ArticlePDF Available

Abstract and Figures

This article provides a description of a land surface temperature (LST) data set generated (and provided in near-real-time or offline) based on infrared data from sen-sors onboard different geostationary (GEO) satellites: Meteosat Second Generation (MSG), Geostationary Operational Environmental Satellite (GOES), and Multifunction Transport Satellite (MTSAT). Given the different characteristics of the imagers onboard each GEO platform, different algorithmic methodologies for the retrieval of LST are presented and implemented – namely the Generalized Split-Window (GSW) algo-rithm and the Dual Algorithm (DA) in its mono-and dual-channel forms – using semi-empirical functions that relate LST to top-of-atmosphere brightness temperatures in infrared window channels. The assumptions and physics underlying each method-ology, as well as the uncertainties of LST estimates, are discussed. The formulations are trained using a data set of radiative transfer simulations for a wide range of atmo-spheric and surface conditions. The performance of each algorithm is then assessed by comparing its output against an independent set of simulations, suggesting that prod-uct uncertainties range from 2 • C (for GSW and the two-channel algorithm) to 4 • C (for the one-channel algorithm). Finally, LST retrievals from different GEO satellites are merged into a single field. In overlapping areas, the average discrepancies between LST products derived from GOES and from the Spinning Enhanced Visible and Infrared Imager (SEVIRI) onboard MSG are within 1 • C during night-time, but may reach 3 • C during daytime. Over those areas, the merged LST field is obtained as a weighted aver-age of available LST retrievals for the same time slot, taking into account the respective error bar.
Content may be subject to copyright.
This article was downloaded by: [b-on: Biblioteca do conhecimento online UL]
On: 26 December 2012, At: 12:37
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
International Journal of Remote
Sensing
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/tres20
Land surface temperature from
multiple geostationary satellites
Sandra C. Freitas a , Isabel F. Trigo a b , João Macedo a , Carla
Barroso a , Ricardo Silva a & Rui Perdigão a b
a Instituto de Meteorologia, 1749-077, Lisbon, Portugal
b Instituto Dom Luiz, 1749-016, Lisbon, Portugal
Version of record first published: 30 Oct 2012.
To cite this article: Sandra C. Freitas , Isabel F. Trigo , João Macedo , Carla Barroso , Ricardo
Silva & Rui Perdigão (2012): Land surface temperature from multiple geostationary satellites,
International Journal of Remote Sensing, DOI:10.1080/01431161.2012.716925
To link to this article: http://dx.doi.org/10.1080/01431161.2012.716925
PLEASE SCROLL DOWN FOR ARTICLE
Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-
conditions
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation
that the contents will be complete or accurate or up to date. The accuracy of any
instructions, formulae, and drug doses should be independently verified with primary
sources. The publisher shall not be liable for any loss, actions, claims, proceedings,
demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
International Journal of Remote Sensing
iFirst, 2013, 1–18
Land surface temperature from multiple geostationary satellites
Sandra C. Freitasa,IsabelF.Trigo
a,b*, João Macedoa,CarlaBarroso
a,RicardoSilva
a,
and Rui Perdigãoa,b
aInstituto de Meteorologia, 1749-077 Lisbon, Portugal; bInstituto Dom Luiz, 1749-016 Lisbon,
Portugal
(Received 13 January 2011; accepted 25 October 2011)
This article provides a description of a land surface temperature (LST) data set
generated (and provided in near-real-time or offline) based on infrared data from sen-
sors onboard different geostationary (GEO) satellites: Meteosat Second Generation
(MSG), Geostationary Operational Environmental Satellite (GOES), and Multifunction
Transport Satellite (MTSAT). Given the different characteristics of the imagers onboard
each GEO platform, different algorithmic methodologies for the retrieval of LST are
presented and implemented – namely the Generalized Split-Window (GSW) algo-
rithm and the Dual Algorithm (DA) in its mono- and dual-channel forms – using
semi-empirical functions that relate LST to top-of-atmosphere brightness temperatures
in infrared window channels. The assumptions and physics underlying each method-
ology, as well as the uncertainties of LST estimates, are discussed. The formulations
are trained using a data set of radiative transfer simulations for a wide range of atmo-
spheric and surface conditions. The performance of each algorithm is then assessed by
comparing its output against an independent set of simulations, suggesting that prod-
uct uncertainties range from 2C (for GSW and the two-channel algorithm) to 4C (for
the one-channel algorithm). Finally, LST retrievals from different GEO satellites are
merged into a single field. In overlapping areas, the average discrepancies between LST
products derived from GOES and from the Spinning Enhanced Visible and Infrared
Imager (SEVIRI) onboard MSG are within 1Cduringnight-time,butmayreach3
C
during daytime. Over those areas, the merged LST field is obtained as a weighted aver-
age of available LST retrievals for the same time slot, taking into account the respective
error bar.
1. Introduction
Land surface temperature (LST) is an important component of the surface radiation bud-
get, since, together with surface emissivity, it controls the upward thermal radiation.
Moreover, LST is intrinsically linked to the partition between sensible and latent heat fluxes
between the surface and the atmosphere (e.g. Dickinson et al. 1991; Caparrini, Castelli,
and Entekhabi 2004), making this a key variable for a wide range of applications, most
notably (i) general model evaluation, either land surface or numerical weather prediction
(NWP) models (Trigo and Viterbo 2003; Mitchell et al. 2004); (ii) data assimilation in
(NWP or land surface) models (e.g. Caparrini et al. 2004; Bosilovich et al. 2007; Qin et al.
2007; Saux-Picart et al. 2009; Ghent et al. 2010); (iii) hydrological applications (Kustas
and Norman 1996; Wan, Wang, and Li 2004; Miglietta et al. 2009; Lacaze et al. 2010); and
*Corresponding author. Email: isabel.trigo@ipma.pt
ISSN 0143-1161 print/ISSN 1366-5901 online
©2013Taylor&Francis
http://dx.doi.org/10.1080/01431161.2012.716925
http://www.tandfonline.com
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
2S.C. Freitas et al.
(iv) climate and environmental monitoring (Jin 2004; Jin, Dickinson, and Zhang, 2005;
Yu, Prive t t e , a n d P i n h e i r o 2 0 0 8 ; Settle et al. 2009). Many of those applications require
global coverage, high-to-moderate spatial resolution, and high-frequency sampling, which
can only be achieved through the use of remotely sensed data.
The choice between LST retrievals from polar orbiters and geostationary (GEO) plat-
forms often corresponds to a compromise between high spatial resolution and global
coverage versus high temporal sampling. A proper characterization of the diurnal cycle can
only be provided by GEO satellites, with the exception of the polar regions where revisit
times of polar orbiters are generally short. Moreover, since LST retrievals from thermal
infrared (TIR) imagers are restricted to clear sky conditions, low observation frequencies
may seriously limit LST availability in regularly cloud covered areas.
Several methodologies have been proposed in the last two decades to estimate LST
from remotely sensed data. Most of these are obtained from measurements of one or more
channels within the TIR atmospheric window from 8 to 13 µm and are limited to clear
sky pixels (Dash et al. 2002). Operational LST retrievals often make use of split-window
algorithms (e.g. Prata 1993; Wan and Dozier 1996; Coll and Caselles 1997). Other studies,
however, have assessed the combination of middle infrared (MIR) bands centred on 3.9
µm(TMIR)withthesplitwindowwithinthe1013µmband(TTIR1 and TTIR2 ,Pinkeretal.
