Homogenization of daily peak wind gust series from Spain and - - PowerPoint PPT Presentation

Homogenization of daily peak wind gust series from Spain and Portugal

José A. Guijarro1, Cesar Azorin-Molina2

1State Meteorological Agency (AEMET), Palma de Mallorca, Spain 2Instituto Pirenaico de Ecología (IPE-CSIC), Zaragoza, Spain

EUMETNET Data Management Workshop

• St. Gallen, Switzerland, 28-30 October 2015

Outline

Introduction Homogenization strategies Impact on extreme wind indexes Conclusions

Introduction

◮ Homogenization of daily series is difficult, due to their

lower noise/signal ratio.

◮ Yet the study of the variability of extreme weather events

requires homogeneous and quality controlled daily series.

◮ Here we apply different strategies to homogenize daily

maximum gust speeds from Portugal and Spain, and analyze their impact on the evaluation of the trends of mean and maximum gusts, the number of days over the 90 percentile and maximum expected gusts for return periods

• f 50, 100 and 200 years.

◮ Question:

Do we really need to homogenize the daily series?

Introduction

◮ Homogenization of daily series is difficult, due to their

lower noise/signal ratio.

◮ Yet the study of the variability of extreme weather events

requires homogeneous and quality controlled daily series.

◮ Here we apply different strategies to homogenize daily

maximum gust speeds from Portugal and Spain, and analyze their impact on the evaluation of the trends of mean and maximum gusts, the number of days over the 90 percentile and maximum expected gusts for return periods

• f 50, 100 and 200 years.

◮ Question:

Do we really need to homogenize the daily series?

Introduction

◮ Homogenization of daily series is difficult, due to their

lower noise/signal ratio.

◮ Yet the study of the variability of extreme weather events

requires homogeneous and quality controlled daily series.

◮ Here we apply different strategies to homogenize daily

maximum gust speeds from Portugal and Spain, and analyze their impact on the evaluation of the trends of mean and maximum gusts, the number of days over the 90 percentile and maximum expected gusts for return periods

• f 50, 100 and 200 years.

◮ Question:

Do we really need to homogenize the daily series?

Introduction

◮ Homogenization of daily series is difficult, due to their

lower noise/signal ratio.

◮ Yet the study of the variability of extreme weather events

requires homogeneous and quality controlled daily series.

◮ Here we apply different strategies to homogenize daily

maximum gust speeds from Portugal and Spain, and analyze their impact on the evaluation of the trends of mean and maximum gusts, the number of days over the 90 percentile and maximum expected gusts for return periods

• f 50, 100 and 200 years.

◮ Question:

Do we really need to homogenize the daily series?

Methodology

◮ The data set consisted of 80 series (7 Portuguese and 73

Spanish) of daily maximum peak wind gusts spanning 54 years (1961-2014).

◮ Corresponding daily series from MM5 simulations at 10 km

resolution were available until 2007 (Murcia University).

◮ Homogenization was performed with Climatol 2.2

(multiplicative model) on:

◮ Average monthly values, using MM5 series as references

when available, and adjusting the daily series with interpolated monthly correction factors.

◮ Direct homogenization of daily values, using MM5 series as

references when available.

◮ Direct homogenization of daily values, without MM5

references.

◮ Annual values of maximum and average wind peak gusts

and number of days over the 90 percentile.

Methodology

◮ The data set consisted of 80 series (7 Portuguese and 73

Spanish) of daily maximum peak wind gusts spanning 54 years (1961-2014).

◮ Corresponding daily series from MM5 simulations at 10 km

resolution were available until 2007 (Murcia University).

◮ Homogenization was performed with Climatol 2.2

(multiplicative model) on:

◮ Average monthly values, using MM5 series as references

when available, and adjusting the daily series with interpolated monthly correction factors.

◮ Direct homogenization of daily values, using MM5 series as

references when available.

◮ Direct homogenization of daily values, without MM5

references.

◮ Annual values of maximum and average wind peak gusts

and number of days over the 90 percentile.

Methodology

◮ The data set consisted of 80 series (7 Portuguese and 73

Spanish) of daily maximum peak wind gusts spanning 54 years (1961-2014).

