Mesoscale-Microscale Coupling: a New Time for the Atmospheric Modeling

Mesoscale-Microscale coupling: A new time for the atmospheric modeling

A. Montorn`es th 6 International Conference on Meteorology and Climatology of the Mediterranean 2017

80 m above ground

10 10' mean Wind speed [m/s]

19/12 21/12 23/12 25/12 27/12 29/12 31/12 Time (10')

Vortex−PBL 3 km

Montorn`es et. al, [email protected] Vortex Mesoscale-Microscale coupling: a new time for the atmospheric modeling th 6 International Conference on Meteorology and Climatology of the Mediterranean 2017

80 m above ground

10 10' mean Wind speed [m/s]

19/12 21/12 23/12 25/12 27/12 29/12 31/12 Time (10')

Measured Vortex−PBL 3 km

80 m above ground: 1 year analysis Mesoscale Microscale

1e+02 1y 1m 1d 3h

1e+00

Spectral density [−] Spectral 1e−02 10min

1e−05 1e−03 Frequency (Hz) label Real Vortex−PBL 3 km

Wind resource assessment approach: mesoscale models

Climate 1000 Years Meteorology Fluid dynamics Climate models 100 Years Wind industry Climate variations 10 Years interest

1 Year ENSO Seasonal Global 1 Month Planetary waves models 1 week Mesoscale Synoptic scale models 1 day Mesoscale 1 h

10 min Microscale Characteristic time 1 min

10 s

Small eddies 1 s Molecular diffusion <0.1 s <1 mm 1 cm 10 cm 1 m 10 m 100 m 1 km 10 km 100 km 1.000 km10.000 km100.000 km

Characteristic length Montorn`es et. al, [email protected] Vortex Mesoscale-Microscale coupling: a new time for the atmospheric modeling th 6 International Conference on Meteorology and Climatology of the Mediterranean 2017

A new age: CFD

Climate 1000 Years Meteorology Fluid dynamics Climate models 100 Years Wind industry Climate variations 10 Years interest

1 Year ENSO Seasonal Global 1 Month Planetary waves models 1 week Mesoscale Synoptic scale models 1 day CFD models Mesoscale 1 h

10 min Microscale Characteristic time 1 min

10 s

Small eddies 1 s Molecular diffusion <0.1 s <1 mm 1 cm 10 cm 1 m 10 m 100 m 1 km 10 km 100 km 1.000 km10.000 km100.000 km

Large eddies Small eddies Input Dissipation energy

Energy flux

Large eddies Small eddies

RANS Resolved Modeled

Average

Large eddies Small eddies

RANS Avg.

LES Dyn. DNS Resolved

Dynamic

Large eddies Small eddies

RANS Avg.

LES Dyn.

DNS Dyn.

DNS LES RANS

Dyn. Dyn. Avg.

Adapted from Maries, A., Haque, M. A., Yilmaz, S. L., Nik, M. B., Marai, G. E.: New Developments in the Visualization and Processing of Tensor Fields, Springer, pp. 137156, D. Laidlaw, A. Villanova. 2012

DNS LES RANS

Dyn. Dyn. Avg.

Seconds Minutes Distributions Different tools for different applications

A new age: Mesoscale-Microscale coupling

Climate 1000 Years Meteorology Fluid dynamics Climate models 100 Years Wind industry Climate variations 10 Years interest

1 Year ENSO Seasonal Global 1 Month Planetary waves models 1 week Mesoscale Synoptic scale models 1 day CFD models Mesoscale 1 h

10 min Microscale Characteristic time 1 min

10 s

Small eddies 1 s Molecular diffusion <0.1 s <1 mm 1 cm 10 cm 1 m 10 m 100 m 1 km 10 km 100 km 1.000 km10.000 km100.000 km

Coupling mesoscale-LES: Challenges

◮ Lateral boundary conditions ◮ Surface layer and Land Surface Model ◮ Terra-Incognita

Coupling mesoscale-LES: Challenges

◮ Lateral boundary conditions ◮ Surface layer and Land Surface Model ◮ Terra-Incognita

Lateral boundary conditions

Microscale LES BC

Mesoscale PBL

Lateral boundary conditions

Mesoscale PBL

Lateral boundary conditions

θnew = θold + θ' MuñozEsparza (2014) + Vortex inhouse R+D

Mesoscale PBL

A new age: Mesoscale-Microscale coupling

Climate 1000 Years Meteorology Fluid dynamics Climate models 100 Years Wind industry Climate variations 10 Years interest