2007; Sun and Pinker 2007), an approach that can be particularly useful in the absence of
two TTIR bands.
An LST product for the area covered by Meteosat Second Generation (MSG, centred at
0longitude) is currently distributed by the EUMETSAT Satellite Application Facility on
Land Surface Analysis (LSA SAF). The LSA SAF generates, archives, and disseminates
LST from SEVIRI (onboard MSG) with a 15 minute frequency at the original satellite spa-
tial resolution (Trigo et al. 2011). The main objective of the work leading to this article was
to extend the spatial coverage of the LSA SAF LST product to areas covered by other GEO
satellites, namely the Geostationary Operational Environmental Satellite (GOES) and the
Multifunction Transport Satellite (MTSAT), centred at 60Wand145
E, respectively. The
merged LST product, obtained from the combination of retrievals from these three GEO
satellites, partially overcomes the area coverage restrictions associated to GEO platforms,
while maintaining a high (hourly) temporal sampling. The GEO LST product presented
here has been generated since August 2009 and can be made available either in near-real-
time (NRT) or offline for the full period (from August 2009 to present) upon request; the
full data set will soon be made available via the Geoland2 portal (Lacaze et al. 2010; http://
www.geoland2.eu/).
This article is structured as follows: the LST algorithms adapted for GOES and MTSAT
imagers are presented in Section 2; the uncertainty associated with the LST estimations is
addressed in Section 3; Section 4 addresses the consistency of the LST algorithms, whereas
Section 5 outlines the data fusion procedure used to obtain the quasi-global LST product.
The article is closed with concluding remarks presented in Section 6.
2. LST algorithms
2.1. Basic underlying assumptions
The LST algorithms used in this study are all designed for top-of-atmosphere (TOA) TIR
and MIR observations within the atmospheric window bands. One of the most relevant fac-
tors in algorithm selection is their expected reliability for operational LST retrievals, both in
terms of accuracy and timeliness. The latter favours the use of semi-empirical relationships
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
International Journal of Remote Sensing 3
between LST and TOA brightness temperatures, which are computationally efficient and
free of the convergence problems of direct emissivity and temperature retrieval methods
(e.g. Faysash and Smith 1999).
The use of two pseudo-continuous channels in the TIR 10–13 µmrange–thesplit-
window channels – for atmospheric correction has proved to be one of the most efficient
methodologies for LST estimation (Prata 1993; Wan and Dozier 1996; Dash et al. 2002).
Other studies have assessed other approaches based on the combination of MIR bands
(centred at 3.9 µm; MIR) with observations in one or two split-window bands (TIR1 and
TIR2; e.g. Pinker et al. 2007; Sun and Pinker 2007). Although there are several caveats
regarding the use of MIR channels for LST operational retrievals, they can become a very
useful source of information, particularly in the absence of two TIR bands. The major prob-
lems regarding the use of TMIR include (Freitas et al. 2010) (i) the uncertainty of surface
emissivity within MIR, which is considerably higher than that of the TIR channels, partic-
ularly over semi-arid regions (Trigo et al. 2008b); and (ii) solar contamination of daytime
TMIR observations must be taken into account.
The ultimate goal of this study is to describe the LST data set currently under production
within the framework of Geoland-2 (Lacaze et al. 2010). This corresponds to the fusion of
LST fields obtained from three GEO satellites – GOES, MSG, and MTSAT. The design of
the algorithms applied to each individual sensor takes into account the channels available
from each sensor, as described in Table 1.
The algorithms developed for GEO satellites are divided into three groups: (i) split-
window methodologies, which make use of two adjacent window channels within the TIR
range (TIR1 and TIR2); (ii) two-channel algorithms, which derive LST from one window
channel in the TIR (TIR1, around 11 µm) and another in the MIR (around 3.9 µm), to be
used when only one TIR channel is available – and for night-time conditions, when TMIR
is not contaminated by solar radiation reflected by the surface; and (iii) the mono-channel
method that corrects the TOA brightness temperature of a single channel, TTIR1,foratmo-
spheric attenuation and surface emissivity; this algorithm is used for daytime conditions,
when only one TIR channel is available. The split-window methodology is applied to the
SEVIRI data, while for MTSAT and GOES observations two different approaches are used:
the two-channel algorithm for night-time and one-channel algorithm for daytime. It should
be noted here that the goal of this study is to develop a set of algorithms that can be used
for the NRT generation of LST based on MTSAT, GOES, and SEVIRI data. Although the
MTSAT imager has two channels in the split-window band (Table 1), only one is currently
being disseminated in NRT to European countries. Given this limitation, we do not consider
here the use of a split-window methodology for MTSAT.
Table 1. Thermal and middle infrared (TIR and MIR) channels
(wavelengths in µm) available from each sensor, considered here
for LST estimation.
SEVIRI GOES imager MTSAT
MIR 3.5–4.4 3.8–4.0 3.5–4.0
TIR1 10.0–11.5 10.2–11.2 10.3–11.3
TIR2 11.2–12.8 11.5–12.5*
Note: *Not disseminated via EUMETCast and therefore not available in
near-real time at the moment.
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
4S.C. Freitas et al.
All aforementioned methodologies are based on semi-empirical formulations, where
LST is expressed as a regression function of TOA brightness temperatures. The algorithms
should minimize LST uncertainties and limit the impact of input error propagation. The for-
mulations are trained for different classes of satellite view angle, atmosphere water vapour
content, and land-cover type.
2.2. Generalized split-window algorithm – SEVIRI LST
The LSA SAF routinely produces LST from SEVIRI onboard MSG. The algorithm, based
on the GSW technique first developed by Wan and Dozier (1996) for the Advanced Very
High Resolution Radiometer (AVHRR) and MODIS, was calibrated for SEVIRI measure-
ments. A full description of the algorithm, product validation, and expected accuracy are
available in Trigo et al. (2008a, 2008b) and Freitas et al. (2010). LST data provided by
the LSA SAF, along with respective user documentation, may be downloaded from http://
landsaf.meteo.pt.
The accuracy of LST retrievals using the GSW algorithm varies considerably with the
satellite viewing angle and atmospheric water vapour content (Trigo et al. 2008a; Freitas
et al. 2010). In the most extreme cases, when the surface is observed through moist atmo-
spheres and large view angles leading to uncertainties of the order of 4Corhigher,LST
retrievals are masked out.
The analysis of algorithm uncertainties is complemented with validation against inde-
pendent sources of data, including in situ observations and LST products retrieved from
other sensors (Trigo et al. 2008a). As shown in Freitas et al. (2010), there is a good agree-
ment with in situ measurements taken in semi-arid regions. For comparisons made between
the LSA SAF SEVIRI LST product and ground observations taken at Gobabeb (Namibia)
over o ne full year period (May 2 008–April 2009), root mean square differences ( RMS)
ranged between 1Cand2
C, while biases varied between –1Candlessthan+0.5C. The
monthly uncertainties estimated for the LST product taking into account expected algo-
rithm and input accuracies lay between 2Cand3
Cforthesameperiod.Thisrangeisof
the same order or even larger than RMS found between the in situ observations and the
satellite estimations (Trigo et al. 2008a; Freitas et al. 2010).
