ASSESSMENT OF ALGORITHMS FOR LAND SURFACE ANALYSIS DOWN-WELLING LONG-WAVE RADIATION AT THE SURFACE

Similar documents
Isabel Trigo, Sandra Freitas, Carla Barroso, Isabel Monteiro, Pedro Viterbo

INTERCOMPARISON OF METEOSAT-8 DERIVED LST WITH MODIS AND AATSR SIMILAR PRODUCTS

THE LAND-SAF SURFACE ALBEDO AND DOWNWELLING SHORTWAVE RADIATION FLUX PRODUCTS

A satellite-based long-term Land Surface Temperature Climate Data Record

Satellite Application Facility on Land Surface Analysis (LSA-SAF/Land SAF): Products and applications

PRECONVECTIVE SOUNDING ANALYSIS USING IASI AND MSG- SEVIRI

OSI SAF SST Products and Services

VALIDATION OF THE OSI SAF RADIATIVE FLUXES

Land Surface Temperature in the EUMETSAT LSA SAF: Current Service and Perspectives. Isabel Trigo

LAND SURFACE ALBEDO AND DOWNWELLING SHORTWAVE RADIATION FROM MSG GEOSTATIONARY SATELLITE (METHOD FOR RETRIEVAL, VALIDATION, AND APPLICATION)

EUMETSAT LSA-SAF EVAPOTRANSPIRATION PRODUCTS STATUS AND PERSPECTIVES

SAFNWC/MSG SEVIRI CLOUD PRODUCTS

Saharan Dust Longwave Radiative Forcing using GERB and SEVIRI

MSG system over view

IMPORTANCE OF SATELLITE DATA (FOR REANALYSIS AND BEYOND) Jörg Schulz EUMETSAT

EUMETSAT products and services for monitoring storms - New missions, more data and more meteorological products

Outgoing Longwave Radiation Product: Product Guide

EUMETSAT SAF NETWORK. Lothar Schüller, EUMETSAT SAF Network Manager

Evapotranspiration monitoring with Meteosat Second Generation satellites: method, products and utility in drought detection.

Surface Radiation Budget from ARM Satellite Retrievals

Plans for the Assimilation of Cloud-Affected Infrared Soundings at the Met Office

NWC-SAF Satellite Application Facility in Support to Nowcasting and Very Short Range Forecasting

REVISION OF THE STATEMENT OF GUIDANCE FOR GLOBAL NUMERICAL WEATHER PREDICTION. (Submitted by Dr. J. Eyre)

Bias correction of satellite data at Météo-France

EUMETSAT SAF NETWORK. Lothar Schüller, EUMETSAT SAF Network Manager

Lecture 4: Radiation Transfer

Remote sensing derived evapotranspiration: comparisons to observations and models and results over the full MSG disk and selected basins

METEOSAT cloud-cleared radiances for use in three/fourdimensional variational data assimilation

Spectrum of Radiation. Importance of Radiation Transfer. Radiation Intensity and Wavelength. Lecture 3: Atmospheric Radiative Transfer and Climate

Lecture 3: Atmospheric Radiative Transfer and Climate

THE ATMOSPHERIC MOTION VECTOR RETRIEVAL SCHEME FOR METEOSAT SECOND GENERATION. Kenneth Holmlund. EUMETSAT Am Kavalleriesand Darmstadt Germany

ESTIMATION OF ATMOSPHERIC COLUMN AND NEAR SURFACE WATER VAPOR CONTENT USING THE RADIANCE VALUES OF MODIS

Improved assimilation of IASI land surface temperature data over continents in the convective scale AROME France model

Towards a better use of AMSU over land at ECMWF

A new formula for determining the atmospheric longwave flux at the ocean surface at mid-high latitudes

Simulation and validation of land surface temperature algorithms for MODIS and AATSR data

A Longwave Broadband QME Based on ARM Pyrgeometer and AERI Measurements

Atmospheric longwave radiation under cloudy skies for HAM simulation programs

WACMOS-ET LST Product. Algorithm Theoretical Basis Document

Seeking a consistent view of energy and water flows through the climate system

The skill of ECMWF cloudiness forecasts

Hyperspectral Observations of Land Surfaces: Temperature & Emissivity

EXPERIENCE IN THE HEIGHT ATTRIBUTION OF PURE WATER VAPOUR STRUCTURE DISPLACEMENT VECTORS

GUEDJ Stephanie KARBOU Fatima RABIER Florence LSA-SAF User Workshop 2010, Toulouse

A HIGH RESOLUTION EUROPEAN CLOUD CLIMATOLOGY FROM 15 YEARS OF NOAA/AVHRR DATA

VALIDATION RESULTS OF THE OPERATIONAL LSA-SAF SNOW COVER MAPPING

May 3, :41 AOGS - AS 9in x 6in b951-v16-ch13 LAND SURFACE ENERGY BUDGET OVER THE TIBETAN PLATEAU BASED ON SATELLITE REMOTE SENSING DATA

The potential impact of ozone sensitive data from MTG-IRS

Data assimilation of IASI radiances over land.

