Model quality objectives based on measurement uncertainty. Part I: Ozone
Model quality objectives based on measurement uncertainty. Part I: Ozone
复制标题
基于测量不确定性的质量目标模型。
DOI：
10.1016/j.atmosenv.2013.05.018
复制
发表时间：
2013
影响因子：
5
通讯作者：
M. Gerboles
中科院分区：
文献类型：
--
作者：
P. Thunis;D. Pernigotti;M. Gerboles
由于模型越来越多地用于政策支持，其评估正成为一个重要问题。可能的评估之一是将模型结果与测量结果进行比较。然后，统计性能指标可以提供有关模型性能的见解，但不能说明模型结果是否已达到给定应用程序的足够质量水平。在清华紫光等人之前的工作中。 （2012 年，简称 T2012）提出了模型质量目标（MQO），该目标基于测量浓度和建模浓度之间的均方根误差除以测量不确定度。在 T2012 中，假设测量不确定度保持恒定，无论浓度水平如何。在当前的工作中，通过量化 O3 特定情况下所有可能的不确定性来源，克服了这一假设。基于这些不确定性源量化，提出了一种简单的关系来制定测量不确定性，然后用更准确的值更新 T2012 中提出的 MQO 和模型性能标准 (MPC)。根据欧洲监测网络 AIRBASE 数据计算的 MQO 和 MPC 可根据地理区域和站点类型深入了解给定应用的预期模型结果质量。这些特定于站的 MQO 和 MPC 的主要优点是将预期模型性能与潜在的测量不确定性相关联。
Since models are increasingly used for policy support their evaluation is becoming an important issue. One of the possible evaluations is to compare model results to measurements. Statistical performance indicators then provide insight on model performance but do not tell whether model results have reached a sufficient level of quality for a given application. In a previous work Thunis et al. (2012, referred to as T2012) proposed a Model Quality Objective (MQO) based on the root mean square error between measured and modeled concentrations divided by the measurement uncertainty. In T2012 the measurement uncertainty was assumed to remain constant regardless of the concentration level. In the current work this assumption is overcome by quantifying all possible sources of uncertainty for the particular case of O3. Based on these uncertainty source quantifications, a simple relationship is proposed to formulate the measurement uncertainty which is then used to update the MQO and Model Performance Criteria (MPC) proposed in T2012 with more accurate values. The MQO and MPC calculated based on the European monitoring network AIRBASE data provide insight on the expected model results quality for a given application, depending on the geographical area and station type. These station specific MQOs and MPCs have the main advantage of relating expected model performances to the underlying measurement uncertainties.