2.3. Dual Algorithm – GOES and MTSAT LST
The Dual Algorithm (DA) was designed for sensors with only one channel in the thermal
window and another in the MIR. The use of the TMIR observations for LST retrieval implies
two important sources of uncertainty, namely the relatively high emissivity range of natural
surfaces within that band and the contamination of TOA observations by solar radiation
reflected at the surface. To avoid the latter, TMIR observations are only used during night-
time. Thus, the DA provides estimation of LST using (i) a two-channel algorithm applied to
measurements from one MIR and one thermal window (TIR1) channel during night-time
and (ii) a mono-channel method requiring a single thermal window channel, TIR1, during
daytime.
AversionoftheDAwasadjustedtoGOESandMTSATimagers,takingintoaccount
the response functions of the respective TIR and MIR channels. Since it is not possible
to currently implement a GSW algorithm for MTSAT, the DA is the formulation used to
estimate MTSAT-based LST in NRT.
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
International Journal of Remote Sensing 5
Following the approach prese nted in Sun, Pinker, and Basara ( 2004), the mono -channel
and two-channel algorithms developed for GOES and MTSAT, respectively, are based on
Equations (1) and (2):
–mono-channelalgorithm
LST =A1+A2TTIR1 +!(LST), (1)
–two-channelalgorithm
LST =B1+B2TTIR1 +B3(TTIR1 TMIR)+!(LST), (2)
where !(LST) is the respective model error and Aiand Bi(i=1, 2, 3) are the model
coefficients. As previously done in the case of the GSW algorithm, the coefficients in
Equations (1) and (2) are fitted to a calibration data set for different classes of total col-
umn water vapour (TCWV), satellite zenith angle (SZA), and land-cover type, as defined
in the International Geosphere-Biosphere Programme (IGBP) database (Belward 1996).
Although not shown, the use of higher order terms did not lead to statistically signif-
icant performances (bias and root mean square error (RMSE) of the same order), or
improvements concerning the propagation of input uncertainties.
The calibration and verification of Equations (1) and (2) was primarily based on a set
of radiative transfer simulations, aimed at producing the TOA TTIR1 and TMIR values that
would be observed by GOES and MTSAT, respectively, for a wide range of surface and
atmospheric conditions. Such simulations were performed using the Moderate Spectral
Resolution Atmospheric Transmittance (MODTRAN4) model (Berk et al. 1999), applied
to the database of clear sky profiles of temperature, humidity, and ozone compiled by
Borbas et al. (2005). The database compiles over 15,700 clear sky profiles taken from
other data sets, such as NOAA88 (Seemann et al. 2003), TIGR (Thermodynamic Initial
Guess Retrieval; Chedin et al. 1985) and TIGR-like (Chevallier 2002). The purpose of the
database is to allow radiative transfer model simulations over a wide variety of the distri-
bution of temperature and humidity profiles, from tropical to arctic atmospheres, and full
annual range (Borbas et al. 2005).
The parameters of Equations (1) and (2) used in the DA algorithm are estimated inde-
pendently for GOES and MTSAT data for each of the 16 land-cover types within the IGBP
database, for eight different classes of TCWV (up to 6 cm), and for 16 classes of SZA (up to
75). Following the methodology used for the SEVIRI GSW algorithm (Freitas et al. 2010),
two calibration data sets were built for GOES and MTSAT, respectively, corresponding to
MODTRAN simulations performed over a subset of 77 profiles chosen from Borbas et al.
(2005) to be representative of a wide range of atmospheric conditions. The simulations
were made for a large sample of surface and viewing conditions, since for each of these
profiles: (i) the surface temperature was set to values ranging between Tskin –15
Cand
Tskin +15C, in steps of 5C, where Tskin is the profile surface or skin temperature; (ii)
emissivities for split-window and MIR channels covered a wide range of realistic values
characteristic of IGBP land-cover types (Trigo et al. 2008b); and (iii) SZA was set to range
from nadir to 75in steps of 5.Thisprocedureensuresthatalloftheabove-mentioned
classes are well populated. The assignment of emissivity ranges to each class of land cover
considers the dominant vegetation and bareground types within each class, which are con-
verted into characteristic vegetation and bareground emissivities, taking into account the
sensor response functions and laboratory spectral reflectance; further details may be found
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
6S.C. Freitas et al.
72.5
0 2.7
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
98
Explained variance (%)
96
94
92
90
80
Total column water vapour (cm) Total column water vapour (cm)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
2.4
2.1
1.8
1.5
1.2
0.9
A2
80
160
240
320
400
480
A1 (K)
Satellite Zenith Angle (°)
67.5
62.5
57.5
52.5
47.5
42.5
37.5
32.5
27.5
22.5
17.5
12.5
7.5
2.5
0.0
Total column water vapour (cm)
0.750.00 1.50 2.25 3.00 3.75 4.50 5.25
Figure 1. Distribution of the mono-channel algorithm parameters for a specific land-cover type
(croplands) and explained variance of the fitted regression (right) as a function of the SZA and
total column water vapour (cm). Classes for which the explained variance is below 85% and/or the
algorithm error exceeds 4C are masked out.
in Peres and DaCamara (2005) and Trigo et al. (2008b). Land-cover emissivities are esti-
mated as a weighted average of vegetation and bareground values, assuming a range of the
fraction of vegetation cover acceptable for each class.
The parameters obtained for GOES mono-channel and of the two-channel algorithms,
for a specific land-cover type (Croplands) are schematically presented in Figures 1 and 2,
respectively, along with the variance of LST explained by the regression. The algorithm
coefficients vary fairly smoothly among classes of water vapour and view zenith angle,
which contributes to less sharp differences in the transition from class to class. However,
in cases where very moist atmospheres are observed with high zenith angles, the semi-
empirical relations between channel brightness temperatures and LST does not reproduce
well the non-linear effects on the total path length. As a consequence, the explained
variance of surface temperature by TOA brightness temperatures decreases (last panel
in Figures 1 and 2) and the uncertainty of the LST algorithm increases. Classes with
explained variance below 85% are ignored. The uncertainty of the mono-channel (used
for daytime) and two-channel (used for night-time) formulations are assessed by compar-
ison with the verification data set, consisting of MODTRAN simulations performed for
all profiles in the Borbas et al. (2005) database not used for calibration purposes – leav-
ing a total of 15,623 cases. The distribution of errors is shown in Figures 3 and 4 for
GOES and in Figures 5 and 6 for MTSAT formulations. While the two-channel methodol-
ogy presents errors similar to those obtained for the GSW (Freitas et al. 2010) – frequently
within the 2K range – the use of a single channel is associated to a considerable uncer-
tainty increase. The strong contrast in the error distribution among Figures 3 and 4 and
Figures 5 and 6 presents evidence of the need for more than one channel for a proper atmo-
spheric correction. As a consequence, the combination of moist/view angle conditions
where the requirement for LST uncertainty (4Corless)ismetissignicantlyreduced
for the mono-channel algorithm (Figure 1) when compared to the two-channel algorithm
(Figure 2).