◮ Corresponding daily series from MM5 simulations at 10 km

resolution were available until 2007 (Murcia University).

◮ Homogenization was performed with Climatol 2.2

(multiplicative model) on:

◮ Average monthly values, using MM5 series as references

when available, and adjusting the daily series with interpolated monthly correction factors.

◮ Direct homogenization of daily values, using MM5 series as

references when available.

◮ Direct homogenization of daily values, without MM5

references.

◮ Annual values of maximum and average wind peak gusts

and number of days over the 90 percentile.

Methodology

◮ The data set consisted of 80 series (7 Portuguese and 73

Spanish) of daily maximum peak wind gusts spanning 54 years (1961-2014).

◮ Corresponding daily series from MM5 simulations at 10 km

resolution were available until 2007 (Murcia University).

◮ Homogenization was performed with Climatol 2.2

(multiplicative model) on:

◮ Average monthly values, using MM5 series as references

when available, and adjusting the daily series with interpolated monthly correction factors.

◮ Direct homogenization of daily values, using MM5 series as

references when available.

◮ Direct homogenization of daily values, without MM5

references.

◮ Annual values of maximum and average wind peak gusts

and number of days over the 90 percentile.

Methodology

◮ The data set consisted of 80 series (7 Portuguese and 73

Spanish) of daily maximum peak wind gusts spanning 54 years (1961-2014).

◮ Corresponding daily series from MM5 simulations at 10 km

resolution were available until 2007 (Murcia University).

◮ Homogenization was performed with Climatol 2.2

(multiplicative model) on:

◮ Average monthly values, using MM5 series as references

when available, and adjusting the daily series with interpolated monthly correction factors.

◮ Direct homogenization of daily values, using MM5 series as

references when available.

◮ Direct homogenization of daily values, without MM5

references.

◮ Annual values of maximum and average wind peak gusts

and number of days over the 90 percentile.

Methodology

◮ The data set consisted of 80 series (7 Portuguese and 73

Spanish) of daily maximum peak wind gusts spanning 54 years (1961-2014).

◮ Corresponding daily series from MM5 simulations at 10 km

resolution were available until 2007 (Murcia University).

◮ Homogenization was performed with Climatol 2.2

(multiplicative model) on:

◮ Average monthly values, using MM5 series as references

when available, and adjusting the daily series with interpolated monthly correction factors.

◮ Direct homogenization of daily values, using MM5 series as

references when available.

◮ Direct homogenization of daily values, without MM5

references.

◮ Annual values of maximum and average wind peak gusts

and number of days over the 90 percentile.

Methodology

◮ The data set consisted of 80 series (7 Portuguese and 73

Spanish) of daily maximum peak wind gusts spanning 54 years (1961-2014).

◮ Corresponding daily series from MM5 simulations at 10 km

resolution were available until 2007 (Murcia University).

◮ Homogenization was performed with Climatol 2.2

(multiplicative model) on:

◮ Average monthly values, using MM5 series as references

when available, and adjusting the daily series with interpolated monthly correction factors.

◮ Direct homogenization of daily values, using MM5 series as

references when available.

◮ Direct homogenization of daily values, without MM5

references.

◮ Annual values of maximum and average wind peak gusts

and number of days over the 90 percentile.

Station locations

−8 −6 −4 −2 2 4 36 38 40 42 44

VX station locations (5 clusters)

Longitude (deg) Latitude (deg) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4849 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80

Data availability

Data availability

1960 1970 1980 1990 2000 2010 20 40 60 80

• Nr. of VX−d data in all stations

Dates

• Nr. of data

Regression observations vs MM5

• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ● ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ● ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ● ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• 5

10 15 20 5 10 15 20 25 30 35

Zaragoza (1961−2007)

