5.1.2 Statistical performance measures

Statistical performance measures can provide meaningful

measures of model accuracy for dense monitoring networks, such as

those for special field studies. However, statistical measures may

give a distorted view of model performance in cases of routine

monitoring networks, where coverage may be sparse. The Technical

Committee should evaluate the adequacy of the existing monitoring

network for conducting statistical tests for performance



It is recommended that, at a minimum, the following

mathematical formulations be applied as measures for model

performance evaluation. These formulations are detailed in

Appendix C.

Unpaired highest-prediction accuracy - This measure quantifies

the difference between the highest observed value and highest

predicted value over all hours and monitoring stations.

Normalized bias test - This test measures the model's ability

to replicate observed patterns during the times of day when

available monitoring and modeled data are most likely to

represent similar spatial scales.

Gross error of all pairs above 60 ppb - In conjunction with

bias measurements, this metric provides an overall assessment

of base case performance and can be used as a reference to

other modeling applications. Gross error can be interpreted

as precision.



n1151 - n1152 - n1153 - n1154 - n1155 - n1156 - n1157 - n1158 - n1159 - n1160 - n1161 - n1162 - n1163 - n1164 - n1165 - n1166 - n1167 - n1168 - n1169 - n1170 - n1171 - n1172 - n1173 - n1174 - n1175 - n1176 - n1177 - n1178 - n1179 - n1180 - n1181 - n1182 - n1183 - n1184 - n1185 - n1186 - n1187 - n1188 - n1189 - n1190 - n1191 - n1192 - n1193 - n1194 - n1195 - n1196 - n1197 - n1198 - n1199 - n1200


   Flag of Portugal 


 castellano: DISPER CUSTIC DESCAR RADIA    italiano:     


 français:    português:  





deutsch: DIS CUS  DES  RAD

castellano: DIS CUS DES  RAD   english: DIS CUS DES RAD  


 português: DIS CUS DES RAD   italiano:   DIS CUS  DES RAD


français:  DIS CUS DES RAD