Monthly Seasonal Weather Outlook

Author : Lenin Del Rio Amador
Publisher :
ISBN 13 :
Total Pages : pages
Book Rating : 4.:/5 (129 download)

DOWNLOAD NOW!

Download or read book The Stochastic Seasonal to Interannual Prediction System written by Lenin Del Rio Amador and published by . This book was released on 2021 with total page pages. Available in PDF, EPUB and Kindle. Book excerpt: "This thesis proposes the new Stochastic Seasonal to Interannual Prediction System (StocSIPS) model of the atmospheric temperature in the macroweather regime, i.e. time scales longer than the lifetime of planetary sized structures (≈10 days). StocSIPS is stochastic and respects two conservation principles: energy balance and scale invariance. The scale-invariance of the fluctuations implies the existence a huge memory in the system that can be exploited for macroweather forecasts and the low intermittency justifies using well-established (Gaussian) techniques. Here, we present StocSIPS, discuss its statistics, use it to perform long-term forecasts and compare it with forecasts from conventional general circulation models (GCMs).GCMs are initial value problems and each has its own climate, requiring complex post-processing correction schemes. In comparison, StocSIPS is a past value problem based on real world historical data that effectively forces predictions to converge to the real-world climate. It extracts the internal variability (weather noise) directly from past data and does not suffer from model drift. It models the temperature as the high-frequency limit of the (fractional) energy balance equation which governs radiative equilibrium processes when the relevant equilibrium relaxation processes are power law, rather than exponential.StocSIPS was first developed as an improvement over the previous ScaLIng Macroweather Model (SLIMM) for monthly and seasonal forecast of globally averaged temperature. The validity of the model is checked through statistical testing and by comparing the theoretically expected skill scores with actual skill scores obtained from hindcast verification. A detailed comparison with the Canadian Seasonal to Interannual Prediction System (CanSIPS) shows that StocSIPS is just as accurate for one-month forecasts, but significantly more accurate for longer lead times.The scaling symmetry, which is the basis of StocSIPS, also holds at the regional level (here, 2.5°×2.5°). This allows applications for predicting the spatially resolved temperature field by treating each grid point as an independent time series. We obtain monthly and seasonal predictions of the surface temperature and show some preliminary comparison with multi-model ensemble (MME) GCM results. For one month lead times, this single pixel based stochastic model shows skill scores similar to the much more complex conventional (GCM) deterministic models.To explore the possibility of improving the predictions at a given place using past temperatures from other grid points, we developed the comprehensive multivariate version of StocSIPS (m-StocSIPS) as a full space-time process. m-StocSIPS allows to show that the regional memory-based StocSIPS forecasts are optimal in this stochastic framework. It also helps to untangle the usually complicated relationship between correlations and causality thanks to the precise notion of Granger causality. For a given position, past information from other locations cannot be used to improve on the forecast obtained as an optimal linear combination of past data: those correlations "were already used" to build the past at that position. By using m-StocSIPS we are able to reproduce the empirical cross-correlation structure of the temperature field over a wide range of time lags. This is made more convincing by making simulations that display numerous realistic but emergent model properties including spatial teleconnection networks and realistic El Niño patterns and indices"--