P2.7 A GLOBAL INFRARED LAND SURFACE EMISSIVITY DATABASE AND ITS VALIDATION

22nd-26th February th International Wind Workshop Tokyo, Japan

Results from the ARM Mobile Facility

For the operational forecaster one important precondition for the diagnosis and prediction of

Validation of Direct Normal Irradiance from Meteosat Second Generation. DNICast

Radiation in climate models.

CTTH Cloud Top Temperature and Height

Extraction of incident irradiance from LWIR hyperspectral imagery

Radiative Equilibrium Models. Solar radiation reflected by the earth back to space. Solar radiation absorbed by the earth

UPDATES IN THE ASSIMILATION OF GEOSTATIONARY RADIANCES AT ECMWF

Long-Term Time Series of Water Vapour Total Columns from GOME, SCIAMACHY and GOME-2

GIFTS SOUNDING RETRIEVAL ALGORITHM DEVELOPMENT

Christian Sutton. Microwave Water Radiometer measurements of tropospheric moisture. ATOC 5235 Remote Sensing Spring 2003

ATMOS 5140 Lecture 1 Chapter 1

Assimilation of Cloud-Affected Infrared Radiances at Environment-Canada

An Alternate Algorithm to Evaluate the Reflected Downward Flux Term for a Fast Forward Model

Derivation of AMVs from single-level retrieved MTG-IRS moisture fields

Data Short description Parameters to be used for analysis SYNOP. Surface observations by ships, oil rigs and moored buoys

Comparison of Long-term Downward Radiation Observations at Tateno with JRA-25 and ERA-40 Data

Global Instability Index: Product Guide

The EUMETSAT Satellite Application Facility on Land Surface Analysis (Land SAF): Proposed Products

Next generation of EUMETSAT microwave imagers and sounders: new opportunities for cloud and precipitation retrieval

COMPARISON OF SIMULATED RADIANCE FIELDS USING RTTOV AND CRTM AT MICROWAVE FREQUENCIES IN KOPS FRAMEWORK

Rosemary Munro*, Graeme Kelly, Michael Rohn* and Roger Saunders

VALIDATION OF MSG DERIVED SURFACE INCOMING GLOBAL SHORT-WAVE RADIATION PRODUCTS OVER BELGIUM

Assimilation of precipitation-related observations into global NWP models

Meteorological product extraction: Making use of MSG imagery

H-SAF future developments on Convective Precipitation Retrieval

Lectures 7 and 8: 14, 16 Oct Sea Surface Temperature

Combining Polar Hyper-spectral and Geostationary Multi-spectral Sounding Data A Method to Optimize Sounding Spatial and Temporal Resolution

Atmospheric Motion Vectors: Product Guide

Masahiro Kazumori, Takashi Kadowaki Numerical Prediction Division Japan Meteorological Agency

Observations needed for verification of additional forecast products

Radiation in the atmosphere

Validation Report for Precipitation products from Cloud Physical Properties (PPh-PGE14: PCPh v1.0 & CRPh v1.0)

AN ACCURACY ASSESSMENT OF AATSR LST DATA USING EMPIRICAL AND THEORETICAL METHODS

Extending the use of surface-sensitive microwave channels in the ECMWF system

OPTIMISING THE TEMPORAL AVERAGING PERIOD OF POINT SURFACE SOLAR RESOURCE MEASUREMENTS FOR CORRELATION WITH AREAL SATELLITE ESTIMATES

Comparison of cloud statistics from Meteosat with regional climate model data

The assimilation of AMSU and SSM/I brightness temperatures in clear skies at the Meteorological Service of Canada

Improving real time observation and nowcasting RDT. E de Coning, M Gijben, B Maseko and L van Hemert Nowcasting and Very Short Range Forecasting

Direct assimilation of all-sky microwave radiances at ECMWF

Remote Sensing of Precipitation

The retrieval of the atmospheric humidity parameters from NOAA/AMSU data for winter season.

