SPHERA (High Resolution REAnalysis over Italy): system setup and - - PowerPoint PPT Presentation

SPHERA

(High Resolution REAnalysis over Italy): system setup and tests

Ines Cerenzia, Tiziana Paccagnella, Andrea Montani Arpae-Emilia Romagna, Regional Service for Environment Protection and Energy, Bologna Italy

International Symposium on Regional Reanalysis, Bonn, July 2018

Overview Overview

• Introduction
• Question1: 1step or 2steps nest into driver model?
• Question2: which deep temperature does assign to the

regional model?

Motivation Motivation

Climate study: provide a high resolution, space and time consistent description of the past decades climate characteristics NWP model evaluation for the operational forecast: provides a testbed for extended period (multi-decadal) evaluation of regional models (e.g. severe weather events) Downstream applications: provide boundary conditions to high resolution models (e.g. coastal erosion and marine sediment, sea- state,air pollution,..)

PM10-2017

Yearly averaged T2m and cumulated precipitation anomaly over Emilia-Romagna (ERACLITO, Antolini et

• al. 2015)

• Model: Consortium for Small Scale Modeling

(COSMO)

–

non-hydrostatic limited area model

–

convection permitting resolution (2.2km)

–

Hourly output

–

Setup based on NWP operational forecast

• Initial state from ERA5
• Boundary conditions:

–

Hourly at the border for upper air fields

–

Sea state daily updated to ERA5

–

Soil/snow/lake autonomous

• Data Assimilation: continuous nudging of

surface/upper air convectional observations (no radar and satellite)

• Temporal Coverage: 1985-2020
• Production: 09/2018-12/2020

Confiuration Confiuration

Q1: 1step or 2steps nest into ERA5?

Q1: 1step or 2steps nest into ERA5?

2step 1step

ERA5 - IFS ERA5 - IFS (31km) (31km) COSMO (10km) COSMO (10km) COSMO (2.2km) COSMO (2.2km) COSMO (2.2km) COSMO (2.2km) COSMO (2.2km) COSMO (2.2km)

2step 1step

ERA5 - IFS ERA5 - IFS (31km) (31km) COSMO (10km) COSMO (10km) COSMO (2.2km) COSMO (2.2km) COSMO (2.2km) COSMO (2.2km) COSMO (2.2km) COSMO (2.2km)

Deep convection: Tiedtke scheme Soil/Snow/Lake autonomous Soil/Snow interpolated from COSMO10km Lake autonomous Soil/Snow/Lake autonomous

Q1: 1step or 2steps nest into ERA5?

Q1: 1step or 2steps nest into ERA5?

Accumulated precipitation 2015 Accumulated precipitation 2015

ERA5 1step 2step

Accumulated precipitation 2015 Accumulated precipitation 2015

ERA5 1step 2step

WINTER

• 1step performs better

than 2step at almost all thresholds (less FA, less BIAS)

• Regional reanalysis

improves against ERA5 for precipitation >25mm (more POD and FA)

Verification using boxes of 0.25°x0.25° Observations not ingested in data assimilation

Daily accumulated precipitation - Obs Daily accumulated precipitation - Obs

SUMMER

• Even larger

improvement of 1step to 2step than winter (less FA, less BIAS)

• Regional reanalysis

improves against ERA5 for precipitation >10mm (more POD and FA)

Verification using boxes of 0.25°x0.25° Observations not ingested in data assimilation

Daily accumulated precipitation - Obs Daily accumulated precipitation - Obs

T emperature at 2m - Obs T emperature at 2m - Obs

Stable PBL Surface warming up

1step performs better than 2 step, especially in uneasy conditions to represent SPHERA improves in bias and rmse over land compared to ERA5

1step performs better than 2step, especially in summertime, and it generates less precipitation than 2 step

• Convection is too pumped in 2step due to the deep convection

parametrization in the intermediate COSMO integration?

• Noise introduced by the intermediate step in the nest? Gravity

waves?

• Soils develop differently?