1 Year ENSO Seasonal Global 1 Month Planetary waves models 1 week Mesoscale Synoptic scale models 1 day Large Eddy Mesoscale 1 h Simulation

10 min Microscale Characteristic time 1 min SGS

10 s

Small eddies 1 s Molecular diffusion <0.1 s <1 mm 1 cm 10 cm 1 m 10 m 100 m 1 km 10 km 100 km 1.000 km10.000 km100.000 km

Vortex approach

Reanalysis

Vortex WRFLES Mesoscale PBL

Vortex approach

Reanalysis

Vortex WRFLES Mesoscale PBL Postprocessing Microscale LES 4 Hz output 10' averages 3'' gust 10' standard deviation 20150 meters above ground

1year timeseries (10min)

110 m gridsize Wind speed, Wind direction, Region 2.5x2.5 km Temperature, Pressure, Richardson Number, PBL Fluxes, Wind Veer, Inflow angle, ...

80 m above ground

10 10' mean Wind speed [m/s]

19/12 21/12 23/12 25/12 27/12 29/12 31/12 Time (10')

Measured Vortex−PBL 3 km

80 m above ground

10 10' mean Wind speed [m/s]

19/12 21/12 23/12 25/12 27/12 29/12 31/12 Time (10')

Measured Vortex−LES 111 m Vortex−PBL 3 km

80 m above ground: 1 year analysis Mesoscale Microscale

1e+02 1y 1m 1d 3h Terra Incognita 1e+00

Spectral density [−] Spectral 1e−02 10min

1e−05 1e−03 Frequency (Hz) label Real Vortex−LES 111 m Vortex−PBL 3 km

Validation exercise

3.1% Offshore 41.7% Flat terrain 25.0% Complex terrain 30.2% Forest

● ●

● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ●● ● ● ● ●● ● ● ● ● ● ●● ●● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ●

●●

●

● ●

● ● ●

● ● ● ● ● ● ● ● ●

● ●

Anemometers: 18.1% 2050 m 68.1% 50100 m 13.8% 100150 m

1year validation Wind speed validated at 96 sites at different mestmast heights Turbulence validated at 56 sites at different mestmast heights

Montorn`es et. al, [email protected] Vortex Mesoscale-Microscale coupling: a new time for the atmospheric modeling Vortex-LES: white paper

th 6 International Conference on Meteorology and Climatology of the Mediterranean 2017

Wind speed speed: Metrics The validation of the wind speed time-series show a high dependence with the geographical features. Table 1: Summaryof the wind speed metrics

Bias RMSE R2 The analysis of the 96 sites used in this validation (%) (ms-1) shows a mean bias of +2.4% with respect to the mean wind speed (Table 1). This metric is minimum for the 10-min 2.6 0.605 off-shore sites with a +0.4%, followed by the complex All Hourly 2.4 2.5 0.637 terrain with/or forest and the complex terrain with a (100%) mean bias around +0.4% and +0.5%, respectively. The Daily 1.7 0.807 sites located in flat terrain experience a higher 10-min 1.9 0.837 understimation of -3.4% with respect to the mean Off-shore wind speed. Hourly 0.4 1.8 0.855 (3.1%) Daily 1.1 0.937 The study of the RMSE show a mean value for all the sites of 2.6 ms-1 for the 10-min time-series that 10-min 2.5 0.593 decreases slightly when values are hourly Flat Hourly -3.4 2.4 0.625 aggregated. The daily averaged values show a mean (41.7%) RMSE of 1.7 ms-1. By groups, the RMSE is minimum Daily 1.6 0.811 in the off-shore sites with a value of 1.9 ms-1, 1.8 ms-1 and 1.1 ms-1 for the 10-min, hourly and daily scales, 10-min 7.1 0.590 Complex respectively. This metric increases slightly for the Hourly 0.5 7.2 0.623 site located in flat terrain showing a mean RMSE of (25.0%) Daily 7.1 0.790 2.5 ms-1 for the 10-min values and decreasing to 1.6 ms-1 for the daily data-set. The Vortex-LES 10-min 6.6 0.610 simulations in complex terrain or complex terrain Complex Hourly 0.4 6.8 0.646 with/or forest experience higher RMSEs with 7.1 ms- Forest (30.2%) 1 and 6.6 ms-1, respectively. In these cases, the Daily 6.9 0.805 variation of the RMSE with the temporal scale becomes negligible. R2 values are produced in the complex terrain The results of the R2 reveals a similar behavior that with/or forest with 0.610, followed by the flat terrain the one observed for the bias and the RMSE, as sites with 0.593 and the complex terrain simulations expected. The averaged value for all the sites is 0.605 with 0.590. The hourly values reproduce the same Montorn`esfor the 10-min et. time-series, al, [email protected] 0.637 for the hourlydata- pattern. The larger R2 is shown by the complex Vortex Mesoscale-Microscalesets and 0.807 for the daily coupling: simulations. a new Put time in forterrain the atmosphewith/or forestric with modeling 0.646, while the other two plainly, the Vortex-LES simulations can reproduce showvalues around0.623 - 0.625 (Table 1). the around the 60% of the intra-day variability of the site, reaching the 80% for the dailyvalues. The daily values experience a different behavior. The highest R2 is reached by the flat terrain simulation By categories, the off-shore site show the highest R2, while the lowest values are produced by the complex- with 0.837, 0.855 and 0.937 for the different time- terrain sites with 0.811 and 0.790, respectively. The scales. The other groups produce similar results for complex terrain with/or forest is an intermediate the 10-min and hourly data-sets. The larger 10-min case with 0.805.