MM5 peak gusts (m/s) Observed peak gusts (m/s) Y = 1.10 + 1.39 X r = 0.72 Y = X

Correlations observations vs MM5

Correlations between observed and MM5 series

r Frequency 0.0 0.2 0.4 0.6 0.8 10 20 30

Inhomogeneities

1960 1970 1980 1990 2000 2010 −4 −2 2 4

VX−d at 2614(26), ZAMORA

Dates Standardized anomalies (observed − computed) 1 10 100 min.d. (km) 454 3883

Shift

1960 1970 1980 1990 2000 2010 −4 −2 2 4

VX−d at P535(75), LISBOA GEOFÍSICO

Dates Standardized anomalies (observed − computed) 1 10 100 min.d. (km) 189 4574

Trend

1960 1970 1980 1990 2000 2010 −4 −2 2 4

VX−d at B278(71), PALMA DE MALLORCA/SON SAN JUAN

Dates Standardized anomalies (observed − computed) 1 10 100 min.d. (km) 123 1249

Relative homogeneity

1960 1970 1980 1990 2000 2010 −4 −2 2 4

VX−d at 1024E(7), SAN SEBASTIÁN,IGUELDO

Dates Standardized anomalies (observed − computed) 1 10 100 min.d. (km) 114 584

Windowed SNHT histogram

Histogram of maximum tV

tVx Frequency 100 200 300 400 2 4 6 8 10 12

Complete SNHT histogram

Histogram of maximum SNHT

SNHT Frequency 1000 2000 3000 4000 5000 2 4 6 8 10 12

Abnormal series reconstruction

7 8 9 10 11 12 13

VX−m at 8368U(57), TERUEL

Running annual means 1960 1970 1980 1990 2000 2010 0.0 0.5 1.0 1.5 2.0 Years Correction factors

Residual inhomogeneities

1960 1970 1980 1990 2000 2010 −4 −2 2 4

VX2−d at 2539(25), VALLADOLID/VILLANUBLA

Dates Standardized anomalies (observed − computed) 1 10 100 min.d. (km) 262 380

Change of variance

1960 1970 1980 1990 2000 2010 −4 −2 2 4

VX2−d at P535(75), LISBOA GEOFÍSICO

Dates Standardized anomalies (observed − computed) 1 10 100 min.d. (km) 192 21

Other homogenizations

Due to these unsatisfactory results, further homogenizations were performed either directly on the daily data or on annual extreme wind indexes, which led to decreasing levels of break detection when compared to the monthly homogenization: Series Breaks Raw (filled) – Monthly+MM5 to daily 171 Daily+MM5 87 Daily 47 Annual indexes: Averages Maximums Days>90% 28 6 25

Trends of mean peak gusts

• Raw

series Monthly+MM5 to daily Daily with MM5 Daily Annual parameters −1.0 −0.5 0.0 0.5 Homogenization methods Trend (m/s/10y)

Breaks 171 87 47 28

Trends of mean daily peak gusts

Trends of annual peak gusts

• Raw

series Monthly+MM5 to daily Daily with MM5 Daily Annual parameters −2 −1 1 2 3 Homogenization methods Trend (m/s/10y)

Breaks 171 87 47 6

Trends of annual maximum peak gusts

Trends of days > 90%

• Raw

series Monthly+MM5 to daily Daily with MM5 Daily Annual parameters −15 −10 −5 5 10 Homogenization methods Trend (days/10y)

Breaks 171 87 47 25

Trends of nr. of days with peak gust > 90 precentile

• Max. expected peak gusts
• Raw50

M+M50 D+M50 Day50 AnP50 Raw100 M+M100 D+M100 Day100 AnP100 Raw200 M+M200 D+M200 Day200 AnP200 20 30 40 50 60 70 Maximum expected peak gusts (m/s) Homogenization method + Return period

Maximum expected peak gusts (m/s) for return periods of 50, 100 and 200 years

Conclusions

◮ In many cases, there is no clear evidence suggesting that

the homogenization of the daily series is needed (especially for computing trends of average values).

◮ But these results, derived from real data, cannot be

conclusive, since we do not know the true solution.

◮ ⇒ Further experiments should be performed with

synthetic data.

Conclusions

◮ In many cases, there is no clear evidence suggesting that

the homogenization of the daily series is needed (especially for computing trends of average values).

◮ But these results, derived from real data, cannot be

conclusive, since we do not know the true solution.

◮ ⇒ Further experiments should be performed with

synthetic data.

Conclusions

◮ In many cases, there is no clear evidence suggesting that

the homogenization of the daily series is needed (especially for computing trends of average values).

◮ But these results, derived from real data, cannot be

conclusive, since we do not know the true solution.

◮ ⇒ Further experiments should be performed with

synthetic data.