On top of TOA brightness temperatures of involved channels, the operational imple-
mentation of the DA (Equations (1) and (2)) makes use of the following data for the
proper coefficient selection: (i) forecasts of TCWV obtained from the European Centre for
Median-range Weather Forecasts (ECMWF); (ii) land-cover classification of each pixel,
according to the IGBP database (Belward 1996); and (iii) the pixel viewing angle. It should
be noticed that the database described in Borbas et al. (2005) is used for algorithm calibra-
tion and verification, while ECMWF TCWV forecasts are used for the NRT production of
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
International Journal of Remote Sensing 7
Satellite Zenith Angle (°)
72.5
67.5
62.5
57.5
52.5
47.5
42.5
37.5
32.5
27.5
22.5
17.5
12.5
7.5
2.5
0.0
B1 (K) B2B3
0.750.00 1.50 2.25 3.00 3.75 4.50 5.25 0.750.00 1.50 2.25 3.00 3.75 4.50 5.25
98
96
94
92
90
80
0.750.00 1.50 2.25 3.00 3.75 4.50 5.250.750.00 1.50 2.25 3.00 3.75 4.50 5.25
Total column water vapour (cm) Total column water vapour (cm) Total column water vapour (cm)Total column water vapour (cm)
01.30
0.0
0.4
1.25
1.20
1.15
1.10
1.05
1.00
15
30
45
60
0.4
0.8
1.2
1.6
2.0
75
Explained variance (%)
Figure 2. As in Figure 1, but for the two-channel algorithm parameters.
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
8S.C. Freitas et al.
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
LSTSeeBor LSTGSW (K)
SZA:(0,2.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(2.5,7.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(7.5,12.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(12.5,17.5)
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
LSTSeeBor LSTGSW (K)
SZA:(17.5,22.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(22.5,27.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(27.5,32.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(32.5,37.5)
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
LSTSeeBor LSTGSW (K)
SZA:(37.5,42.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(42.5,47.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(47.5,52.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(52.5,57.5)
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
LSTSeeBor LSTGSW (K)
SZA:(57.5,62.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
SZA:(62.5,67.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
SZA:(67.5,72.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
SZA:(72.5,77.5)
Figure 3. Distribution of LST errors, defined as the difference between LSTSeeBor (i.e. the surface
temperature used for MODTRAN simulations) and LSTGSW (surface temperature obtained with the
GSW algorithm), obtained for the GOES mono-channel algorithm through comparison with the veri-
fication database. The distributions are shown for different classes of satellite zenith angle (indicated
in the bottom left of each panel – values are in degrees) and water vapour content, W(plotted on the
x-axis in each diagram). The lines within each box plot correspond to the lower quartile, median, and
upper quartile, respectively, while the whiskers extend to remaining data.
LST. The ECMWF data are received in an operational mode and correspond to a 3 hourly
forecast with steps between 12 and 36 hours (maximum) available globally with a spatial
resolution of about 16 km. The fields are linearly interpolated in time to hourly and bi-
linearly interpolated in space to pixel resolution. GOES and MTSAT are also subject to
acloudmaskprocedure,basedonthealgorithmproposedbyDerrienandGléau(2005),
since the DA is applicable to clear sky pixels only. As addressed next, the uncertainty of
each LST value is also estimated, taking into account the uncertainty of the respective
algorithm (Figures 3–6) and that of the input data.
3. Error analysis
The assessment of an error bar for each estimated LST value (GOES or MTSAT based)
considers the uncertainties of the retrieval algorithm, heavily dependent on the total optical
path between the sensor and the surface, which is essentially determined by the viewing
geometry and the total water vapour content (Figures 3–6). On top of that, LST uncertainty
estimates consider the propagation of input errors, namely (i) sensor noise, (ii) uncertainties
in surface emissivity (for GSW algorithm) or variability within the land-cover type, and
(iii) statistics of TCWV forecast errors. The assessment of the impact of input errors on
LST estimation is assessed through sensitivity tests, as detailed in Freitas et al. (2010)
for the GSW. For this purpose, LST retrievals from simulated radiances, using inputs to
DA with realistic errors (i.e. within the respective error distribution) are compared with the
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
International Journal of Remote Sensing 9
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
LSTSeeBor LSTGSW (K)
SZA:(0,2.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(2.5,7.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(7.5,12.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(12.5,17.5)
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
LSTSeeBor LSTGSW (K)
SZA:(17.5,22.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(22.5,27.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(27.5,32.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(32.5,37.5)
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
LSTSeeBor LSTGSW (K)
SZA:(37.5,42.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(42.5,47.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(47.5,52.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(52.5,57.5)
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
LSTSeeBor LSTGSW (K)
SZA:(57.5,62.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
SZA:(62.5,67.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
SZA:(67.5,72.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
SZA:(72.5,77.5)
Figure 4. As in Figure 3, but for the two-channel algorithm developed for GOES MIR and TIR
channels.
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
LSTSeeBor LSTGSW (K)
SZA:(0,2.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(2.5,7.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(7.5,12.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(12.5,17.5)
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
LSTSeeBor LSTGSW (K)
SZA:(17.5,22.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(22.5,27.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(27.5,32.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(32.5,37.5)
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
LSTSeeBor LSTGSW (K)
SZA:(37.5,42.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(42.5,47.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(47.5,52.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(52.5,57.5)
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
LSTSeeBor LSTGSW (K)
SZA:(57.5,62.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
SZA:(62.5,67.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
SZA:(67.5,72.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
SZA:(72.5,77.5)
Figure 5. As in Figure 3, but for the mono-channel algorithm developed for the MTSAT 11 µm
channel.
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
10 S.C. Freitas et al.
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
LSTSeeBor LSTGSW (K)
SZA:(0,2.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(2.5,7.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(7.5,12.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(12.5,17.5)
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
LSTSeeBor LSTGSW (K)
SZA:(17.5,22.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(22.5,27.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(27.5,32.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(32.5,37.5)
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
LSTSeeBor LSTGSW (K)
SZA:(37.5,42.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(42.5,47.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(47.5,52.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
SZA:(52.5,57.5)
5
0
5
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
LSTSeeBor LSTGSW (K)
SZA:(57.5,62.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
SZA:(62.5,67.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
SZA:(67.5,72.5)
0.00 0.75 1.50 2.25 3.00 3.75 4.50 5.25
W (cm)
SZA:(72.5,77.5)
Figure 6. As in Figure 3, but for the two-channel algorithm developed for MTSAT MIR and TIR
channels.
012345
012345
012345
(c)(b)(a)
Figure 7. Maps of LST error bar (C) for 21 July 2010 at 12:00 UTC, over the (a) GOES disc, using
the mono-channel algorithm; (b) MSG disc, using the GSW algorithm; and (c) MTSAT disc, using
the two-channel algorithm. The coordinates of each sub-satellite point (latitude, and longitude in E)
are also indicated.
original surface temperature used in the simulations. Figures 7(a)–(c)showexamplesofthe
error bars associated with LST retrieved with the mono-channel algorithm, GSW algorithm,
and two-channel algorithm, respectively; panel (b)showstheestimateduncertaintyofLST
retrieved from SEVIRI/MSG using the GSW algorithm.
It should be noted that despite the large differences between LST algorithms used
(mono-channel in the case of GOES and MTSAT in panels (a)and(c), respectively,
and GSW in the case of SEVIRI/MSG in panel (b)), the error bars shown in Figure 7
present similar values. Uncertainties higher than 2Candeven3
Cobservedover
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
International Journal of Remote Sensing 11
northern Africa are associated to relatively high uncertainties in surface emissivity, fur-
ther enhanced by low atmospheric water content characteristic of that region. Pixels with
LST error bars higher than 4Caremaskedout,althoughusersmaychoosestricter
values.
4. Consistency analysis of the LST algorithms
The algorithms developed for different GEO satellites are verified for consistency in over-
lapping areas, which limits this analysis to LST retrievals from SEIVIRI/MSG and GOES.