EUMETSAT PLANS. Dr. K. Dieter Klaes EUMETSAT Am Kavalleriesand 31 D Darmstadt Germany

The LSA-SAF Albedo products

Observational Needs for Polar Atmospheric Science

Xianglei Huang University of Michigan Xiuhong Chen & Mark Flanner (Univ. of Michigan), Ping Yang (Texas A&M), Dan Feldman and Chiancy Kuo (LBL, DoE)

ATMOS 5140 Lecture 7 Chapter 6

OBJECTIVE USE OF HIGH RESOLUTION WINDS PRODUCT FROM HRV MSG CHANNEL FOR NOWCASTING PURPOSES

COMPARING PERFORMANCE OF SOLARGIS AND SUNY SATELLITE MODELS USING MONTHLY AND DAILY AEROSOL DATA

Transcription:

ASSESSMENT OF ALGORITHMS FOR LAND SURFACE ANALYSIS DOWN-WELLING LONG-WAVE RADIATION AT THE SURFACE Isabel F. Trigo, Carla Barroso, Sandra C. Freitas, Pedro Viterbo Instituto de Meteorologia, Rua C- Aeroporto, 1700-049 Lisboa, Portugal Abstract The Satellite Application Facility on Land Surface Analysis (LSA SAF) has been generating Downwelling Surface Longwave Flux (DSLF) from the Meteosat Second Generation (MSG) satellite, on a pixel-by-pixel basis, since the beginning of 005. The retrieved DSLF corresponds to instantaneous values, estimated every 30-minutes, for the whole Meteosat disk. DSLF can only be indirectly inferred from remotely sensed data. The LSA SAF approach makes use of separate bulk parameterization schemes suitable for clear and cloudy conditions, respectively. DSLF retrievals benefit from the signature of clouds and different cloud types on IR (Infrared) and VIS (Visible) channels, complemented with information on atmosphere water content and near surface air temperature available from Numerical Weather Prediction (NWP) fields. The comparison against in situ measurements (mostly obtained from BSRN - Baseline Surface Radiation Network - stations) suggests the LSA SAF DSLF is generally underestimated. This is particularly apparent for clear cases, with biases of the order of 10 Wm - to 0 Wm -. Cloudy pixels also tend to exhibit negative biases (mostly within 10 Wm - to 30 Wm -, for European sites), but higher dispersion than in clear cases. As a step forward to eliminate the detected biases, this work presents an assessment of different DSLF algorithms, valid for clear and cloudy conditions together with a new proposed formulation, applicable to all conditions. The different schemes are compared with modelled data MODTRAN and with in situ (BSRN) measurements. The modelled fluxes are estimated for the TIGR-like database that samples temperature and humidity profiles within ECMWF (European Centre for Medium-Range Weather Forecasts) re-analyses (ERA-40). This database presents a comprehensive and balanced set of atmospheric profiles, suitable for calibration/validation of radiative models/schemes. The new proposed algorithm expresses the effects of cloud cover, atmospheric temperature and humidity through parameterization of the emisssivity and the effective temperature. The calibration of this new parameterization scheme makes use of the MODTRAN DSLF values. When the algorithm is evaluated against in situ data, it reveals an overall better performance than the remaining formulations, and proves to be the most stable under moist and dry conditions. INTRODUCTION The Downwelling Surface Radiative Flux (DSLF), defined as the irradiance reaching the surface in the thermal infrared part of the spectrum (4-100 µm), is one of the components involved in the surface radiation budget. Estimations of this parameter are potentially important for the validation/verification of numerical weather forecast models, climate monitoring and also to assess energy and agricultural needs. The DSLF can be obtained through radiative transfer model calculations, if the properties of the overlaying air column are well known. Such atmospheric properties can be obtained from radio soundings, which are usually infrequent and geographically sparse. Furthermore, there are insufficient

