VI Seminar Homogenization, Budapest 2008 M.Mendes, J.Neto, A.Silva, L.Nunes, P.Viterbo Instituto de Meteorologia, Portugal “Characterization of data sets for the assessment of inhomogeneities of climate data series, resulting from the automation of the observing network in Mainland Portugal

VI Seminar Homogenization, Budapest 2008 Current IM network with overlapping observations 30 sites with Automatic Weather Stations (AWS) and Conventional Stations (CS) 2 sites also with Present Weather Sensors (WW)

VI Seminar Homogenization, Budapest 2008 The problem: continuation of conventional data series with data from Automatic Weather Station?

VI Seminar Homogenization, Budapest 2008 OVERLAPPING PERIODS OF AWS AND CS DATA

VI Seminar Homogenization, Budapest 2008 Station Features Automatic Weather Stations Type I - AWS1 (15)Type II – AWS2 (15) Temperature (air,ground) Humidity Wind Global radiation Pressure Precipitation Interactive terminal (TIC) 10 minute records Temperature (air, ground) Humidity Wind Global radiation Precipitation 10 minute records Conventional Stations Principal – CS1 (15) Simple – CS2 (15) Temperat.(air, ground) Humidity Wind Sunshine duration Pressure Precipitation Visual parameters (cloud cover, cloud type, present & past weather, horiz. visibility) Daily records 1,2 hourly records/day Profissional Observers Temperature (air & ground) Wind Sunshine duration Precipitation Visual parameters (cloud cover, present & past weather, horizontal visibility,...) Daily records 1,2 hourly records/day Volunteers Present Weather Systems (2) Precipitation sensor Horizontal visibility sensor

VI Seminar Homogenization, Budapest 2008 Conventional Observations and Instruments Piranómetro Radiation screen Thermometers Thermo-hygrograph Rain gauge Sunshine- recorder Evaporation pan Mercury Barometer Visual observations

VI Seminar Homogenization, Budapest 2008 Automatic weather station sensors and equipments Wind vane and anemometer Pyranometer Rain detector Radiation shield with temp. & hum. sensors Rain gauge AWS with solar panel GSM Antena Data acquisition system

VI Seminar Homogenization, Budapest 2008 Data records/failures AWS vs CS (10 years data) AIR TEMPERATURE CodeParameter T009Air temperature at 09 UTC T015Air temperature at 15 UTC T018Air temperature at 18 UTC Tmax Maximum temperature (09-09 UTC) AWS – 144 consecutive 10 minute records and/or 24 consecutive hour records * CS – 1 record Tmin Minimum temperature (09-09 UTC) AWS – 144 consecutive 10 minute records and/or 24 consecutive hour records * CS – 1 record Tmn1 Maximum temperature (09-18 UTC) AWS – 60 consecutive 10 minute records and/or 10 consecutive hour records * CS – 1 record Tmx1 Minimum temperature (00-10 UTC) AWS – 66 consecutive 10 minute records and/or 11 consecutive hour records * CS – 1 record AWS - % FAILURE DAYS Nr.NDaysT009T015T018TmaxTminTmn1Tmx CS - % FAILURE DAYS NrNdaysT009T015T018TmaxTmin

VI Seminar Homogenization, Budapest 2008 Bias results for air temperature (Differences between observations AWS-CS) BIAS T T T Tmin Tmn Tmax Tmx < to > > 11 cases for T09, 21 cases for Tmin, 19 cases for Tmax

VI Seminar Homogenization, Budapest 2008 Spatial distribution of Bias results for air temperature (AWS-CS)

VI Seminar Homogenization, Budapest 2008 Bias monthly results for air temperature TmaxJanFebMarAprMayJunJulAugSepOctNovDec Tmin JanFebMarAprMayJunJulAugSepOctNovDec

VI Seminar Homogenization, Budapest 2008 Example of statistic analysis for individual series (1/4) StatisticsJanFebMarAprMayJunJulAugSepOctNovDecTot Ndays MeanCS MeanAWS MinCS MinAWS MaxCS MaxAWS StdDevCS StdDevAWS AssimetryCS AssimetryAWS KurtosisCS KurtosisAWS Bias Tmin.: Cabril

VI Seminar Homogenization, Budapest 2008 Example of statistic analysis for individual series (2/4) Tmax.: Cabril StatisticsJanFebMarAprMayJunJulAugSepOctNovDecTot Ndays MeanCS MeanAWS MinCS MinAWS MaxCS MaxAWS StdDevCS StdDevAWS AssimetryCS AssimetryAWS KurtosisCS KurtosisAWS Bias

VI Seminar Homogenization, Budapest 2008 Example of statistic analysis for individual series (3/4) Tmin: Lisboa StatisticsJanFebMarAprMayJunJulAugSepOctNovDecTot Ndays MeanCS MeanAWS MinCS MinAWS MaxCS MaxAWS StdDevCS StdDevAWS AssimetryCS AssimetryAWS KurtosisCS KurtosisAWS Bias