Since the respective LST products are obtained using independent data sources and method-
ologies, the consistency between them may be regarded as an indirect assessment of LST
uncertainty. This is then a first step towards the full validation of LST products, a process
that necessarily requires the extension of LST verification against in situ data, part of future
work.
Here we show the results of the comparison between LST estimated with GSW with that
estimated with the DA over the GOES/MSG overlapping region (South America). Both
LST fields are projected onto the plate carrée projection with a resolution of 0.05×0.05.
The comparison is performed for pixels classified as ‘clear sky’ for both satellites over land
areas, where these have similar satellite view angles, and for two months considered to be
representative of two contrasting seasons: January and July 2010.
The differences between the LST fields retrieved from MSG/SEVIRI and GOES
imagery are shown in Figures 8(a)and(c)(for16January2010)andFigures9(a)and(c)
(for 16 July 2010), for two time slots representative of night-time and daytime, respectively;
panels to the right show the respective scatter plots of MSG LST versus GOES LST.
The scatter plots in Figures 8 and 9 reveal that LST exhibits higher dispersion at day-
times than at night-times. The range of LST values is larger in July than in January, since
the former encompasses retrievals from the tropics and mid-latitude winter areas. In both
months, however, GOES LST tends to be warmer than that obtained from SEVIRI/MSG.
Even so, the highest point density stretches along the 1:1 line, suggesting good consistency
between the two LST fields.
The discrepancies between the two fields are mostly within GOES (and to a lesser
extent SEVIRI/MSG) LST error bars, further supporting the consistency between both
retrievals and the robustness of the algorithm parameters. This is illustrated in Figure 10,
which shows the daily cycle of LST from both sensors for a specific location in a flat
region located in northern Brazil (6.75S; 41.6W). Figure 10 also points towards a marked
diurnal cycle of the differences among the two sensors. Larger discrepancies are obtained
close to local noon (time slots between 15 and 18:00 UTC), when a higher uncertainty
method is used for LST retrieval from GOES (the mono-channel algorithm) and LST spatial
variability is higher. The impact of surface anisotropy on the two LST retrievals is further
enhanced by the nearly symmetric view zenith angles of GOES and MSG over the South
American region under study (e.g. Barroso et al. 2005; Pinheiro, Privette, and Guillevic
2006; Rasmussen et al. 2010).
The distributions of the differences between GOES and SEVIRI/MSG LST within the
over lapping area are shown in Figur e 11 for the full period under analysis. The spread of
LST discrepancies is larger for January than July, and for daytime than night-time slots, as
already suggested by Figures 8 and 9. During the night period, median and upper quar-
tile values are less than 1Cand2
C, respectively. These are very close to the upper
limit of LST error bars for that region (see example shown in Figure 7), meaning that
the discrepancies are mostly consistent with product uncertainties.
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
12 S.C. Freitas et al.
30.0
20.0
10.0
0.0
10.0
20.0
30.0
40.0
30.0
20.0
10.0
0.0
10.0
20.0
30.0
40.0
70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0
70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0
4.0
40
40
40
40
30
30
30
30
20
20
20
20
MSG LST (°C)
MSG LST (°C)
GOES LST (°C)
GOES LST (°C)
10
10
10
10
0
0
0
0
3.2
2.4
1.6
0.8
0.0
0.8
1.6
2.4
3.2
4.0
4.0
3.2
2.4
1.6
0.8
0.0
0.8
1.6
2.4
3.2
4.0
Latitude (°)
Latitude (°)
Longitude (°)
Longitude (°)
(a)
(c)(d)
(b)
Figure 8. (a,c)Spatialdistributionofdifferences(
C) between the LST retrieved with
MSG/SEVIRI and the LST retrieved with GOES imagery, both projected onto the plate carrée pro-
jection with a resolution of 0.05×0.05, for 16 January 2010. (b,d) Scatter plot of LST from
MSG/SEVIRI against LST from GOES, also for 16 January 2010. Cases (a,b)and(c,d)referto
03:00 UTC and 15:00 UTC, respectively.
The distribution of LST uncertainties shows considerably higher variability during
daytime (09:00–21:00 UTC). The minimum differences (median value of about –1Cin
January and 0CinJuly)attainedat12:00UTCarecompatiblewiththeSun-satellite
geometry, favouring SEVIRI to view a larger portion of sunlit surfaces than GOES. The
opposite holds for the afternoon part of the day, in accordance with the warmer GOES bias
when compared with SEVIRI between 15 and 21:00 UTC. During daytime, it becomes
particularly difficult to distinguish among the different causes for LST discrepancies,
such as algorithms used, higher LST spatial variability, and differences in TOA obser-
vations. The comparison between GOES and SEVIRI/MSG LST retrievals is consistent
with previous studies that focused on the impact of observation angles on LST estima-
tions (Barroso et al. 2005; Pinheiro, Privette, and Guillevic 2006; Trigo et al. 2008b;
Rasmussen et al. 2010). Both surface and TOA measurements are strongly influenced by
surface heterogeneity, which in turn depends strongly on dominant surface types and terrain
orography.
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
International Journal of Remote Sensing 13
(c)
30.0
20.0
10.0
0.0
10.0
20.0
30.0
40.0
30.0
20.0
10.0
0.0
10.0
20.0
30.0
40.0
70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0
70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0
4.0
40
40
40
40
30
30
30
30
20
20
20
20
MSG LST (°C)MSG LST (°C)
GOES LST (°C)
GOES LST (°C)
10
10
10
10
0
0
0
0
3.2
2.4
1.6
0.8
0.0
0.8
1.6
2.4
3.2
4.0
4.0
3.2
2.4
1.6
0.8
0.0
0.8
1.6
2.4
3.2
4.0
Latitude (°)
Latitude (°)
Longitude (°)
Longitude (°)
(a)
(d)
(b)
Figure 9. As Figure 8, but for July 2010. (a)and(b) are for 03:00 UTC and (c)and(d)arefor
15:00 UTC.
50
40
30
20
10
0
00 03 06 09 12 15 18 21 00 03 06 09 12 15 18 21
50
40
30
20
10
0
LST (°C)
LST (°C)
MSG
GOES
Night
MSG
GOES
Night
Time (UTC) Time (UTC)
(a)(b)
Figure 10. LST values and error bars for 3 h spaced time slots throughout a chosen day: (a)
16 January, 6.75S, 41.60W and (b) 16 July 2010, 6.75
S, 41.60W. S q u a r e s a n d t r i a n g l e s
correspond to the LST from MSG/SEVIRI and from GOES, respectively. Night-times are shaded.
5. LST from a constellation of geostationary satellites
The algorithms described in Section 3 allow the estimation of LST on a pixel-by-pixel basis,
corresponding to instantaneous fields produced every 15 minutes in the case of SEVIRI
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
14 S.C. Freitas et al.
(a)(b)
10
8
6
0 3 6 9 12 15 18 21 24
4
2
0
2
4
6
8
10
Time (UTC)
0 3 6 9 12 15 18 21 24
Time (UTC)
(GOES LST) – (MSG LST) (°C)
10
8
6
4
2
0
2
4
6
8
10
(GOES LST) – (MSG LST) (°C)
Figure 11. Distribution of the differences between GOES minus SEVIRI LST for the full month of
(a) January 2010 and (b) July 2010, presented for 3 hourly time slots. The lines within each box plot
correspond to the lower quartile, median, and upper quartile, respectively, while the whiskers extend
to remaining data.
