Development of an empirical model for seasonal forecasting over the Mediterranean
In the frame of MEDSCOPE project, which mainly aims at improving predictability on seasonal timescales over the Mediterranean area, a seasonal forecast empirical model making use of new predictors based on a collection of targeted sensitivity experiments is being developed. Here, a first version of...
Published in: | Advances in Science and Research |
---|---|
Main Authors: | , , , , |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
Copernicus Publications
2019
|
Subjects: | |
Online Access: | https://doi.org/10.5194/asr-16-191-2019 https://noa.gwlb.de/receive/cop_mods_00000151 https://noa.gwlb.de/servlets/MCRFileNodeServlet/cop_derivate_00000125/asr-16-191-2019.pdf https://asr.copernicus.org/articles/16/191/2019/asr-16-191-2019.pdf |
Summary: | In the frame of MEDSCOPE project, which mainly aims at improving predictability on seasonal timescales over the Mediterranean area, a seasonal forecast empirical model making use of new predictors based on a collection of targeted sensitivity experiments is being developed. Here, a first version of the model is presented. This version is based on multiple linear regression, using global climate indices (mainly global teleconnection patterns and indices based on sea surface temperatures, as well as sea-ice and snow cover) as predictors. The model is implemented in a way that allows easy modifications to include new information from other predictors that will come as result of the ongoing sensitivity experiments within the project. Given the big extension of the region under study, its high complexity (both in terms of orography and land-sea distribution) and its location, different sub regions are affected by different drivers at different times. The empirical model makes use of different sets of predictors for every season and every sub region. Starting from a collection of 25 global climate indices, a few predictors are selected for every season and every sub region, checking linear correlation between predictands (temperature and precipitation) and global indices up to one year in advance and using moving averages from two to six months. Special attention has also been payed to the selection of predictors in order to guaranty smooth transitions between neighbor sub regions and consecutive seasons. The model runs a three-month forecast every month with a one-month lead time. |
---|