陶守正 （）　2008.03.18 01:02:20

三、世界气象组织（WMO）评分办法的相关原文及中译文

Table 4 shows the general form for the three possible two by two contingency tables referred to above (the third is the table for the near normal category and the combined above and below normal category). In Table 4, T is the grand sum of all the proper weights applied on each occurrence and non-occurrence of the events.

表格4展示了上面所说的三个可能的2×2列连报的普通形式。(第三类是接近正常和高于与低于正常合并的类型). 表格4中, T 是每个发生和没发生事件的所有恰当的权重的总和。

Table 4: General ROC contingency table for deterministic forecasts.

		Observations
		occurrences	non-occurrences
forecasts	occurrences	O1	NO1	O1+ NO1
	non-occurrences	O2	NO2	O2+ NO2
		O1+ O2	NO1+ NO2	T

表格 4: 确定性预报的普通 ROC列联表

		观测
		发生	未发生
预报	发生	O1	NO1	O1+ NO1
	未发生	O2	NO2	O2+ NO2
		O1+ O2	NO1+ NO2	T

The 2X2 table in Table 4 may be constructed from the 3X3 table described in Table 3 by summing the appropriate rows and columns.

表格4中的2X2表格是由表格3中3X3表格合并相近的行和列而得到的。

In Table 4, O₁ represents the correct forecasts or hits:

表格4中, O₁代表正确预报或准确:

(OF) being 1 when the event occurrence is observed and forecast; 0 otherwise. The summation is over all grid points or stations.

当预报发生事件在观测中也发生了，则(OF)=1；否则(OF)= 0。 总和是所有栅格点或站点的加和。

NO₁ represents the false alarms:

NO₁ 表示误报：

(NOF) being 1 when the event occurrence is not observed but was forecast; 0 otherwise. The summation is over all grid points or stations.

当预报发生的事件在观测中未发生，则(NOF)= 1 ；否则(NOF)=0 。总和是所有栅格点或站点的加和。

O₂ represents the misses:

O₂ 表示遗漏:

(ONF) being 1 when the event occurrence is observed but not forecast; 0 otherwise. The summation is over all grid points or stations.

当预报未发生的事件却在观测中发生时，则(ONF) = 1 ；否则(ONF)=0 。总和是所有栅格点或站点的加和。

NO₂ represents the correct rejections:

NO₂ 表示正确的否决：

(NONF) being 1 when the event occurrence is not observed and not forecast; 0 otherwise. The summation is over all grid points or stations.

当预报和观测中为未发生的事件时，(NONF)= 1 ；否则(NONF)= 0 。总和是所有栅格点或站点的加和。

when verification is done at stations or at single grid points.

此时，i是在一个有限的地理区域内的站点或单独的栅格点上；

at grid point i, when verification is done on a grid.

此时，i在栅格面上，而非栅格交点上

the latitude at grid point i.

为栅格点的纬度

When verification is done at stations, the weighting factor is one. Consequently, the number of occurrences and non-occurrences of the event are entered in the contingency table of Table 4.

当在站点上进行验证时, 权重因素为1。所以, 将发生和未发生事件的数目输入到列联表4中。

However, when verification is done on a grid, the weighting factor is cos(q_i), where q_i is the latitude at grid point i. Consequently, each number entered in the contingency table of Table 5, is, in fact, a summation of the weights properly assigned.

如果当验证是在栅格平面上，权重为cos(q_i),其中 q_i 栅格点i的纬度。所以，输入列联表5中的数字实际上是 指定的权重的加和。

Using stratification by observations (rather than by forecast), the Hit Rate (HR) is defined as (referring to Table 4):

利用观测的分层(而非预报的分层) ，准确率(HR)为 (O为表格4中的数据):

The range of values for HR goes from 0 to 1, the latter value being desirable. An HR of one means that all occurrences of the event were correctly forecast.

HR 的范围为0 到 1, 可以取到1。 HR=1意味着所用发生的事件被正确的预报。

The False Alarm Rate (FAR) is defined as:

误报率(FAR)为:

The range of values for FAR goes from 0 to 1, the former value being desirable. A FAR of zero means that in the verification sample, no non-occurrences of the event were forecast to occur.

FAR取值范围为 0 到 1,可以取到0。FAR=0意味着在验证样本中未发生的事件也被准确预报。

Hanssen and Kuipers score (see Hanssen and Kuipers, 1965 and Stanski et al, 1989) is calculated for deterministic forecasts. Hanssen and Kuipers score (KS) is defined as:

Hanssen 和 Kuipers 的评分 (见 Hanssen 和 Kuipers, 1965以及 Stanski et al, 1989) 用于确定性预报的计算。Hanssen 和 Kuipers score (KS) 定义式为:

The range of KS goes from -1 to +1, the latter value corresponding to perfect forecasts (HR being 1 and FAR being 0). KS can be scaled so that the range of possible values goes from 0 to 1 (1 being for perfect forecasts):

KS的取值范围为-1到 +1, KS=1表示精确预报(即HR = 1 ； FAR = 0)。 KS 的取值范围也可转化为0 到 1 (KS=1表示精确预报):

The advantage of scaling KS is that it becomes comparable to the area under the ROC curve for probabilistic forecasts (see section 3.33) where a perfect forecast system has an area of one and a forecast system with no information has an area of 0.5 (HR being equal to FAR).

标准化 KS的优点是使其在概率性预报的ROC曲线的区域内可以用于 比较(见 3.33)。一个精确的预报系统ROC曲线区域的面积为1，无信息的预报系统ROC曲线区域的面积为0.5(HR=FAR) 。

Þ Contingency tables for deterministic categorical forecasts (such as in Table 3) are mandatory for level 3 verification in the core SVS. These contingency tables can provide the basis for the calculation of several scores and indices such as the Gerrity Skill Score, the LEPSCAT or the scaled Hanssen and Kuipers score and others.

Þ 确定性分类预报的列联表 (例如表格3) 受控于Core SVS中的层次3的验证。这些列联表可以为几种评分和索引提供基矗例如： Gerrity 技术评分(GSS)， LEPSCAT 或者 标准化的Hanssen 和 Kuipers的评分以及其他模型。

网站简介　投稿指南　特别声明