Diagnostic Studies of Numerical Long-Range Predictions

Colucci建议通过数值模型研究延长（30天）预测的准确性，并确定初始状态和演化状态的动态特征，这些特征区分了模型表现出低技能的时期、高技能的时期和预测技能急剧变化的时期，而在最初定义的状态中变化很小。要使用的技术是“案例合成”，根据相关属性（在这种情况下是预测技能）的客观度量对气候记录进行细分，并为每个子类别构建合成模式。通过发现复合材料之间物理上有意义的和可解释的差异来表明成功，然后可以使用这些差异来识别特定的模型缺陷。它高度依赖于实践者的流星学和气候学见解。在美国国家科学基金会之前的支持下，Colucci过去的成就表明他在这方面是完全合格的。这项研究将有助于美国全球变化研究计划的一个目标，即在1996-2000年期间实施季节气候的可操作数值预报模式。这样一个模型将成为后续预测多年气候变化（由自然和/或人为原因引起）的模型的垫脚石，具有足够的可信度，可作为在私营、国家和国际部门内制定资源管理政策的有用指南。

Colucci proposes to study the accuracy of extended (30-day) forecasts by a numerical model, and to identify those dynamical characteristics of initial and evolving states that distiguish between periods in which the model exhibits low skill, from those with high skill, and those in which forecast skill changes drasti- cally with minor variations in the initially defined state. The technique to be used is "case compositing", in which the climato- logical record is subdivided according to some objective measure of the relevant property (in this case, forecast skill) and composite patterns are constructed for each of the subcategories. Success is shown by finding physically meaningful and interpretable differ- ences between the composites, which may then be used to identify specific model deficiencies. It is highly dependent on the meteor- ological and climatological insights of the practitioner. Colucci's past accomplishments under prior NSF support show that he is well qualified in this respect. The study will contribute to one object- ive of the US Global Change Research Program, viz., an operational numerical forecast model for seasonal climate to be implemented during the 1996-2000 period. Such a model will serve as a stepping stone to subsequent models predicting multi-annual climate changes (induced by natural and/or man-made causes) with sufficient credi- bility to serve as useful guides in the framing of resource- management policy within the private, national, and international sectors.