Differences

This shows you the differences between two versions of the page.

--- aerocom:faq_aerocom [2015-06-22 13:53:49]
michaels
+++ aerocom:faq_aerocom [2022-05-31 09:29:31] (current)
@@ Line 1: / Line 1: @@
+ATTENTION - THIS WIKI PAGE IS NO LONGER UPDATED - PLEASE GO TO [[http://aerocom.met.no/|aerocom.met.no]]FOR LATEST INFO
 ====== Frequently asked questions AeroCom Web Interface ======
@@ Line 93: / Line 95: @@
 ==== Definition of Statistics used in Score tables and in Scatter plots ====
-"Obs": Observational mean constructed from all values in scatter plot (e.g. from yearly station means : Example)
+Explanation for abbreviations used on scatter plots and in score tables:
+"#": Number of "observations" used for statistics (eg daily, monthly or yearly values)
+"#st": Number of sites
+"Obs": Observational mean constructed from all values in scatter plot (e.g. means from yearly station means are different from all daily value means : [[http://aerocom.met.no/cgi-bin/aerocom/surfobs_annualrs.pl?PROJECT=cci-Aerosol&MODELLIST=CCI-AEROSOL-products&FULL=2abbrev&INFO=nohover&PERFORMANCE=ind&YEARFILTER=ALLYEARS&PSFILTER=ALLVARS&Type0=SCATTERLOG&Ref0=AERONETSun2.0&Run0=AATSR_SU_v4.2&Parameter0=OD550_AER&Station0=WORLD&Year0=an2002&Period0=YEARLY&Type1=SCATTERLOG&Ref1=AERONETSun2.0&Run1=AATSR_SU_v4.2&Parameter1=OD550_AER&Station1=WORLD&Year1=an2002&Period1=mALLYEARdaily|Example]])
+"Mod": Model/Dataset mean constructed from all values in scatter plot, eg correspond to "Obs" values consistently.
+"NMB": Normalized mean bias, = (Mod-Obs)/Obs
+"RMS": Root Mean Square Error: SQRT(Sum( (y'-x')^2)/#)
+"R": Pearson Correlation Coefficient
+"RMSbc": Bias corrected RMS: SQRT(Sum( (ybc'-x')^2)/#) ; ybc' contains all model/data values corrected for the NMB bias. ybc=y*(Sum(x)/Sum(y))
+"RMSspatial": RMS from yearly mean station vales in observations and models. Model values consistently chosen for times when observations are available.
+"RMStemporal": Residual RMS after subtracting RMSspatial from RMS ( RMStemporal = SQRT( RMS^2 - RMSspatial^2 ), assuming error propagation and no covariance in temporal and spatial fields)
+...additional statistics in score tables
+"Spearman Rank Correlation": https://en.wikipedia.org/wiki/Spearman%27s_rank_correlation_coefficient
+"Spatial yearly Mean Corr Coeff": Pearson correlation coefficient from yearly mean values
+"Seasonal Anomaly Corr Coeff": Correlation coefficient of local anomalies wrt to annual mean.
+"Slope fit forced through zero": Regression of model on obs data, slope
+"Regression coefficient": Regression of model on obs data, regression
+"Regression constant, offset": Y intercept
+"STDDEV(Model)/STDDEV(Data)". Ratio of standard deviations
+"Score (mean relative bias )": sum(abs(tm-td))/# [ tm= model values, td= obs values ]
-For daily data analysis, RMS is split into the "spatial" RMS, calculated from monthly/seasonal/yearly averages and the "temporal" RMS, which is the residual (RMS^2temp=RMS^2-RMS^2spatial)
+"Taylor score": 4*(1+r)/((1+0.9)*(sm/sd+sd/sm)*(sm/sd+sd/sm) ) [ sm= Stddev of model, sd=Stddev of observations ]
 ==== Known bugs/problems/wishes ====
 -mountain sites are compared from surface level model output, use relative height\\