VORTEX FdC S.L. Parc Tecnologic Barcelona/Carrer Marie Curie 8-1 /08042 Barcelona Spain / vortexfdc.com Vortex-LES: white paper

Table 1: Summaryof the Weibull metrics Off-shore A k Freq. (%) (%) (sm-1) All 10-min2.8 2.3 0.015 (100%)

Off-shore Vortex-LES: white paper 10-min0.3 8.3 0.007 (3.1%)

Flat 10-min-3.2 1.0 0.018 (41.7%)

Complex th 10-min 8.0 7.7 0.014 Table 1: Summaryof the Weibull metrics 6 International Conference(25.0%) on Meteorology and Climatology of the Mediterranean 2017 Vortex-LES:Complex white paper Off-shore Forest 10-min 6.8 -2.9 0.014 (30.2%) Fig ??: Example of a comparison between the Weibull A k Freq. Table 1: Summaryof the Weibull metrics distribution for the measurementsSite030 and Vortex-LES Wind speed: Weibull Off-shore -1 The analysis of the Weibull parameters shows a good for Site018. (%) A k Freq. (%) (sm ) Vortex-LES:agreement white between paper the real and model wind (%) (%) (sm-1) 0.075 distributions. Generally, the Vortex-LES simulations All reproduceTable 1: Summaryof all10-min the wind the Weibull regimes2.8 metrics being 2.3 the calms 0.015 and All extreme(100%) events a high improvements with respect to Off-shore 10-min2.8 2.3 0.015 0.050 the mesoscale simulations.A k Freq. Off-shore -1 (100%) 10-min(%) 0.3(%) (sm 8.3) 0.007 (3.1%) AllThe validation is focused on three parameters: i) the 10-min2.8 2.3 0.015 Probability Density [−] (100%) relativeFlat error in the scale parameter A, relative error 0.025 10-min-3.2 1.0 0.018 Off-shorein the(41.7%) shape parameter k and the mean differences of Off-shore 10-min0.3 8.3 0.007 10-min0.3 8.3 0.007 (3.1%)both distributions (Appendix ?). Table ?? shows a Complex summary of the10-min metrics for8.0 all the sites7.7 and0.014 by the (3.1%) Flat(25.0%) 0.000 different10-min categories-3.2 aforementioned. 1.0 0.018 Figs ?? and ?? (41.7%) 0 10 20 30 showsComplex two samples of the results for the site 18th and Wind speed [m/s] Complex Site018 theForest site10-min 87th. 10-min8.0 6.87.7 0.014-2.9 0.014 Flat terrain Flat (25.0%)(30.2%) Fig ??: Example of a comparison between the Weibull 10-min-3.2 1.0 0.018 Complex distribution for the measurements and Vortex-LES The averaged error for all the sites is +2.8% and 0.075 (41.7%) Forest+2.3%The analysis10-min for A of and the6.8 k, Weibull respectively,-2.9 parameters0.014 while shows the a mean good for Site018. (30.2%) Fig ??: Example of a comparison between the Weibull differenceagreement between between the the real real and andthe modeled modeldistribution wind for the measurements and Vortex-LES -1 distributionsdistributions. is Generally, of around the0.015 Vortex-LES sm . simulationsfor Site018. The analysis of the Weibull parameters shows a good 0.050 Complex reproduce all the windk−1 regimes beingk the calms and 10-min 8.0 7.7 0.014 agreementf(x between)= k thex realexp and− modelx wind distributions.Theextreme off-shore Generally, eventsA A a thesites high Vortex-LES improvements simulate simulationsA the with best respect wind to reproducedisrtributionthe mesoscale all the wind with simulations. regimes a near-zero being the error calms in Aand and a mean (25.0%) Probability Density [−] extremedifference events a high between improvements both distributions with respect to of around 0.025 the mesoscale0.007The validation sm simulations.