0 10 20 30 40 50 60
10 0 10 20 30 40 50 60
10 0 10 20 30 40 50 60
10
(c)(b)(a)
Figure 12. LST (C) estimated for 21 July 2010 at 12:00 UTC over the (a) GOES disc, using the
mono-channel algorithm; (b) MSG disc, using GSW algorithm; and (c) MTSAT disc, using the two-
channel algorithm. The coordinates of each sub-satellite point (latitude, and longitude in E) are also
indicated.
and hourly in the case of GOES and MTSAT. LST data fusion consists of re-projecting and
merging LST retrievals from the different GEO satellites (GOES, MSG, and MTSAT) into
asinglefield,yieldingproductcoveringmostoftheAmericancontinent,Europe,Africa,
part of Asia, and Australia (Figure 12).
The final LST product is obtained by averaging all of the LST values from pixels within
agridboxof0.05
×0.05(regular in latitude and longitude). The number of pixels within
agridboxdependsonthe(i)distancetothesub-satellitepoint(increasingwithdecreasing
distances) and (ii) number of pixels that were effectively processed (no clouds, nor missing
input data). In this context, all of the averages are weighted, taking into account the respec-
tive error bar so that the final product is less influenced by low-quality retrievals. At the
end, four data layers are added to the LST product with additional information crucial for
the product applications: (i) the error bar on the product grid; (ii) the mean value of the
time of acquisition; (iii) the fraction of pixels that were effectively processed; and (iv) a
quality flag with satellite information, missing data, cloudiness, and land–water mask.
Figures 12(a)–(c) show examples of the LST retrieved with the mono-channel algo-
rithm (over the GOES disc), the GSW algorithm (over the MSG disc), and the two-channel
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
International Journal of Remote Sensing 15
LST
150
80
60
40
20
0
20
40
60
80
100 50
20
10
0
0
10
20
30
40
50
60
50 100 150
Figure 13. LST product (C) estimated for 21 July 2010 at 12:00 UTC.
algorithm (over the MTSAT disc), respectively, for 21 July 2010 at 12:00 UTC. Figure 13
shows the merged LST product resulting from data fusion of the fields, as depicted in
Figure 12.
AfullyglobalLSTproductwillrequiredatafrompolar-orbitingsatellitesandfrom
aGEOsatellitepositionedovertheIndianregiontobeavailableafterthebeginningof
production in version 2 of the product.
6. Concluding remarks
Here we propose a methodology to generate high-frequency (hourly) LST with a large area
coverage, encompassing North and South America, Europe, part of Asia, and Australia,
through the use of a constellation of three GEO satellites – MSG, GOES, and MTSAT.
Given the difference in channels available from the sensors onboard these platforms, it is
necessary to apply different, but consistent, retrieval algorithms; these were calibrated and
validated using MODTRAN simulations of TOA brightness temperatures for a wide range
of atmospheric and surface conditions.
The use of the GSW algorithm is limited to SEVIRI/MSG, which has two adjacent
channels within the thermal window channel available for NRT processing. As for the
imagers onboard MTSAT and GOES, only one channel is available in such a region.
In those cases, LST is retrieved using an additional channel (positioned in the MIR region)
during the night-time and a single TIR channel during daytime. This strategy avoids the
solar contamination of MIR observations. Together with the LST field, gridded error bars
are also provided. These are estimated with a careful characterization of the algorithm
uncertainties and the uncertainty of the input variables and its propagation through the LST
algorithm, based on simulated radiances.
The consistency between LST retrievals from GOES and SEVIRI/MSG is assessed
for the overlapping region of the two discs. It is shown that, despite the differences in
sensors and algorithm, night-time average discrepancies are generally below 1C. The mean
and range of differences increase during daytime and in the warmer period of the year.
In these latter cases, it is particularly difficult to separate the influence of the sun-viewing
geometry from the differences in algorithms and input data on the final results, due to the
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
16 S.C. Freitas et al.
high heterogeneity of LSTs (Barroso et al. 2005; Pinheiro, Privette, and Guillevic 2006;
Rasmussen et al. 2010).
A complete assessment of the LST retrievals presented here requires their comparison
with an independent source, preferably obtained from in situ observations and retrievals
from other sensors. The comparison between GOES and SEVIRI/MSG-based LST values
may be regarded as a first exercise of the latter, which nevertheless needs to be extended
to longer periods and to other sensors. Validation against ground data is still in a prelim-
inary stage, and shall be the subject of a later study. In the case of SEVIRI/MSG, LST
values were compared with observations taken in Evora (Southern Portugal) and Gobabeb
(Namibia) (Trigo et al. 2008a; Freitas et al. 2010), suggesting that the retrievals are gen-
erally below 2C and within the NRT estimations of error bars. Full validation of the
quasi-global LST product, however, requires that such exercises be extended to LST values
estimated from GOES and MTSAT.
The LST product described in this work has been freely available since August 2009,
and continues to be generated within the framework of project Geoland-2 (Lacaze et al.
2010). To completely fulfil the land areas of the globe, one more GEO satellite (such as
Meteosat-7) is needed to cover central Asia. In addition, a polar-orbiting satellite may be
used to cover the polar regions. Therefore, the subject of future work shall be the adapta-
tion of algorithms for such satellites and the fusion of those retrievals into a global, high
resolution LST product.
Acknowledgements
This work has been developed within the framework of Geoland-2, the pilot project for the Global
Monitoring for the Environment and Security (GMES) Land Monitoring Core Service; Geoland-
2 is funded by the European Community’s Seventh Framework Programme (FP7/2007–2013) under
grant agreement 218795. Instituto Dom Luiz, Pest-OE/CTE/LA0019/2011-IDL, also supported part
of this work. LST products derived from SEVIRI/MSG are provided by the EUMETSAT Satellite
Application Facility on Land Surface Analysis (LSA SAF).
References
Barroso, C., I. F. Trigo, F. Olesen, C. DaCamara, and M. P. Queluz. 2005. “Intercalibration of NOAA
and Meteosat Window Channel Brightness Temperatures.” International Journal of Remote
Sensing 26: 3717–33.
Belward, A. S., ed. 1996. “The IGBP-DIS Global 1 km Land Cover Data Set (DISCover) – Proposal
and Implementation Plans.” IGBP-DIS Working Paper No. 13. IGBP-DIS Office, Météo-France,
Tou l ou se, F ra n ce , 61 pp .
Berk, A., G. Anderson, P. Acharya, J. Ghetwynd, L. Bernstein, E. Shettle, M. Matthew, and S. Adler-
Golden. 1999. MODTRAN4 User’s Manual. Air Force Research Laboratory. Space Vehicles
Directorate. Air Force Material Command. HANSCOM AFB, MA 01731–3010.
Borbas, E., S. W. Seemann, H. -L. Huang, J. Li, and W. P. Menzel. 2005. “Global Profile Training
Database for Satellite Regression Retrievals with Estimates of Skin Temperature and Emissivity.”
Proceedings of the International ATOVS Study Conference-XIV, Beijing, China, May 25–31,
763–70.
Bosilovich, M. G., J. D. Radakovich, A. da Silva, R. Todling, and F. Verter. 2007. “Skin Temperature
Analysis and Bias Correction in a Coupled Land-Atmosphere Data Assimilation System.”
