Vert Vertical ical wi wind nd pro profil file: e: Assessme - - PowerPoint PPT Presentation

Vert Vertical ical wi wind nd pro profil file: e: Assessme Assessment nt of reg

• f regiona

ional l re rean analyses alyses Comparing reanalyses with tower measurements and vertically extrapolated wind

Christopher Frank1,2, Bernhard Pospichal2, Susanne Crewell2

1 Hans-Ertel-Centre for Weather Research – Climate Monitoring Branch 2 Institute for Geophysics and Meteorology, University Cologne

ISRR, Bonn, 17 July 2018

Motivation

ISRR 2018

• Wind is often highly variable in space and time  Gridded wind speed

information are important for: e.g. public safety, traffic, renewable energy...

• Site assessment for wind turbines needs best possible wind speed

information (on various temporal and spatial scales)

• State of the art assessments are often based on 1 year measurements

taken significantly below hub-height. Hub-height wind speed is then estimated by vertical extrapolation.  Error-prone, time consuming, and expensive

• Consistent and reliable gridded wind source could revolutionize site

assessment Central question: How do reanalyses perform compared to vertically extrapolated wind measurements and what are advantages of new regional reanalyses?

• Glob. reanalyses

COSMO-REA6 Europe ERA-INTERIM foreced 1995 - 2015 40 levels 6 km COSMO-REA2 Germany COSMO-REA6 forced 2007 - 2013 50 levels 2 km

Bollmeyer et al. (2014) REA2 REA2 specifications: Convection-permitting, Latent-heat nudging

ISRR 2018

Regional reanalyses

Era-Interim ~80 km 60 levels Often used in climate studies MERRA-2 ~50 km 72 hybrid-eta levels Often used in renewable energy studies

Tower measurements (Reference)

ISRR 2018

Bollmeyer et al. (2014)

REA2

Cabauw Billwerder Karlsruhe Falkenberg

• 4 towers
• 10 min averages
• 2007-2013

Available height levels of wind speed from tower and reanalyses.

Karlsruhe Falkenberg Billwerder Cabauw

Extrapolation methods for surface 10m wind

ISRR 2018

Three methods from simple to complex:

• 1. 𝑄𝑀-𝑑𝑝𝑜𝑡𝑢: Constant exponent 𝛽 = 0.2
• 2. 𝑄𝑀-𝐼𝑇: Stability and roughness dependent exponent:

𝑑0, 𝑑1, 𝑑2 are dependent on Pasquill-Gifford-SRDT stability classes

• 3. 𝑄𝑀-2𝑀: 2 level based exponents:

Power-law based extrapolation: The extrapolations are based on 10 and 40m measurements, except for Karlsruhe where a local forest forces the reference heights to 40 and 60m.

Marginal distribution

ISRR 2018

• PL_const and PL_HS show underestimation
• Distribution of PL_2L similar to that of the reanalyses

Billwerder 2007-2013 Falkenberg 2007-2013 Wind speed Wind speed

Bias

ISRR 2018

• Roughness

problems in Karlsruhe (regional rea‘s show advantage)

• PL_2L provides

best results at low roughness sites

Cabauw Karlsruhe Billwerder Falkenberg

Mean absolute error (BC_MAE)

ISRR 2018

Cabauw Karlsruhe

• Reanalyses

perform better than extrapolations in ~100 m

• PL_2L shows best

quality – but needs the seconds measuring height.

Billwerder Falkenberg

Stability condition

ISRR 2018

Extrapolation methods are stability dependent  Based on the measured temperature gradient we define stabile, unstable, and neutral classes Cabauw

MAE under stable conditions

• Accuracy of

extrapolated wind decreases fast even faster than for all conditions.

• Especially in low

heights regional reanalyses perform better.

ISRR 2018

Cabauw Billwerder Karlsruhe Falkenberg

ISRR 2018

MAE under unstable conditions

• Extrapolations,

and also reanalyses perform better under unstable conditions.

• No systematically

working best extrapolation method.

Billwerder Falkenberg Karlsruhe Cabauw

Power estimate uncertainty

Regional reanalyses show smaller site dependence than the global ones, and the extrapolation methods. Errors are smaller in lower heights and less than about 25% above 50m.

ISRR 2018

Cabauw Billwerder Karlsruhe Falkenberg

Conclusion Embedding this study

ISRR 2018

• COSMO-REA6 and COSMO-REA2 outperformed the global reanalyses

(Especially close to ground).

• Under stable and neutral conditions
• If only 10m wind extrapolations exist, reanalyses perform better in heights

above 60m

• If two measurement heights exist, reanalyses perform better in about 3

times above the higher reference height

• In well-mixed conditions no systematic was found.

This work is embedded in a broader study to exploit regional reanalyses for renewable energy purposes

• Solar radiation

(Frank et al., Solar Energy 2018)

• wind

(current work, will also be submitted soon)

• Compund and counter-acting event analyses