direct measurements of this quantity, since the surface radiation observation network is particularly limited for the case of longwave observations (Niemelä et al., 001). Satellite measurements, which provide data with a wide coverage and temporal samplings of up to 15 minutes (in the case of the Meteosat Second Generation), are in the best position to allow retrieval of radiative fluxes over large areas. Clouds have a strong effect on the longwave radiation transfer because they modify the atmospheric emissivity at certain wavelengths; they are almost completely opaque to infrared radiation and prevent the escape of longwave radiation into space. DSLF is greater when clouds are present, especially in the case when low clouds are warmer than the surface. In the thermal window part of the spectrum the downward fluxes are dominated by the concentration and temperature of the atmospheric water vapour, but if clouds are present, most radiation is emitted at cloud bottom. Temperature and height of the optically thick thermal surface may be determined by NWP models or vertical sounders (TOVS), while cloud properties may be inferred from satellite data. Several methods have been developed aiming to estimate the DSLF from top of atmosphere satellite observations. The Satellite Application Facility on Land Surface Analysis (LSA SAF) makes use of a semi-empirical method to obtain DSLF every 30 minutes from Meteosat Second Generation TOA observations, on a pixel-by-pixel basis. The DSLF retrievals benefit from the signature of clouds and different cloud types on IR (Infrared) and VIS (Visible) channels, obtained from the Nowcasting SAF (NWC SAF; http://nwcsaf.inm.es/) complemented with information on atmosphere water content and near surface air temperature available from NWP fields. The latter, obtained from ECMWF (European Centre for Medium-Range Weather Forecasts) 1 to 4 hours forecast, indirectly include information from atmospheric sounders and other available observations, and thus correspond to the best knowledge of atmospheric profiles for each time-slot. The LSA SAF approach to estimate DSLF considers separate bulk parameterization schemes suitable for clear (Prata, 1996) and cloudy conditions (Josey et al., 003), respectively. In this paper an evaluation of four different longwave (LW) radiation parameterization schemes is presented, which include the clear and cloudy algorithms currently used by the LSA SAF, a parameterization for clear pixels from Dilley and O Brien (1998), and a new proposed parameterization. The latter one corresponds to a modified version of the algorithm developed by Prata (1996), applicable to all situations. The parameterizations are compared with values simulated by MODTRAN and with in situ measurements from a set of Baseline Surface Radiation Network (BSRN) stations. PARAMETERIZATION SCHEMES There are several approaches to compute DSLF from bulk parameterization schemes that are based on the Stefan-Boltzmann law, considering that the atmospheric layer immediately above the surface emits IR radiation at a temperature T, and with an effective emissivity ε : 4 F = σε T (1) Table 1 summarizes the different assumptions/formulations adopted by the parameterizations assessed in this study. For clear conditions, the emissivity ε is typically defined as a function of the atmosphere humidity and the screen level air temperature (Niemelä et al., 001). In the formulation proposed by Prata (1996), applicable for clear, the effective emissivity depends explicitly on the water vapour content of the atmosphere ( w ), and the coefficients are empirical constants derived from observational data. Other formulation applicable under clear conditions developed by Dilley and O Brien (1998) considers the emissivity as a function of w and screen level air temperature. In this case, the parameters were determined by comparing model irradiances with irradiances computed from radiative transfer model computations. The LSA SAF has used this latter algorithm since the

beginning of its pre-operational activities (in 005) to retrieve DSLF for clear conditions. Nevertheless validation studies conducted by the LSA SAF (LSA SAF Validation Report_1.6, 007) have revealed that the formulation derived by Prata (1996) presented slightly better results, particularly for the European sites. Scheme ε T Applicable Conditions Prata (1996) 1 w w 1 + exp 1. + 3 10 10 1 w = total column water vapour (mm) T T = two meters air temperature (K) Clear Sky Dilley and O Brien (1998) τ = 1 exp ( 1.66τ ) 1 T w.3 1.88 0.74 + 73 5 T Clear Sky Josey et al. (003) 1 T + 10.77n +.34n 18. 44 +.84 Td T 4.01 ( ) 0 + All Sky Conditions n=cloud fraction; T =dew point (K) Prata MODIFIED w w 1 1 + exp α + β 10 10 α, β, m - new parameters cloud cover dependent m T + γ T T ) + δ ( d γ, δ - new parameters cloud cover dependent All Sky Conditions Table 1: Terms of eq. 1 in the parameterizations under analysis. The parameterization scheme proposed by Josey et al. (003) is expressed in terms of the surface temperature adjustment necessary to obtain the effective temperature of a blackbody, which emits a radiative flux equivalent to the atmospheric longwave. This effective temperature ( T ) is a function of the total cloud amount and of the dew point depression. For the case of DSLF retrieved by the LSA SAF the required information on clouds is obtained from NWC SAF software (http://nwcsaf.inm.es/) and the temperature and humidity of the atmosphere are obtained from ECMWF forecasts. This formulation, applicable under all conditions was calibrated with cruise measurements over the Atlantic Ocean (Josey et al., 003). The new formulation proposed in this study (Prata MODIFIED ) is based on the scheme first developed by Prata (1996), but with both the emisssivity and the effective temperature adjusted according to the cloud cover, temperature and atmospheric water vapour content. This new formulation is applicable to all situations, considering that DSLF is given by: PRATAMOD CLOUD ( n) DSLFPRATAMOD CLEAR DSLF = ndslf + 1 () where n is the fraction of cloud cover and DSLF ( PRATAMOD CLOUD DSLF ) is the new modified PRATAMOD CLEAR Prata formulation with coefficients calibrated for cloudy (clear) conditions.

