Jeff Settle (ESSC, UK)

Albedo requirements of the UK National
Centre for Earth Observation (NCEO)
& the UKMO
Jeff Settle,University of Reading/NCEO

RADAGAST - Radiative flux studies in
the Sahel (an AMMA-related project)
MOSES, JULES, and the UKMO
requirements for satellite albedo
products

RADAGAST
(Radiative Atmospheric Divergence using ARM
mobile facility, GERB data and AMMA STations)
Broad band flux data from geostationary orbit are
combined with continuous surface measurements to
give the first temporally well-sampled estimates of
flux divergence across the atmosphere - (i.e. the
amount of radiation absorbed in the atmosphere).
Has to be estimated as a residual. Estimates of this
(global average) vary from 67 Wm –2 (Kiehl & Trenberth
1997) to 95 Wm –2 (Wild et al 1995).

RADAGAST Methodology
1. Observe the top of
atmosphere fluxes
using the GERB and
SEVIRI instruments
on Meteosat
2. Observe the surface fluxes
using the ARM sites and use
additional data to
characterize the fluxes over
the GERB footprint
3. Calculate the flux
divergence across the
atmosphere as the difference
between the surface and top
of atmosphere fluxes
GERB determines the spatial scale at which comparisons must
be made: ~ 50km at surface

The RADAGAST Surface Sites
Evidently, the ground
station (AMF = ARM
Mobile Facility) at
Niamey airport
belongs to a GERB
pixel (large squares)
that is not very
uniform. A second
set of instruments
seem to have been
better placed.
How does this affect
the comparisons we
make/made?

Sampling Error in the Surface Radiation
Budget
SRB = SW ↓−SW ↑+LW ↓−LW ↑
= SW ↓ 1− a
( ) − εσT s
a = surface albedo
U = SRB AMF − SRB GERB
( )
4 4
+ ε LW ↓−σTS
We need to estimate E U
( ) and Var U
( ).

Components of the Sampling Error
We decompose U as the sum of 5 components
U = U 1 + ... U 5
Where each component is of the form:
p × ( FGERB – F ) AMF or F × ( pGERB – p ) AMF
where F is a flux and p a surface parameter. For example
U1 = ( 1−a ) ( AMF SW ↓AMF −SW ↓ ) GERB
U2 = SW ↓ ( GERB aGERB − aAMF), etc.
(Random)
(Systematic)

Albedo variability within GERB pixels, and
GERB–AMF albedo differences
(1 GERB pixel = 10 4 MODIS pixels)

Sampling errors (“U”) by month
Albedo U1 U2 U3 U4 U5 U
Jan 0.252 ±2.06 8.45 ±1.70 -4 0 4.45±2.26
Feb 0.255 ±2.35 10.04 ±1.86 -4 0 6.04±2.21
Mar 0.266 ±3.44 10.58 ±2.30 -4 0 6.58±3.20
Apr 0.259 ±3.85 11.45 ±2.85 -4 0 7.45±3.40
May 0.234 ±7.80 10.39 ±2.66 -4 0 6.39±7.89
Jun 0.226 ±5.60 10.07 ±2.32 -4 0 6.07±6.03
Jul 0.222 ±9.12 9.34 ±1.64 -4 0 5.34±9.34
Aug 0.196 ±12.00 8.87 ±1.63 -4 0 4.87±12.2
Sep 0.200 ±6.89 9.06 ±2.65 -4 0 5.06±7.00
Oct 0.226 ±4.77 10.08 ±2.15 -4 0 6.08±4.66
Nov/Dec 0.248 ±3.25 9.67 ±1.29 -4 0 5.67±3.10
Random part of the sampling error vs day-to-day SRB variability
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
SRB 52.7 70.1 73.4 86.7 96.1 114 116 122 116 101 65.0 45.9
σ(SRB) 7.12 6.7 9.9 8.0 20.8 20.6 38.6 40.4 37.2 15.8 6.6 4.4
U 2.26 2.21 3.20 3.40 7.89 6.03 9.34 12.2 7.00 4.66 5.67 -

Correcting for the systematic
albedo error
Why not correct the upwelling short wave
flux observed using the average MODIS
albedo value over the GERB pixel?

