MSGID: 1:317/3 646ee4b3   
   PID: hpt/lnx 1.9.0-cur 2019-01-08   
   TID: hpt/lnx 1.9.0-cur 2019-01-08   
    New method predicts extreme events more accurately    
    Columbia Engineers develop machine-learning algorithm to better   
   understand and mitigate the impact of extreme weather events, which are   
   becoming more frequent in our warming climate.    
      
     Date:   
         May 24, 2023   
     Source:   
         Columbia University School of Engineering and Applied Science   
     Summary:   
         A new study has used global storm-resolving simulations and machine   
         learning to create an algorithm that can deal separately with   
         two different scales of cloud organization: those resolved by a   
         climate model, and those that cannot be resolved as they are too   
         small. This new approach addresses the missing piece of information   
         in traditional climate model parameterizations and provides a way   
         to predict precipitation intensity and variability more precisely.   
      
      
         Facebook Twitter Pinterest LinkedIN Email   
      
   ==========================================================================   
   FULL STORY   
   ==========================================================================   
   With the rise of extreme weather events, which are becoming more   
   frequent in our warming climate, accurate predictions are becoming more   
   critical for all of us, from farmers to city-dwellers to businesses   
   around the world. To date, climate models have failed to accurately   
   predict precipitation intensity, particularly extremes. While in nature,   
   precipitation can be very varied, with many extremes of precipitation,   
   climate models predict a smaller variance in precipitation with a bias   
   toward light rain.   
      
   Missing piece in current algorithms: cloud organization Researchers   
   have been working to develop algorithms that will improve prediction   
   accuracy but, as Columbia Engineering climate scientists report, there   
   has been a missing piece of information in traditional climate model   
   parameterizations -- a way to describe cloud structure and organization   
   that is so fine-scale it is not captured on the computational grid   
   being used. These organization measurements affect predictions of both   
   precipitation intensity and its stochasticity, the variability of random   
   fluctuations in precipitation intensity. Up to now, there has not been   
   an effective, accurate way to measure cloud structure and quantify   
   its impact.   
      
   A new study from a team led by Pierre Gentine, director of the Learning   
   the Earth with Artificial Intelligence and Physics (LEAP) Center, used   
   global storm-resolving simulations and machine learning to create an   
   algorithm that can deal separately with two different scales of cloud   
   organization: those resolved by a climate model, and those that cannot be   
   resolved as they are too small. This new approach addresses the missing   
   piece of information in traditional climate model parameterizations   
   and provides a way to predict precipitation intensity and variability   
   more precisely.   
      
   "Our findings are especially exciting because, for many years, the   
   scientific community has debated whether to include cloud organization in   
   climate models," said Gentine, Maurice Ewing and J. Lamar Worzel Professor   
   of Geophysics in the Departments of Earth and Environmental Engineering   
   and Earth Environmental Sciences and a member of the Data Science   
   Institute. "Our work provides an answer to the debate and a novel solution   
   for including organization, showing that including this information   
   can significantly improve our prediction of precipitation intensity and   
   variability."  Using AI to design neural network algorithm Sarah Shamekh,   
   a PhD student working with Gentine, developed a neural network algorithm   
   that learns the relevant information about the role of fine-scale cloud   
   organization (unresolved scales) on precipitation. Because Shamekh did   
   not define a metric or formula in advance, the model learns implicitly   
   -- on its own -- how to measure the clustering of clouds, a metric   
   of organization, and then uses this metric to improve the prediction   
   of precipitation. Shamekh trained the algorithm on a high-resolution   
   moisture field, encoding the degree of small-scale organization.   
      
   "We discovered that our organization metric explains precipitation   
   variability almost entirely and could replace a stochastic   
   parameterization in climate models," said Shamekh, lead author of the   
   study, published May 8, 2023, by PNAS. "Including this information   
   significantly improved precipitation prediction at the scale relevant   
   to climate models, accurately predicting precipitation extremes and   
   spatial variability."  Machine-learning algorithm will improve future   
   projections The researchers are now using their machine-learning approach,   
   which implicitly learns the sub-grid cloud organization metric, in climate   
   models. This should significantly improve the prediction of precipitation   
   intensity and variability, including extreme precipitation events, and   
   enable scientists to better project future changes in the water cycle   
   and extreme weather patterns in a warming climate.   
      
   Future work This research also opens up new avenues for investigation,   
   such as exploring the possibility of precipitation creating memory,   
   where the atmosphere retains information about recent weather conditions,   
   which in turn influences atmospheric conditions later on, in the climate   
   system. This new approach could have wide-ranging applications beyond   
   just precipitation modeling, including better modeling of the ice sheet   
   and ocean surface.   
      
       * RELATED_TOPICS   
             o Earth_&_Climate   
                   # Weather # Global_Warming # Climate #   
                   Environmental_Awareness   
             o Computers_&_Math   
                   # Computer_Modeling # Mathematical_Modeling #   
                   Computer_Programming # Distributed_Computing   
       * RELATED_TERMS   
             o Global_climate_model   
             o Temperature_record_of_the_past_1000_years o   
             Climate_model o Computer_simulation o Weather_forecasting   
             o Global_warming_controversy o Artificial_neural_network   
             o Alan_Turing   
      
   ==========================================================================   
   Story Source: Materials provided by   
   Columbia_University_School_of_Engineering_and_Applied Science. Original   
   written by Holly Evarts. Note: Content may be edited for style and length.   
      
      
   ==========================================================================   
   Journal Reference:   
      1. Sara Shamekh, Kara D. Lamb, Yu Huang, Pierre Gentine. Implicit   
      learning   
         of convective organization explains precipitation stochasticity.   
      
         Proceedings of the National Academy of Sciences, 2023; 120 (20)   
         DOI: 10.1073/pnas.2216158120   
   ==========================================================================   
      
   Link to news story:   
   https://www.sciencedaily.com/releases/2023/05/230524181937.htm   
      
   --- up 1 year, 12 weeks, 2 days, 10 hours, 50 minutes   
    * Origin: -=> Castle Rock BBS <=- Now Husky HPT Powered! (1:317/3)   
   SEEN-BY: 15/0 106/201 114/705 123/120 153/7715 218/700 226/30 227/114   
   SEEN-BY: 229/110 112 113 307 317 400 426 428 470 664 700 291/111 292/854   
   SEEN-BY: 298/25 305/3 317/3 320/219 396/45   
   PATH: 317/3 229/426