SUMMER 12UTC

14km 4km 275m 20m Higher energy in the free atmosphere in 2step than 1step Energy spectra

• f the vertical

wind speed at different altitudes 24/07/2015 at 12UTC: several convective cells

• ver Italy

Lower energy in the CBL in 2step than 1step

1step: self-developing soil at 2.2km 2step: interpolated from the intermediate resolution run at 10km

Soil diferences Soil diferences

Soil can have a large effect on long time scales due to its inertia In regional models, the temperature of the deepest soil level is set:

• By vertically interpolating the driver model.. but not always

meaningful (Hdeep in ERA5 = -1.89m >> Hdeep in COSMO =

• 14.58m)
• By boundary conditions, e.g. flux =0 or T=const .. but it misses the

interannual thermal changes and the global warming Idea: Parametrize it using the soil features of the driver model at the deepest levels

Q2: Which deep soil temperature? Q2: Which deep soil temperature?

Parametrizations Parametrizations

1) 5 stations over Europe with multi-years deep soil temperature observations over (0.5m-12m) 2) 3 parametrizations of the soil annual thermal cycle (iiven two time series of T soil at two shallower depths, and assumini constant thermal difusion below the iiven levels) 1) Method of amplitude (Evett, 2002) 2) Method of phase (Verhoef et al, 1996) 3) Three-yearly runnini mean delayed by a time lai <===== 3) Reconstruct the Tdeep feld at -14.58m required by COSMO 4) Re-run for 12months+6months (spin-up) 1)

Diferences of Tdeep Diferences of Tdeep

Standard solution in COSMO: 1) Report deepest temperature of ERA5 to deepest level of SPHERA (-1.89m! =-14.58m 2) Keep it constant from the initialization time Compared to time lai method ------> Diferences depend on the

• Initialization time
• Soil type (lonier lai in Po valley than
• ver Alps)
• Time passed since initialization

Diferences can be as larie as 3°C

Diferences of Tdeep Diferences of Tdeep

Standard solution in COSMO: 1) Report deepest temperature of ERA5 to deepest level of SPHERA (-1.89m! =-14.58m 2) Keep it constant from the initialization time Compared to time lai method ------> Diferences depend on the

• Initialization time
• Soil type (lonier lai in Po valley than
• ver Alps)
• Time passed since initialization

Diferences can be as larie as 3°C

Questions: 1) How large is the impact into the soil? 2) Is there an effect at the surface and in the atmpsphere? 3) After how much time?

Conclusions Conclusions

• SPHERA at 2.2km will cover Italy from 1985-2020. First 5 years will be

ready by the end of 2018

• Direct nest (1step) into ERA5 with a 1:15 resolution step between the

inteiration domains performs better than traditional resolution steppini (1:3:5) for precipitation and temperature at 2m.

• Reason is likely the diferent soil status in 1step and 2step.
• Deep soil temperature should be considered when settini loni-term
• inteiration. Update it to a simple runnini mean of the deepest

temperature available from the driver model could already be a iood solution ……. Thanks for your attention!

1)

Additional Material

COSMO-1Nest COSMO-2Nest

• COSMO-2I
• IC : Era5 il primo giorno, poi inizializzazione a caldo
• BC: Era 5 /1h
• Suolo/Neve/Laghi autonomi
• SST aggiornata a ERA5 ogni giorno
• Periodo: 2015 + 6mesi di inizializzazione (7/2014-12/2014)
• Osservate: SYNOP/SHIP/TEMP/PILOT/ AIREP (METAR

durante i primi 5mesi di inizializzazione)

• Vers:INT2LM 2.04, COSMO 5.3 in single precision (no

urban)

• Dominio: 16 punti griglia più largo rispetto a COSMO-2I in
• gni direzione

COSMO-10M

• IC : Era5 il primo giorno, poi inizializzazione a caldo
• BC: Era 5 /1h
• Suolo/Neve/Laghi autonomi
• SST aggiornata a ERA5 ogni giorno
• Periodo: 2015 + 6mesi di inizializzazione (7/2014-12/2014)
• Osservate: SYNOP/SHIP/TEMP/PILOT/ AIREP (METAR durante

i primi 5mesi di inizializzazione)

• Vers:INT2LM 2.04, COSMO 5.3 in single precision (no urban,

Tiedtke convection)

• Dominio: COSMO-5M
• COSMO-2I
• IC : COSMO-10M il primo giorno, poi inizializzazione a caldo
• BC: COSMO-10M /1h
• Suolo/Neve/SST interp. da COSMO-10M; laghi autonomi
• Vers:INT2LM 2.04, COSMO 5.3 in single precision (no urban)
• Dominio:COSMO-2I

JJA average of the T2m difference 1step-2step