Journal of the Meteorological Society of Japan 85A: 205–28.
Caparrini, F., F. Castelli, and D. Entekhabi. 2004. “Variational Estimation of Soil and Vegetation
Turb ule nt Tr an s fer a nd He at Fl ux Pa ram ete r s fro m Seq uen ces o f M ul t ise nso r Ima ger y .Wa t e r
Resources Research 40. doi:10.1029/2004WR003358.
Chedin, A., N. A. Scott, C. Wahiche, and P. Moulinier. 1985. “The Improved Initialization Inversion
Method: A High Resolution Physical Method for Temperature Retrievals from Satellites of the
TIROS-N Series.” Journal of Climate and Applied Meteorology 24: 128–43.
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
International Journal of Remote Sensing 17
Chevallier, F. 2002. Sampled Databases of 60-level Atmospheric Profiles from the ECMWF Analyses.
Numerical Weather Prediction Satellite Application Facility Research Report [NWP SAF
Research Report] no. 4, January 2002.
Coll, C., and V. Caselles. 1997. “A Split-Window Algorithm for Land Surface Temperature from
Advanced Very High Resolution Radiometer Data: Validation and Algorithm Comparison.”
Journal of Geophysical Research 102: 16697–713.
Dash, P., F. M. Göttsche, F. S. Olesen, and H. Fischer. 2002. “Land Surface Temperature
and Emissivity Estimation from Passive Sensor Data: Theory and Practice–Current Trends.”
International Journal of Remote Sensing 23: 2563–94.
Derrien, M., and H. Gléau Le. 2005. “MSG/SEVIRI Cloud Mask and Type from SAFNWC.”
International Journal of Remote Sensing 26: 4707–32. doi:10.1080/01431160500166128.
Dickinson, R. E., A. Henderson-Sellers, C. Rosenzweig, and P. J. Sellers. 1991. “Evapotranspiration
Models with Canopy Resistance for Use in Climate Models, a Review.” Agricultural and Forest
Meteorology 54: 373–88.
Faysash, A., and E. A. Smith. 1999. “Simultaneous Land Surface Temperature-Emissivity Retrieval
in the Infrared Split Window.” Journal of Atmospheric and Oceanic Technology 16: 1673–89.
Freitas, S. C., I. F. Trigo, J. M. Bioucas-Dias, and F. Göttsche. 2010. “Quantifying the Uncertainty
of Land Surface Temperature Retrievals from SEVIRI/Meteosat.” IEEE Transactions on
Geoscience and Remote Sensing 48. doi:10.1109/TGRS.2009.2027697.
Ghent, D., J. Kaduk, J. Remedios, J. Ardö, and H. Balzter. 2010. “Assimilation of Land Surface
Temperature into the Land Surface Model JULES with an Ensemble Kalman Filter.” Journal of
Geophysical Research 115: D19112. doi:10.1029/2010JD014392.
Jin, M. 2004. “Analysis of Land Skin Temperature Using AVHRR Observations.” Bulletin of the
American Meteorological Society 85: 587–600.
Jin, M., R. E. Dickinson, and D. Zhang. 2005. “The Footprint of Urban Areas on Global Climate as
Characterized by MODIS.” Journal of Climate 18: 1551–65.
Kustas, W. P., and J. M. Norman. 1996. “Use of Remote Sensing for Evapotranspiration Monitoring
over L and Surfac es.” Hydrological Sciences Journal 41: 495–515.
Lacaze, R., G. Balsamo, F. Baret, A. Bradley, J.-C. Calvet, F. Camacho, R. D’Andrimont, S. C.
Freitas, H. Makhmara, V. Naeimi, P. Pacholczyk, H. Poilvé, B. Smets, K. Tansey, I. F. Trigo,
W. Wagner, M. Weiss. 2010. “GEOLAND2 – Towards an Operational GMES Land Monitoring
Core Service; First Results of the Biogeophysical Parameter Core Mapping Service.” In ISPRS
TC VII Symposium – 100 Years ISPRS, Vienna, Austria, 5–7 July 2010, IAPRS, Vol. XXXVIII,
Par t 7 B,editedbyW.WagnerandB.Székely,3549.
Miglietta, F., B. Gioli, Y. Brunet, R. W. A. Hutjes, A. Matese, C. Sarrat, and A. Zaldei. 2009.
“Sensible and Latent Heat Flux From Radiometric Surface Temperatures at the Regional Scale:
Methodology and Evaluation.” Biogeosciences 6: 1975–86.
Mitchell, K. E., D. Lohmann, P. R. Houser, E. F. Wood, J. C. Schaake, A. Robock, B. A. Cosgrove,
J. Sheffield, Q. Duan, L. Luo, R. W. Higgins, R. T. Pinker, J. D. Tarpley, D. P. Lettenmaier, C. H.
Marshall, J. K. Entin, M. Pan, W. Shi, V. Koren, J. Meng, B. H. Ramsay, and A. A. Bailey. 2004.
“The Multi-Institution North American Land Data Assimilation System (NLDAS): Utilizing
Multiple GCIP Products and Partners in a Continental Distributed Hydrological Modeling
System.” Journal of Geophysical Research 109. doi:10.1029/2003JD003823.
Peres, L. F., and C. C. DaCamara. 2005. “Emissivity Maps to Retrieve Land-Surface Temperature
From MSG/SEVIRI.” IEEE Transactions on Geoscience and Remote Sensing 43: 1834–44.
Pinheiro, A. C. T., J. L. Privette, and P. Guillevic. 2006. “Modeling the Observed Angular Anisotropy
of Land Surface Temperature in a Savanna.” IEEE Transactions on Geoscience and Remote
Sensing 44. doi:10.1109/TGRS.2005.863827.
Pinker, R. T., D. Sun, M. Miller, and G. J. Robinson. 2007. “Diurnal Cycle of Land Surface
Temp e r atu r e i n a Dese r t E n cro a c hme n t Z one as O b s erve d f rom Sa t elli t e s.” Geophysical Research
Letters 34. doi:10.1029/2007GL03186.
Prata, A. J. 1993. “Land Surface Temperature Derived from the Advanced Very High Resolution
Radiometer and the Along-Track Scanning Radiometer: 1. Theory.” Journal of Geophysical
Research 98: 16689–702.
Qin, J., S. Liang, R. Liu, H. Zhang, and B. Hu. 2007. “A Weak-Constraint-Based Data Assimilation
Scheme for Estimating Surface Turbulent Fluxes.” IEEE Transactions on Geoscience and Remote
Sensing Letters 4. doi:10.1109/LGRS.2007.904004.
Rasmussen, M. O., A. C. Pinheiro, S. R. Proud, and I. Sandholt. 2010. “Modeling Angular
Dependences in Land Surface Temperatures from the SEVIRI Instrument Onboard the
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
18 S.C. Freitas et al.
Geostationary Meteosat Second Generation Satellites.” IEEE Transactions on Geoscience and
Remote Sensing 48. doi:10.1109/TGRS.2010.2044509.
Saux-Picart, S., C. Ottlé, B. Decharme, C. André, M. Zribi, A. Perrier, B. Coudert, N. Boulain,
B. Cappelaere, L. Descroix, D. Ramier. 2009. “Water and Energy Budgets Simulation over
the AMMA-Niger Super-Site Spatially Constrained with Remote Sensing Data.” Journal of
Hydrology 375: 287–95.
