and feedbacks on the atmospheric circulation. Using a global analysis of fall SSS anomalies, we find even stronger predictability of winter time precipitation in the US Southwest. In all cases, SSS is superior to SST as a seasonal predictor. In addition, we have used global SSS and SST as predictors to win a sub-seasonal rainfall forecasting contest for the US West. We hypothesize that remote teleconnections arise from rain-generated SSS anomalies that are indicative of diabatic heating of the upper atmosphere which generates Rossby waves from the tropics to the extratropics.  SSS is superior to SST because it is an indicator of the large latent heat flux from the ocean which dominates the sensible flux by an order of magnitude as the prime driver of the atmosphere.