-1 . The is focused shape parameter on three parameters: is overestimated i) the Complex withanerrorrelative error of8.3%. in the scale parameter A, relative error The validationin the shape is focused parameter on three k and parameters: the mean i) differences the of relative error in the scale parameter A, relative error 0.000 Theboth flat distributions terrain sites (Appendix tendReal to underestimate ?). TableVortex−LES ?? the shows scale a Forest 10-min 6.8 -2.9 0.014 in the shape parameter k and the mean differences of 0 10 20 parametersummary of with the an metrics averaged for error all the of -3.2%. sites and The by shape the Wind speed [m/s] both distributions (Appendix ?). Table ?? shows a Fig ??: Exampleparameterdifferent of categories a iscomparison slightly aforementioned. overestimated Figs with between ?? a and mean ?? the Weibull (30.2%) summary of the metrics for all the sites and by the Fig ??: Example of a comparison between the Weibull Montorn`esdifferent et. al,errorshows [email protected] categories of two 1.0%. samples aforementioned. The of difference the results Figs between for ?? the and site the ?? 18th real and Vortex distribution for the measurements and Vortex-LES distributionshows formodeledthe two site samples the 87th. frequencies of the measurements results shows for thea mean site 18th error and of 0.018 smand- Vortex-LES Mesoscale-Microscale coupling: a new time for the atmospheric modeling Flat terrain the site1 . 87th. for Site087. Flat terrain for Site018. The averaged error for all the sites is +2.8% and The analysis of the Weibull parameters shows a good The averaged+2.3% for error A forand all k, the respectively, sites is +2.8% while and the mean VORTEX FdC S.L. Parc Tecnologic Barcelona/Carrer Marie Curie 8-1 /08042 Barcelona Spain / vortexfdc.com +2.3%difference for A and between k, respectively, the real while and the the mean modeled wind difference between the real and the modeled wind agreement between the real and model wind distributions is of around 0.015 sm-1 . distributions is of around 0.015 sm-1 . distributions. Generally, the Vortex-LES simulations TheThe off-shore off-shore sites simulate sites simulate the best the wind best wind disrtributiondisrtribution with a withnear-zero a near-zero error in A error and a in mean A and a mean reproduce all the wind regimes being the calms and differencedifference between between both distributions both distributions of around of around 0.0070.007 sm-1 . sm The-1 shape. The parameter shape parameter is overestimated is overestimated extreme events a high improvements with respect to withanerrorwithanerror of8.3%. of8.3%. The flat terrain sites tend to underestimate the scale The flat terrain sites tend to underestimate the scale the mesoscale simulations. parameter with an averaged error of -3.2%. The shape parameter with an averaged error of -3.2%. The shape parameter is slightly overestimated with a mean errorparameter of 1.0%. The is difference slightly between overestimated the real withand Fig a mean ??: Example of a comparison between the Weibull Fig ??: Example of a comparison between the Weibull modelederror frequencies of 1.0%. shows The differencea mean error between of 0.018 sm the- realdistribution and for the measurements and Vortex-LES 1 . modeled frequencies shows a mean error of 0.018for Site087. sm- distribution for the measurements and Vortex-LES The validation is focused on three parameters: i) the 1 . for Site087. relative error in the scale parameter A, relative error VORTEX FdC S.L. Parc Tecnologic Barcelona/Carrer Marie Curie 8-1 /08042 Barcelona Spain / vortexfdc.com VORTEX FdC S.L. Parc Tecnologic Barcelona/Carrer Marie Curie 8-1 /08042 Barcelona Spain / vortexfdc.com in the shape parameter k and the mean differences of both distributions (Appendix ?). Table ?? shows a summary of the metrics for all the sites and by the different categories aforementioned. Figs ?? and ?? shows two samples of the results for the site 18th and the site 87th. Flat terrain

The averaged error for all the sites is +2.8% and +2.3% for A and k, respectively, while the mean difference between the real and the modeled wind distributions is of around 0.015 sm-1 .

