In this work we analyze predictability and complexity of monthly rainfall temporal series recorded from 1950 to 2012, at 133 gauging stations in Pernambuco state, northeastern Brazil. To this end we use the complexity entropy causality plane (CECP) and Fisher Shannon plane (FS) formed by information quantifiers permutation entropy, permutation statistical complexity, and Fisher information measure, which serve to evaluate randomness and structural organization of the underlying process. By comparing the locations of analyzed stations in CECP and FS and performing statistical significance test, we distinguish rainfall regime in the deep inland, drier Sertão region, from the intermediate inland Agreste, and coastal, tropical humid Zona da Mata regions. We also perform time dependent CECP and FS analysis and for Sertão region identify the periods of higher entropy (lower complexity and Fisher information measure) which are related to El Niño episodes and historical droughts in northeastern Brazil. Our work represents a contribution to establishing the use of information theory based methods in climatological studies.

在这项工作中，我们分析了1950年至2012年在巴西东北部伯南布哥州133个测量站记录的每月降雨时间序列的可预测性和复杂性。为此，我们使用由信息量词置换熵，置换统计复杂度和Fisher信息量度形成的复杂度熵因果平面（CECP）和Fisher香农平面（FS），以评估基础过程的随机性和结构组织。通过比较CECP和FS中分析站的位置并进行统计显着性检验，我们区分了深内陆，干燥Sertão地区，中陆内陆阿格里斯特和沿海，热带湿润的Zona da Mata的降雨状况地区。我们还执行时间依赖性CECP和FS分析，并针对Sertão地区确定与巴西东北部厄尔尼诺事件和历史干旱有关的较高熵（较低复杂度和Fisher信息测度）的时期。我们的工作为建立基于信息论的气候学研究方法做出了贡献。