A Global Daily Gauge-based Precipitation Analysis, Part I- .ppt

《A Global Daily Gauge-based Precipitation Analysis, Part I- .ppt》由会员分享，可在线阅读，更多相关《A Global Daily Gauge-based Precipitation Analysis, Part I- .ppt（18页珍藏版）》请在麦多课文档分享上搜索。

1、A Global Daily Gauge-based Precipitation Analysis, Part I: Assessing Objective Techniques,Mingyue Chen & CPC Precipitation Working Group CPC/NCEP/NOAA,The 32th Annual Climate Diagnostics & Prediction Workshop Oct. 22-26, 2007, Tallahassee, FL,Background,Problems with current CPC precipitation analys

2、esDue to historical reasons, the current precipitation analyses at CPC do not take advantage of all available gauge and satellite data and present inconsistencies among various products (e.g. global analysis does not match with regional analyses);A project is under way at CPC to generate a unified p

3、recipitation analyses with improved quality and consistent quantity;The first step of the project is to construct a unified analysis of gauge-based daily precipitation over global land;To do this, we have to To unify daily gauge observation reports available at various places of CPC; and To select a

4、n objective analysis technique to define the gauge-based analyses of daily precipitation; To produce the analysis for an extended period and on a real-time basis;Objective of this talk is to report the algorithm selection and data construction that we have done;,Strategy of Algorithm Selection,Asses

5、s the performance of three widely used gauge interpolation techniques and pick up the one with the best statisticsThe three gauge interpolation algorithms: Cressman (1959) Distance weighting Used to generate current regional analyses over US-Mexico and S. America Shepard (1968) Distance weighting Us

6、ed to generate GTS-based daily analysis over global land Optimal Interpolation (OI) of Gandin (1965) Implementation of Xie et al. (2007) Interpolation of ratio of daily total to daily clim with orographic adjustments Used to generate regional gauge analysis over East Asia,Unified Daily Gauge Data,De

7、nse gauge networks from special CPC collections over US, Mexico, and S. America;GTS gauge network elsewhereDaily reports available from 17,000 stations,Daily Prcp Analyses for Jan.5, 2005,Precipitation Analyses: Similar patterns, larger raining areas and smoother distributions in Cressman,Correlatio

8、n between Daily Analyses,Calculated for 2005;Very high correlation between OI and Shepard;Less desirable correlation between Cressman and other analyses over areas covered by less gauges;,Cross-Validation Tests,Withdraw daily precipitation reports at 10% stations selected randomly;Define the analyze

9、d values of precipitation at the 10% withdrawn stations by interpolating gauge reports at the remaining 90% of the stations;Repeat this process for 10 times so that each station is withdrawn once;Compare the analyzed values with the withdrawn station observations to assess the performance of the alg

10、orithms,Performance for Different Regions,Cross-Validation Tests Results for 2005OI presents the highest correlation and small bias over most regions,Histograms of Rainfall Intensity,All three sets of analyses present lower frequencies for no-rain strong rainfall events compared to gauge observation

11、sCressman yields substantially reduced / inflated frequencies for no-rain / light rain events,Gauge Network Density Impacts Tests,To examine how the objective techniques perform in interpolating station reports from gauge networks of different densities;Select the CONUS as our test region for the av

12、ailability of a very dense;Create gauge-based analysis of daily precipitation with a subset of all available gauge data and compare the analysis with original station data,Correlation & Bias,Quality of the gauge analysis degrades as gauge network density becomes sparseOI performs the best withthe hi

13、ghest correlation and the smallest biasesin most cases,Histograms of Rainfall Intensity,100% Network,10% Network,1% Network,Cressman spreads raining area substantially with sparse networksOI reproduces the PDF very well,Construction of the Historical Analysis,OI is selected to construct the gauge-ba

14、sed daily precipitation analysis over the global land areasQuality control performed for the daily station reports;Historical analysis created for a 28-year period from 1979 to 2006;Time series of numbers of gauge reports available,Example for January 8, 1998,Structure of precipitation well depicted

15、 over various parts of the global land areas,Comparison with Existing Analysis,Comparison with CPC existing regional analysis over US for January 8, 1998;Existing analysis is created using the Cressman method;The new analysis presents finer structure in better agreements with station data;,Existing,

16、New,Comparison with Existing Analysis,Comparison with CPC existing regional analysis over S. America for January 8, 1998;,Existing,New,Station,Comparison of Area Mean Prcp,Time series of mean precipitation over NW US 35N-45N; 110W- 120W;Green: Existing analysisRed: New analysis Close agreements in t

17、ime series;New analysis shows slightly larger precipitation due to improved analysis method,Existing,New,Summary,Performance of three widely used gauge interpolation methods has been examined through cross-validation tests and gauge network density impact tests.Based on the assessments, the Optimal

18、Interpolation (OI) algorithm is selected to define our gauge-based analysis of daily precipitation.Quality control (QC) is performed for daily gauge precipitation reports.Gauge-based analyses have been created by interpolating the QCed station data using the OI algorithm for an extended period from 1979 to 2006.Preliminary comparisons showed improved quality of our new analysis compared to existing CPC analyses.Further work is underway to check the new analysis and to apply it for various climate studies.,