The off-shore sites simulate the best wind disrtribution with a near-zero error in A and a mean difference between both distributions of around 0.007 sm-1 . The shape parameter is overestimated withanerror of8.3%.

The flat terrain sites tend to underestimate the scale parameter with an averaged error of -3.2%. The shape parameter is slightly overestimated with a mean error of 1.0%. The difference between the real and Fig ??: Example of a comparison between the Weibull modeled frequencies shows a mean error of 0.018 sm- distribution for the measurements and Vortex-LES 1 . for Site087.

VORTEX FdC S.L. Parc Tecnologic Barcelona/Carrer Marie Curie 8-1 /08042 Barcelona Spain / vortexfdc.com Vortex-LES: white paper

Table 1: Summaryof the wind direction metrics Complex terrain Bias MAE (deg) (deg)

All 10-min3 18 (100%)

Off-shore 10-min-2 35 (3.1%)

Flat 10-min0 34 (41.7%)

Complex 10-min2 34 (25.0%) Complex Fig ??: Example of a comparison between the Weibull Forest 10-min10 31 distribution for the measurements and Vortex-LES (30.2%) for Site018. one wind rose sector (Table ?). The preformance is Complex forest similar for the off-shore, flat terrain and complex terrain sites with a mean MAE of around 34º. In the off-shore sites, the wind direction is slightly underestimated with -2º, while the complex terrain sites show a slightly positive bias of +2º. Finally, the flat terrain sites produce a near-zero bias.

The larger mean bias is observed in the complex terrain with/or forest sites with +10º. However, the mean MAE is significantlylower with 31º·

th 6 International Conference on Meteorology and Climatology of the Mediterranean 2017 Vortex-LES: white paper