VI Seminar Homogenization, Budapest 2008 Example of statistic analysis for individual series (4/4) StatisticsJanFebMarAprMayJunJulAugSepOctNovDecTot Ndays MeanCS MeanAWS MinCS MinAWS MaxCS MaxAWS StdDevCS StdDevAWS AssimetryCS AssimetryAWS KurtosisCS KurtosisAWS Bias Tmax: Lisboa

VI Seminar Homogenization, Budapest 2008 Statistical Testing: Total data mean values (AWS-CS) TmaxJanFebMarAprMayJunJulAugSepOctNovDecTotal TminJanFebMarAprMayJunJulAugSepOctNovDecTotal Z- values: two tailed test (significance levels: 10%, 5% and 1%) For each month, results are significant (90%) for most of the stations; For each station results may change between Tmax and Tmin

VI Seminar Homogenization, Budapest 2008 Statistical testing of monthly data differences to normal values Z- values: two tailed test (significance levels: 10%, 5% and 1%) Nr.YM vs. AWS vs. CS AWS vs. CS Nr.YM vs. AWS vs. CS AWS vs. CS Tmin Nr.YM vs. AWS vs. CS AWS vs. CS Nr.YM vs. AWS vs. CS AWS vs. CS Tmax

VI Seminar Homogenization, Budapest 2008 Climatological analysis of extreme values Tmin Nr.M Cold Days (AWS) Tmin<-10ºC Cold Days (CS) Tmin<-10ºC Tropical Night (AWS) Tmin>20ºC Tropical Night (CS) Tmin>20ºCTN10 (ºC) T<TN10 (AWS) T<TN10 (CS)TN90 (ºC) T>TN90 (AWS) T>TN90 (CS)Ndays At Lisboa AWS detects more tropical nights than the CS, the opposite at Cabril

VI Seminar Homogenization, Budapest 2008 Climatological analysis of extreme values At Lisboa AWS detects less warm, summer and tropical days than CS, at Cabril there is seasonal dependancy Tmax Nr.M Warm Days (AWS) Tmax>20º C Summer Days (AWS) Tmax>25º C Summer Days (CS) Tmax>25º C Tropical Day T1 (AWS) Tmax>30º C Tropical Day T1 (CS) Tmax>30º C Tropical Day T1 (AWS) Tmax>35º C Tropical Day T1 (CS) Tmax>35º CTX10 (ºC) T<TX1 0 (AWS) T<TX1 0 (CS) TX90 (ºC) T>TX9 0 (AWS) T>TX9 0 (CS)Ndays

VI Seminar Homogenization, Budapest 2008 Connection with the Project “SIGN”: Signatures of environmental change in the observations of the Geophysical Institutes Recovery of 19th and early 20th century Portuguese historical meteorological data M.Valente,M.Barros,L.Nunes,E.Alves,R.Trigo,E.Pinhal,F.Coelho,M.Mendes,J.Miranda This work presents the joint efforts of the 3 Portuguese Geophysical Institutes (of Lisbon, Oporto and Coimbra) and the Portuguese Meteorology Institute to convert to a digital database the historical meteorology data, recorded since 1856 until 1940 in several publications by the institutes. The different sets of historical data contain monthly, daily and sometimes hourly records of pressure, temperature, precipitation, humidity, wind speed and direction, cloud cover, evaporation & ozone. The published data cover several stations in mainland Portugal, the Azores and Madeira islands and in former Portuguese African and Asian colonies. One of the aims is to use the data to study the changes that have taken place in the historical records during the last 150 years, when the recovered data are joined with the post-1941 data stored in the Meteorology Institute digital database. The other aim is to make the data available to the meteorology community at large. Direct observations of pressure data for Lisbon and for the period were prioritized and have been manually digitized, being later subjected to quality control tests. Digital historical records of Lisbon temperature, relative humidity and precipitation data have been obtained through corrected OCR techniques applied to published hourly or bi-hourly tables. Preliminary digital results are also available for several stations in mainland Portugal, Azores and Madeira. All datasets are subjected to an initial quality control test, to detect wrong values, with more comprehensive tests to be applied at later stages. At the same time, detailed metadata files are being compiled for each station. First analysis results for the digital historical database are available.

VI Seminar Homogenization, Budapest 2008 Final remarks/questions Availability of 10 years of daily data x 30 stations Overlapping data series have been characterized and compared, Some results regarding air temperature have been shown, but many other variables (humidity, pressure, …) have also been analyzed, For Tmax & Tmin 2/3 of stations have bias +/-0.25ºC, for T09 only 1/3 of stations There is a problem with missing data from AWS, which lowers the confidence, Climatological extremes are different if calculated with AWS or CS! For air temperature (well behaved variable, 2 types of inhomogenities were shown: seasonal dependence and offset For most of the stations, conventional observations will stop in a couple of years (only few sites will remain for more years), so, we’ll have to rely on AWS data, Then, most recent “break-point” of the series will be known (CS=>AWS), An homogenization plan is required! First for monthly data and then daily data... Continuation of the SIGN project is desirable IM-Portugal welcomes cooperation in this filed (in relation with COST HOME?)