How well does MODIS capture
the spatial variability in albedo?
DABEX (Dust and biomass experiment) January 2006
FAAM = FACILITY for AIRBORNE ATMOSPHERIC MEASUREMENTS

Flight from NIA to BAN, 500ft 10:30 19th January 2006
BB Albedo calculated every second

FAAM albedo vs MODIS albedo
NB: “Adjustment” is to increase the MODIS value by 13-14%

Albedo comparisons
through the year
(2006).
Black figures are
MODIS;
the red line is the
median observed value
over the same period,
the blue lines are the
minimum and
maximums in each
period. (AMF albedos
calculated over 11:00-
13:00)

Daily albedo at NIA airport
(total all-day SW↑÷ total SW↓)
Possible bias if
cloudy days’
albedos are very
different from
those on clear
days

Treatment of the Land Surface in
the Unified Model
Original scheme: MOSES (Met Office Surface Exchange
Scheme). “Effective” parameters defined at scale of
atmospheric grid cell.
2001- MOSES2. Land surface in each cell is now a mosaic
of smaller tiles with different surface cover types. Fluxes
calculated for each tile and combined.

Four options:
Albedo from file
Albedo in the UM
Albedo for each tile: either fixed as a function of cover
type, or as a function of LAI and SZA using Sellers’ 2stream
model. LAI can be imposed, or generated from a
dynamic vegetation model (TRIFFID in the case of the
UM). Soil albedo given as fn(soil colour).
Landcover from file, albedo as fn(landcover)
LAI from file, albedo from canopy RT model (currently Sellers
1985) as fn(LAI, SZA, soil colour, landcover)
Dynamic Veg. Model –> LAI –>Albedo (canopy RT)

JULES
Further development of the Land component of the Met
Office/Hadley Centre GCM takes place in a complex
scheme known as JULES (Joint UK Land Environment
Simulator).
Based on MOSES2, as currently used in the Unified
Model.
Being developed to model the carbon cycle, the nitrogen
cycle, hydrology applications, crop modelling,etc. and
general Earth System Modelling.
Open to ‘the community’, code is free.
http://www.jchmr.org/jules/

The Spectral Simulator
This is a collection of ‘observation operators’ being
developed to predict top-of-canopy and top-ofatmosphere
radiances for a given set of surface
parameters. Predicts surface reflectance/albedo,
temperature (emissivity?) and microwave emission.
Comparisons with albedo/surface reflectance products
will be critical. (Timescale ~ 3-4 years)
BRF product (directional reflected radiance) might be
even more important for exercising surface models.
The JSS is an important part of the Land Data
Assimilation Scheme under development for JULES

Value of GlobAlbedo to UKMO
Adding to long - term climate record
Model evaluation
- resolving diurnal cycle of net radiation
- seasonal cycle of albedo - large uncertainties
between models.
- improvement of JULES, the surface scheme
(MODIS already used for this)
Potential for real-time assimilation?
Data for reanalyses

Incorporation of improved albedo in the Hadley Centre Model
From CLASSIC report, 2006

From CLASSIC report, 2006

Improved Albedo in the Hadley centre GCM
From CLASSIC report, 2006

Requirements of the UKMO for GlobAlbedo
Spatial
Resolution
Temporal
Resolution
Spectral
Resolution
Directional
requirement
NWP Climate
1km resolution
for some runs
NRT?
Hourly radiation
calculations
BB, VIS/NIR BB, VIS/NIR
(defn of NIR?)
N216 (~60km) in 3-5 years time.
10 day average, and 10 day average of diurnal cycle
(hourly radiation calculation in near future)
Nothing Calibration of the vegetation sub-models in JULES
Other format: netcdf, please.
price: free, please
Well-characterised errors and unce rtainties must be provided for all
products (esp. for NRT DA)

Conclusions
• Experience of the MODIS albedo has been very
positive (RADAGAST, calibration of MOSES-2).
• UKMO very interested in GlobAlbedo. Will be
used
– in climate studies for: calibration, initialisation,
validation of JULES
– in high resolution mode as ancillary data for NWP
– via DA in NRT(?)
– in reanalyses
• Products at a range of resolutions (1- 60km,
90km).
• Product should factor in the diurnal cycle.
• Products must have well characterised errors.