Table 1: Summaryof the wind direction metrics Complex terrain Wind direction:Bias MetricsMAE and rose Vortex-LES: white paper (deg) (deg) All 10-min3 18 Real Vortex−LES Fig ??: Example of a comparison(100%) between the Weibull Site016 Vortex-LES: white paper Table 1: Summaryof the wind direction metrics N N Complex terrain 16% Wind Speed distribution for the measurements and Vortex-LES (m/s) Off-shore Bias MAE 12% 10-min-2 35 (0,3] Table(3.1%) 1: Summaryof the wind(deg) direction(deg) metrics 8% for Site018. (3,6] Complex terrain 4% FlatAll (6,9] 10-minBias03 MAE 3418 0% W E W E (41.7%)(100%) (9,12] The mean error in the scale parameter(deg) increases(deg) (12,15] Complex significantly in the complexOff-shoreAll terrain10-min and2 forest 34 (15,18] (25.0%) 10-min-23 1835 categories with +8.0% and +6.8%,(100%)(3.1%) respectively. The (18,19.8] behavior for the shape parameterComplex is completly S S Fig ??: Example of a comparison between the Weibull Off-shoreForestFlat 10-min10 31 10-min-20 3534 distribution for the measurements and Vortex-LES (30.2%)(41.7%)(3.1%) different. For complex terrain sites, the k is Real Vortex−LES Site078 for Site018. N N Wind Speed overestimated with an averagedComplex error of +7.7%. In (m/s) Flat 10-min2 34 25% one wind(25.0%) rose sector10-min (Table ?).0 The preformance 34 is contrast, the scale parameter(41.7%) for complex terrain 20% (0,3] Complex forest similarComplex for the off-shore, flat terrain and complex 15% (3,6] 10% Fig ??: Example ofwith/or a comparison forest between thetends Weibull toterrain beComplexunderestimated sites with10-min a mean MAE102 of aroundwith 3431 34º. a In the (6,9] Forest 5% off-shore(25.0%) sites, the wind direction is slightly (9,12] distribution formean the measurements error of and around Vortex-LES -2.9%.(30.2%) Both categories show a 0% W E W E underestimated with -2º, while the complex terrain (12,15] for Site018. similar mean error in the distributionsComplex of 0.014 sm-1 . Fig ??: Example of a comparison between the Weibull sites showForest a slightly10-min positive bias10 of +2º. 31 Finally, the (15,18] one wind rose sector (Table ?). The preformance is distribution for the measurements and Vortex-LES flat terrain(30.2%) sites produce a near-zero bias. (18,21] Complex forest similar for the off-shore, flat terrain and complex for Site018. (21,24] terrain sites with a mean MAE of around 34º. In the The validation of theThe wind larger meandirection bias is observed shows in the an complex S S (24,25.6] terrainoneoff-shore wind with/or rose sites, sector forest the (Table sites wind with ?). direction The +10º. preformance However, is slightly the is averaged biasComplex of 3º forest and ameansimilarunderestimated mean MAE for isMAE the significantlylower off-shore,with of -2º, 18º, while flat less terrainwith the complex31º· than and complex terrainFig ??: Example ofthe windroses for two sites. terrainsites show sites a slightlywith a mean positive MAE bias of ofaround +2º. Finally, 34º. In the off-shoreflat terrain sites,sites produce the wind a near-zero direction bias. is slightly VORTEX FdC S.L. Parc Tecnologicunderestimated Barcelona/Carrer with -2º, Marie while Curie the complex 8-1 /08042 terrain Barcelona Spain / vortexfdc.com sitesThe larger show a mean slightly bias positive is observed bias of in+2º. the Finally, complex the flatterrain terrain with/or sites forest produce sites a near-zero with +10º. bias. However, the mean MAE is significantlylower with 31º· Fig ??: Example of a comparison between the Weibull Montorn`esThe et. al,larger [email protected] mean bias is observed in the complex Vortex distribution for the measurements and Vortex-LES terrain with/or forest sites with +10º. However, the for Site018. Mesoscale-Microscalemean MAE is significantlylowercoupling: a new with time 31º· for the atmospheric modeling

The mean error in the scale parameter increases significantly in the complex terrain and forest categoriesFig ??: Example with of +8.0% a comparison and +6.8%, between respectively. the Weibull The behavior for the shape parameter is completly distribution for the measurements and Vortex-LES different. For complex terrain sites, the k is overestimatedfor Site018. with an averaged error of +7.7%. In Fig ??: Example of a comparison between the Weibull contrast, the scale parameter for complex terrain distributionThe mean error for the in measurementsthe scale parameter and Vortex-LES increases with/orsignificantly forest in tends the to complex be underestimated terrain and with forest a for Site018. meancategories error with of around +8.0% -2.9%. and +6.8%, Both categories respectively. show The a similar mean error in the distributions of 0.014 sm-1 . Thebehavior mean for error the in shape the scale parameter parameter is completly increases significantlydifferent. For in complex the complex terrain terrain sites, and the forest k is categoriesTheoverestimated validation with with +8.0% of thean and averaged wind +6.8%, direction error respectively. of shows +7.7%. The an In behavioraveragedcontrast, bias the for of scale the 3º and shape parameter a mean parameter MAE for complex of 18º, is completly less terrain than Fig ??: Example ofthe windroses for two sites. different.with/or forest For tends complex to be terrain underestimated sites, the with k is a VORTEXoverestimatedmeanerror FdC S.L. ofParcaround with Tecnologic an -2.9%. averaged Barcelona/Carrer Both error categories Marie of +7.7%. Curie show 8-1 In/08042 a Barcelona Spain / vortexfdc.com contrast,similar mean the error scale in parameter the distributions for complex of 0.014 terrain sm-1 . with/or forest tends to be underestimated with a meanThe validation error of around of the -2.9%. wind Both direction categories shows show an a similaraveraged mean bias error of 3º in and the a distributions mean MAE of of 18º, 0.014 less sm than-1 . Fig ??: Example ofthe windroses for two sites.

VORTEXThe validation FdC S.L. Parc Tecnologicof the Barcelona/Carrer wind direction Marie shows Curie 8-1 an /08042 Barcelona Spain / vortexfdc.com averaged bias of 3º and a mean MAE of 18º, less than Fig ??: Example ofthe windroses for two sites.

