Please use this identifier to cite or link to this item:
Research center of Embrapa/Collection: Embrapa Meio Ambiente - Artigo em anais de congresso (ALICE)
Date Issued: 2006
Type of Material: Artigo em anais de congresso (ALICE)
Authors: MAIA, A. de H. N.
Additional Information: ALINE DE HOLANDA NUNES MAIA, CNPMA; H. MEINKE, Department of Primary Industries and Fisheries, Austrália.
Title: Assessing uncertainty of seasonal probabilistic forecasts: distribution-free confidence limits.
Publisher: In: INTERNATIONAL CONFERENCE ON SOUTHERN HEMISPHERE METEOROLOGY AND OCEANOGRAPHY, 8., 2006, Foz do Iguaçu,PR. [Anais?]. Foz do Iguaçu,PR: ICSHMO, 2006. p. 569-573.
Language: pt_BR
Description: Probabilistic climate information, including climate forecasts, often rely on time series data of prognostic variables (Y, eg. rainfall or yield), represented as cumulative distribution probabilities functions (CDFs) or their complement, probability of exceeding functions (POEs). Useful information for decision-making is then derived from such distributions and expressed as Y percentiles or the probability of Y exceeding a certain threshold c (Pr[Y> c]). Such estimates are frequently reported without any measure of uncertainty. The degree of uncertainty associated with such estimates depends on the length of the time series and their internal variability. Lack of uncertainty assessments can lead to misguided beliefs about the true performance of the forecast systems, possibly resulting in inappropriate actions by the decision maker (Potts et al. 1996; Jolliffe 2004; Maia et al. 2006). However, even when uncertainty estimates are provided, these are often based on methods that rely on assumptions of data being normally distributed. This is in spite of the well-known fact that distributions of important climate variables, such as rainfall, are notoriously skewed, particularly in areas with strong seasonality (eg. high frequencies of ?zero? rainfall amounts). As an alternative for Normal-based procedures, we therefore propose the use of distribution free methods for constructing percentile and POE confidence limits as described in Hahn and Meeker (1991) and implemented into ?The Capability Procedure?? of the SAS® System. Such distribution-free tools are particularly useful for spatial uncertainty assessments that would otherwise require a tedious, location-by-location checking of assumptions regarding underlying probability distributions (Maia et al., 2006). Here, we discuss the rationale, advantages and limitations of both, parametric and nonparametric approaches. We illustrate the use of distribution-free methods by assessing the uncertainty of percentiles and POEs estimates for 3-monthly rainfall series from selected locations in Australia and the Southeast of South America.
Thesagro: Estatística
Previsão do tempo
Data Created: 2015-10-08
Appears in Collections:Artigo em anais de congresso (CNPMA)

Files in This Item:
File Description SizeFormat 
2006AA011.pdf3,03 MBAdobe PDFThumbnail

FacebookTwitterDeliciousLinkedInGoogle BookmarksMySpace