NEW DSLF FORMULATION - CALIBRATION The new parameterization presented here was calibrated with data obtained from radiative transfer model calculations. The MODerate spectral resolution atmospheric TRANSsmittance algorithm (MODTRAN4; Berk et al., 000) was used to compute synthetic DSLF (hereafter DSLFMODTRAN) for TIGR-like database (Chevallier et al., 001). The TIGR-like database consists of a sample of atmospheric profiles (temperature, moisture, etc) collected from ECMWF reanalysis (ERA-40), which are representative for radiative transfer modelling. The coefficients of the new parameterization scheme were estimated separately for clear and conditions, corresponding to profiles in the TIGRlike database with total cloud cover above 90% and below 10%, respectively. PARAMETERIZATIONS VERSUS MODTRAN The different parameterizations are compared with MODTRAN simulated fluxes. The comparison between the schemes under study and MODTRAN simulations shows a fairly good agreement for all algorithms. In the case of cloudy conditions, the scheme developed by Josey et al. (003), agrees well with MODTRAN simulations for fluxes below approximately 300 Wm - (Fig.1), which are typical of mid-tohigh latitudes corresponding to the regions where the algorithm was calibrated; higher DSLF values are strongly underestimated. The new parameterization shows a low bias of 3. Wm - with a root mean square error (RMSE) of 11.4 Wm - against the observed bias of 54.4 Wm - and a RMSE of 5.5 Wm - obtained with Josey et al. (003) formulation. Figure 1: Scatterplot of DSLF values estimated using a bulk parameterization scheme (see color code) versus MODTRAN simulations, for cloudy conditions. In the case of clear conditions, the algorithm by Josey et al. (003) reveals again a clear conditional bias, with underestimations of high DSLF values (Fig.). Table contains the mean differences and root mean square differences between each scheme and MODTRAN. The mean differences are considerably low for all the algorithms under analysis. The new formulation shows a good fit to the MODTRAN simulations with bias and RMSE, of -0.4 Wm - and 9.1 Wm -, respectively.

Figure : As in Fig. 1, but for clear conditions. Scheme Bias (Wm - ) RMSE (Wm - ) Prata (1996) 11.6 17.6 Dilley&O Brien(1998) -0.7 19. Josey et al. (003) -0.7 38.1 Prata_modified -0.4 9.1 Table : Mean differences (Bias) and mean square differences (RMSE) for the different parameterization schemes under study, observed in the comparison with MODTRAN simulated data, for the case of clear conditions. The analysis of model performance with atmospheric water content (for clear conditions) shows very clearly that Josey et al. (003) algorithm has increasing biases for higher total column water vapour (TCWV) values (Fig.3). It should be noted, however, that the number of points that falls into this higher classes is considerably smaller than those corresponding to TCWV<10 mm (35 points from a sample of 488 profiles). In any case it is clear that the modified Prata scheme seems to be the most stable under both dry and moist conditions; the worst mean difference is found to be approximately 10 Wm - for the most extreme TCWV atmospheric content (above 40 mm), in which only 1 cases fall into. Figure 3: Mean differences between parameterized DSLF (see colour code) and MODTRAN, estimated per classes of total column water vapour, for clear sly conditions.

PARAMETERIZATIONS VERSUS IN SITU OBSERVATIONS Since the new parameterization scheme was calibrated with MODTRAN simulations, it can only be properly validated against independent observational data. Estimations of DSLF obtained from the formulations presented above are compared with long-wave fluxes measured at the four BSRN stations indicated in Fig. 4. The observation period ranges from January-December 005 to July- November 005 (in the case of Tamanrasset). Figure 4: Location of stations with in situ measurements of IR downward fluxes at the surface, used in the current study. For clear situations, all the considered schemes, with the exception of Josey et al. (003) at Tamanrasset, show mean differences between estimated and observed DSLF of less than about 0 Wm - and mean root square differences < 30Wm - (Fig. 5). Clear Sky: Figure 5: Mean differences (left) and mean square differences (right) between DSLF estimated with the parameterization schemes under study (see colour code) and in situ observations, for clear conditions. For partially cloudy conditions Josey et al. (003) formulation is shown to perform worse for all of the stations, with mean differences between estimated and observed DSLF greater than about 30 Wm - and RMSE also > 30 Wm -. Under these conditions, RMSE is also higher for the Prata MODIFIED (of about 0 Wm - for most of the stations), but the bias tend to remain below 15 Wm - for most sites.