VORTEX FdC S.L. Parc Tecnologic Barcelona/Carrer Marie Curie 8-1 /08042 Barcelona Spain / vortexfdc.com Vortex-LES: white paper

th 6 International Conference on Meteorology and Climatology of the Mediterranean 2017 a) Complex-terrain b) Off-shore -0.4% Wind speed: VerticalShear (real): proﬁles 0.224 Shear (real): 0.103 Vortex-LES: white paperShear (model): 0.195 Shear (model): 0.075 +0.5% Shear error: -12.7% Shear error: -27.8% a) Complex-terrain b) Off-shore Site028 Site030 +0.8% 100 ●● -0.4% 100 Shear (real): 0.224 Shear (real): 0.103 Shear (model): 0.195 Shear (model): 0.075 ● ● +0.5% +0.5% Shear error: -12.7% Shear error: -27.8% 90 -0.4% 80 ● ● +0.8% +0.4%

● ● +0.5%

80 ●● -0.4% Height [m] Height [m] 60 ● ● +0.4% +0.5%

70 ● ● +0.5% +2.6%

40 ● ● +2.6% +0.9% +2.9% 60 ● ● +0.9% ● ● +2.9%

7 8 9 8 9 10 Wind speed [m/s] Wind speed [m/s]

● Real ● Vortex−LES Fig. 2: Analysis of the wind profile for a site in complex-terrain (left-hand plot) and an off-shore site (right- hand plot). TheFig. numbers 2: Analysis indicate the of anual the bias wind at each profile height normalized for a site with in respect complex-terra to the real meanin wind (left-hand plot) and an off-shore site (right- speed. hand plot). The numbers indicate the anual bias at each height normalized with respect to the real mean wind speed. Generally, the number of sites with measurements at An example of this fit is ilustrated with two examples more than one height is reduced or the customers do in complex-terrain and in off-shore in Fig. ??. not share the data-sets with us. Consequently, the Generally, the number of sitesAll with sites showmeasurements a similar behavior, at theAn bias example tend to be of this fit is ilustrated with two examples Montorn`esanalysis et. al, of [email protected] the vertical profiles is reduced to 7 of the positive near to the surface drifting to near-zero of Vortex 96 sites used in this validation. Particularly, these Mesoscale-Microscale coupling:more thana new time one for height the atmosphe is reducedric modelingslightlynegative or the customers values at the upperdo levels.in complex-terrain and in off-shore in Fig. ??. sites are Site002, Site007, Site028, Site030, Site038, Site045 and Site050not (Tableshare ??). the data-sets with us. Consequently, the All sites show a similar behavior, the bias tend to be analysis of the vertical profilesThe is physical reduced reason to 7 of of these the patterns is in the surface features. The first levelspositive of the realnear to the surface drifting to near-zero of Table 1: Summaryof the vertical levels used for the 96 sites used in this validation.atmosphere Particularly, is highly perturbed these on the local features validation of the vertical profiles slightlynegative values at the upper levels. sites are Site002, Site007, Site028,(trees, buildings Site030, or waves, Site038, among others), that cannot Site Site045Heights and with available Site050 real (Table ??).be resolved by the Vortex-LES at 100 m. The model measurements parameterizes these effects by assumingThe physical some reason of these patterns is in the preevaluated roughnesses, enough for mesoscale Site002 10,20,40,80m surface features. The first levels of the real Table 1: Summaryof the verticalapplications levels but used more for important the in the case of Site007 20,30,40,50,60m microscale simulations. Consequently,atmosphere the model is highly perturbed on the local features Site028 60, 80,validation 100 m of the vertical profiles show a tendency to overestimate the(trees, lower buildings levels. or waves, among others), that cannot Site030 33,40,50,60,70,80,90,100m Site Heights withNevertheless, available thesereal levels have a minorbe impact resolved in the by the Vortex-LES at 100 m. The model Site038 20,30,50,70m wind industry and it is an issue that will be solved Site045 40, 50, 60 m measurementsduring the following years, as thisparameterizes new approach these effects by assuming some Site050 30,35,50,100m penetrates to the market. preevaluated roughnesses, enough for mesoscale Site002 10,20,40,80m applications but more important in the case of The analysis of the verticalSite007 profiles is performed20,30,40,50,60m by The higher levels (i.e. between 50 and 100 m above comparing the real and modeled wind shears based the ground) are less affected bymicroscale the surface simulations. Consequently, the model onalogfit,as Site028 60, 80,characteristics, 100 m and thus, the bias decreases significantly. show a tendency to overestimate the lower levels. Site030 33,40,50,60,70,80,90,100m Nevertheless, these levels have a minor impact in the Site038 20,30,50,70mBelow 50 m, the bias fluctuates betweenwind +5 and industry +20% and it is an issue that will be solved with respect to the mean wind speed, while between being V1 and V2 the anualSite045 mean wind speed at two40, 50, 60 m during the following years, as this new approach different heights, z1 and z2, respectively, and the 50 and 100 m the bias ranges between -5% and +5% , shear coefficient. Site050 30,35,50,100malthough in some cases reaches valuespenetrates around the to the market. 10%.

