comparison between climate forecast system reanalysis ... · deeper analysis - tocantins basin...

Comparison between Climate Forecast SystemReanalysis (CFSR) weather data and data frommeteorological stations in Brazil to evaluate the

suitability of CFSR data for SWAT

Jose A.F. Monteiro1,2 Bjorn Gucker1

Raghavan Srinivasan2

1Department of Biosystems Engineering, Federal University of Sao Joao del-Rei, MG, BR2Spatial Analysis Laboratory, Texas A & M University, TX, USA

SWAT Conference, 31.07.2014

1

Hydrological Modeling - Input Data Challenges

Input DataI Static data

– Topography (DEM)– Soil type– Soil cover (natural vegetation vs. land use)– (River map)

I Dynamic data (weather data)– Temperature– Precipitation– (Solar radiation)– (Relative humidity)– (Wind speed)

– Impossible to reassess!

2

Hydrological Modeling - Input Data Challenges

Input DataI Static data

– Topography (DEM)– Soil type– Soil cover (natural vegetation vs. land use)– (River map)

I Dynamic data (weather data)– Temperature– Precipitation– (Solar radiation)– (Relative humidity)– (Wind speed)

– Impossible to reassess!

2

Hydrological Modeling - Input Data Challenges

Input DataI Static data

– Topography (DEM)– Soil type– Soil cover (natural vegetation vs. land use)– (River map)

I Dynamic data (weather data)– Temperature– Precipitation– (Solar radiation)– (Relative humidity)– (Wind speed)– Impossible to reassess!

2

Alternatives to observed weather data

Name Organization Spatial res. Temporal res. PeriodCRU University of

East Anglia0.5◦ Monthly 1901–2012

MERRA NASAGMAO

∼0.5◦ Hourly 1979–present

CFSR NCEP 0.5◦ 6 hourly 1979–2010

CFSR

I First climate reanalysis that includes atmosphere and ocean

I More accurate representation of observed mean precipitation intropical regions (Wang et al. 2011, Clim Dyn)

3

CFSR vs. INMET observations in Brazil

Comparing CFSR to observedweather data

I INMET observed data– 209 stations: 1980–2010– (15 in the Tocantins Basin)– Uneven distribution

I CFSR for Brazil– 12,201 ’stations’– (642 in the Tocantins Basin)

4

Meta-analysis CFSR

Geographic Consistency - Monthly Precipitation

5

Meta-analysis CFSR

Geographic Consistency - Maximum Daily Temperature

6

Meta-analysis CFSR

Geographic Consistency - Minimum Daily Temperature

7

Meta-analysis CFSR

Daily Anomalies: Frequency Distribution

Anomaly = CFSR − INMET

I INMET stations paired withclosest CFSR station

I Frequency distribution µ

I Frequency distribution SD

8

Meta-analysis CFSR

Daily Anomalies: Linear Regressions

Anomaly = CFSR − INMET

I Anomaly time series → linearregression

I f (x) = a + bx

9

Deeper Analysis - Tocantins Basin

Annual Precipitation Time-Series

INMET vs. CFSR

10

Deeper Analysis - Tocantins Basin

Annual Precipitation Time-Series

INMET vs. CFSRINMET vs. Int. 1

I CF = CFSRINMET

CFSRCF = Interpolation

I Corrected by meansof monthly ratios(months as factors,i.e. Januaries,Februaries, etc.)

10

Deeper Analysis - Tocantins Basin

Annual Precipitation Time-Series

INMET vs. CFSRINMET vs. Int. 1INMET vs. Int. 2

I Corrected by meansof monthly ratios(months as timeseries, i.e. January1980, February 1980,etc.)

10

Deeper Analysis - Tocantins Basin

Annual Precipitation Time-Series

INMET vs. CFSRINMET vs. Int. 1INMET vs. Int. 2INMET vs. Int. C

I Combination of bothprevious methods.

I Whenever possible,use of Int. 1

I Whenever INMETlack, use of Int. 2

10

Deeper Analysis - Tocantins Basin

Applying the Correction Factor

11

Deeper Analysis - Tocantins Basin

Interpolated Data Set Used in SWAT

12

Conclusion

I In the broad picture, CFSR represented Brazilian weather and istherefore a good alternative to observed data, especially forlarge-scale projects.

I Nevertheless, we advice a comparison with observed data previous tothe use of CFSR for hydrologic modeling, and, if necessary, applycorrection methods.

13

Acknowledgments

I Swiss National Science Foundation

I Jaclyn Tech, sending CFSR data

I ANA’s personal, sending flow data

Thank you for your time!

14