Partially cloudy: Figure 6: As in Fig. 5, but for partially cloudy conditions. The comparisons of parameterizations versus in situ observations for overcast situations is presented in Fig. 7; both Josey et al. (003) and Prata MODIFIED schemes show a very good agreement, with the exception of Tamanrasset, with biases less then 10 Wm - and mean square differences of less then 30 Wm -. Overcast: Figure 7: As in Fig. 5, but for overcast conditions. CONCLUSIONS This study presents an assessment of a set of bulk parameterizations of IR downwelling fluxes at the surface. Two of the considered formulations are applicable only to clear conditions, the one presented by Prata (1996) and the one by Dilley and O Brien (1998). The former is the algorithm currently being used by the LSA SAF for the estimation of clear DSLF; for cloudy conditions the methodology relies on the formulation proposed by Josey et al. (003). In an attempt to derive a formula applicable to both clear and cloudy conditions, the formulation first proposed by Prata (1996) was adapted to make a correction of both the emisssivity and the effective temperature taking into account the cloud cover, temperature and atmospheric water vapour content. This new parameterization was calibrated using data simulated by MODTRAN for TIGR-like database (Chevallier et al., 001).

The performance of the four considered formulations are evaluated by comparing the longwave flux estimated by the different schemes with data simulated by MODTRAN, and to a set of local measurements from BSRN stations located over Europe (namely Roissy, Carpentras and Payerne) and over Africa (Tamanrasset). The comparisons between DSLF obtained from the different schemes and data modelled with MODTRAN reveal good agreement for all the algorithms, except for Josey et al. (003) in cases of high atmospheric water vapour contents (greater than about 10 mm). Nevertheless the new proposed formulation shows the best results, with biases of 0.4 Wm - and 3. Wm - for the cases of clear and cloudy conditions, respectively and is also found to be very stable for both dry and moist conditions. When compared against in situ observations, the algorithms show fairly good performance over the European sites, and specially for the extreme situations of clear and overcast, with mean differences not greater than 0 Wm - in most of the cases. Over Africa the results are quite similar, except for Josey et al. (003), that strongly underestimates the observed flux for all situations. Overall the new proposed formulation, Prata MODIFIED, applicable for clear and cloudy cases, presents the most consistent statistics, when compared with MODTRAN simulated and observed IR fluxes. ACKNOWLEDGMENTS This work has been carried out within the scope of LSA SAF, co-funded by EUMETSAT. REFERENCES Berk, A., G.P. Anderson, P.K. Acharya, J.H. Chetwynd, L.S. Bernstein, E.P. Shettle, M.W. Matthew, and S.M. Alder-Golden, (000): MODTRAN4 Version User s Manual Air Force Res. Lab., Space Vehicles Directorate, Air Force Material Command, Hanscom AFB, MA, 000. Chevallier, F., (001): Sampled databases of 60-level atmospheric profiles from the ECMWF analyses. Numerical Weather Prediction Satellite Application Facility Research Report [NWP SAF Res. Rep.] no. 4, Jan 00. Dilley, A.C. and D.M. O Brien (1998): Estimating downward clear long-wave irradiance at the surface from screen temperature and precipitable water, Q. J. R. Meteorol. Soc., 14, 1391-1401. Josey, S.A., Pascal, R.W., Taylor, P.K., Yelland, M.J., (003): A New Formula For Determining the Atmospheric Longwave Flux at Ocean Surface at Mid-High Latitudes. J Geophys. Res., doi:10.109/00jc00141. LSA SAF (007): Validation Report, version 1.6. https://landsaf.meteo.pt/. Niemelä, S., P Räisänen, H Savijärvi (001): Comparison of surface radiative flux parametrizations. Part I: Longwave radiation. Atmosph. Res., 58, 1-18. Prata, A.J. (1996): A new long-wave formula for estimating downward clear- radiation at the surface, Q. J. R. Meteorol. Soc., 1, 111-1151.