VORTEX FdC S.L. ParcThe Tecnologic analysis Barcelona/Carrer of the Marie vertical Curie 8-1 /08042 profiles Barcelona Spain is /performedvortexfdc.com by The higher levels (i.e. between 50 and 100 m above comparing the real and modeled wind shears based the ground) are less affected by the surface onalogfit,as characteristics, and thus, the bias decreases significantly.

Below 50 m, the bias fluctuates between +5 and +20% being V1 and V2 the anual mean wind speed at two with respect to the mean wind speed, while between different heights, z1 and z2, respectively, and the 50 and 100 m the bias ranges between -5% and +5% , shear coefficient. although in some cases reaches values around the 10%.

VORTEX FdC S.L. Parc Tecnologic Barcelona/Carrer Marie Curie 8-1 /08042 Barcelona Spain / vortexfdc.com th 6 International Conference on Meteorology and Climatology of the Mediterranean 2017

Wind standardTable 1: Summaryof deviation: the wind direction metrics Metrics Bias RMSE R2 (ms-1) (ms-1) (-)

All -0.05 0.5 0.278 (100%) mirceavoda.100

Rainfed croplands -0.1 0.4 0.223 (16.4%) 0.8

Cropland (50-70%) Vegetation (20-50%) 0.0 0.5 0.171 (10.9%) 0.7

Vegetation (50-70%)

Cropland (20-50%) 0.0 0.4 0.251 SD [m/s] (10.9%) 0.6

Deciduous forest (40%, >5m) -0.1 0.5 0.261 0.5 (9.1%)

Evergreen forest (40%, >5m) -0.1 0.6 0.309 0 500 1000 1500 2000 (10.9%) Time

Forest (50-70%) lab Real Vortex−LES Grassland (20-50%) -0.1 0.5 0.416 (7.3%)

Grassland (50-70%) Forest (20-50%) -0.1 0.6 0.260 (1.8%)

Shrubland (>15%, <5 m) 0.0 0.4 0.271 (1.8%) Montorn`es et. al, [email protected] Vortex Herbaceous Mesoscale-Microscalevegetation coupling: (>15%) a new0.0 time for0.4 the atmosphe0.246 ric modeling (7.3%)

Sparse vegetation (<15%) -0.1 0.6 0.256 (7.3%)

Grassland, woody, flooded (>15%) 0.0 0.4 0.372 (1.8%)

Bared areas (9.1%) 0.0 0.3 0.261

Off-shore (5.5%) 0.0 0.4 0.321

VORTEX FdC S.L. Parc Tecnologic Barcelona/Carrer Marie Curie 8-1 /08042 Barcelona Spain / vortexfdc.com th 6 International Conference on Meteorology and Climatology of the Mediterranean 2017

Turbulence intensity Site008 Site026

30 30

20 20

Turbulence intensity [%] Turbulence 10 intensity [%] Turbulence 10

0 0

5 10 15 20 25 5 10 15 20 25 10−min wind speed [ms−1] 10−min wind speed [ms−1]

= 100 δws TI Real Vortex−LES

Turbulence Intensity

TI(%) validated at 58 sites

Which metric to use? 1. MAE between TI-model against TI-obs weighted by bin-ocurrence 2. MAE at 15 m/s bin

Average Std Dev MAE 1.8 0.9 MAE-15 1.9 1.1

Montorn`es et. al, [email protected] Vortex Mesoscale-Microscale coupling: a new time for the atmospheric modeling Mesoscale-Microscale coupling: A new time for the atmospheric modeling

A. Montorn`es