Seemann, S. W., J. Li, W. P. Menzel, and L. E. Gumley. 2003. “Operational Retrieval of Atmospheric
Temp e r atu r e , Mois t u re, a n d O zone f r om MOD I S I nfr a r ed Rad i ance s . Journal of Applied
Meteorology 42: 1072–91.
Settle, J. J., N. A. Bharmal, G. J. Robinson, and A. Slingo. 2009. “Sampling Uncertainties in Surface
Radiation Budget Calculations in RADAGAST.” Journal of Geophysical Research 113: D00E02.
doi:10.1029/2008JD010509.
Sun, D., and R. T. Pinker. 2007. “Retrieval of Surface Temperature from the MSG-SEVIRI
Observations: Part I. Methodology.” International Journal of Remote Sensing 28: 5255–72.
Sun, D., R. T. Pinker, and J. B. Basara. 2004. “Land Surface Temperature Estimation from the Next
Generation of Geostationary Operational Environmental Satellites: GOES M–Q.” Journal of
Applied Meteorology 43: 363–72.
Trigo, I. F., C. C. DaCamara, P. Viterbo, J.-L. Roujean, F. Olesen, C. Barroso, F. Camacho-de
Coca, D. Carrer, S. C. Freitas, J. García-Haro, B. Geiger, F. Gellens-Meulenberghs, N. Ghilain,
J. Meliá, L. Pessanha, N. Siljamo, and A. Arboleda. 2011. “The Satellite Application
Fac il ity on Land Surface Analy si s. International Journal of Remote Sensing 32: 2725–44.
doi:10.1080/01431161003743199.
Trigo, I. F., I. T. Monteiro, F. Olesen, and E. Kabsch. 2008a. “An Assessment of Remotely
Sensed Land Surface Temperature.” Journal of Geophysical Research 113: D17108.
doi:10.1029/2008JD010035.
Trigo, I. F., L. F. Peres, C. C. DaCamara, and S. C. Freitas. 2008b. “Thermal Land Surface Emissivity
Retrieved from SEVIRI/Meteosat.” IEEE Transactions on Geoscience and Remote Sensing 46.
doi: 10.1109/TGRS.2007.905197.
Trigo, I. F. and P. Viterbo. 2003. “Clear-Sky Window Channel Radiances: A Comparison Between
Observations and the ECMWF Model.” Journal of Applied Meteorology 42: 1463–79.
Wan, Z. E., and J. Dozier. 1996. “A Generalized Split-Window Algorithm for Retrieving Land
Surface Temperature from Space.” IEEE Transactions on Geoscience and Remote Sensing 34:
892–905.
Wan, Z., P. Wang, and X. Li. 2004. “Using MODIS Land Surface Temperature and Normalized
Difference Vegetation Index Products for Monitoring Drought in the Southern Great Plains,
USA.” International Journal of Remote Sensing 25: 61–72.
Yu, Y., J. L. Privette, and A. C. Pinheiro. 2008. “Evaluation of Split-Window Land Surface
Temp e r atu r e A lgor i t hms f o r G ene r a ting C l i mat e D a ta Re c o rds. IEEE Transactions on
Geoscience and Remote Sensing 46. doi:10.1109/TGRS.2007.909097.
Downloaded by [b-on: Biblioteca do conhecimento online UL] at 12:37 26 December 2012
... The SR data are, in turn, used to compute the NDVI, which is then converted to FVC values (Carlson & Ripley 1997) and used together with previously computed ASTER emissivity values for the bare ground to obtain the corresponding Landsat emissivity (Carlson & Ripley 1997;Rubio et al. 1997). Finally, the SMW algorithm is applied to the TOA TB of the Landsat TIR band (Li et al. 2013;Duguay-Tetzlaff et al. 2015); the algorithm coefficients are mapped onto the Landsat image based on TWVC from NCEP and taking the classes defined in SWM algorithm (Equation (3)) (Sun et al. 2004;Freitas et al. 2013). ...
Article
Full-text available
With the drastic change in global climate, the wide distribution of natural lakes over the Qinghai-Tibet Plateau (TP) has attracted extensive attention due to their high climate sensitivity. In this paper, we investigated the dynamics of Paiku Co, the largest inland lake in the Qomolangma Natural Reserve, with the associated response to climate change in the past three decades. The methods used contain the water index method, the spatial and temporal fusion model, the statistical mono-window algorithm, and multi-variable linear regression. Lake area fluctuated greatly in 1990–2000, followed by a continuous shrinkage in 2000–2010, and stabled after that 2010–2020. We forecasted that Paiku Co would enter a slow expansion period. Conjoint analysis with climate factors showed that the area variation of Paiku Co was not significantly related to precipitation change, but negatively related to the change of air temperature and lake temperature. We found that the lake change was not dominated by a single factor but showed different climate sensitivity in each period. Especially, there was a common inflection point around 2013 that might herald the occurrence of a new trend of climate change. This article provides new ideas and solutions for the research of lakes in the Qinghai-Tibet Plateau and offers a reference for water resource management.
... In addition, the morning warming rate has been shown to be directly related to soil moisture status (Anderson et al., 2007;Piles et al., 2016;Van de Griend et al., 1985). Hence, geostationary (GEO) satellites such as Geostationary Operational Environmental Satellites (GOES)-R Advanced Baseline Imager (ABI) , Meteosat Second Generation (MSG) Spinning Enhanced Visible and InfraRed Imager (SEVIRI) (Freitas et al., 2009), and the Feng Yun meteorological satellites (Tang et al., 2008), have become an optimal choice for observing sub-hourly LST across a wide spatial coverage (Freitas et al., 2013). ...
Article
Full-text available
By characterizing high-frequency surface thermal dynamics at a medium spatial scale, hourly land surface temperatures (LST), retrieved from geostationary satellite thermal infrared (TIR) observations, shows great potential to be used across a range of scientific applications; however, cloud cover typically leads to data gaps and degraded retrieval accuracy in TIR LST products, such as the official Advanced Baseline Imager (ABI) LST product. Studies have focused on the LST gap reconstruction; however, most interpolation-based methods only work for a short-term cloud duration and are unable to adequately compensate for cloud effects, and traditional surface energy balance (SEB)-based methods are able to handle cloud coverage while they are not feasible for use at night. Moreover, few studies have concentrated on recovering the abnormal retrievals of partial cloud pixels. In this study, an all-sky diurnal, hourly LST estimation method based on SEB theory was proposed; the proposed method involved three major steps: 1) an original spatiotemporal dynamic model of LST was constructed from ECMWF Reanalysis v.5 (ERA5); 2) clear-sky ABI LST was then assimilated to the dynamic model to generate a continuous LST series; 3) the diurnal cloud effects were superimposed on cloudy time estimated by an innovative optimization method from satellite radiation products. A 2-km, all-sky, hourly LST product was produced over the contiguous US and Mexico from July 2017 to June 2021. Validation was conducted using ground measurements at 18 sites from Surface Radiation (SURFRAD) and core AmeriFlux networks, and produced an overall root-mean-square error (RMSE) of 2.44 K, a bias of − 0.19 K, and an R 2 of 0.97 based on 408,300 samples. For clear-sky samples, the RMSE values were 2.37 and 2.24 K for day and nighttime, respectively, which was a notable improvement over the corresponding values of the official ABI LST product (2.73 and 2.86 K, respectively). The RMSE values on cloudy-sky were 2.78 and 2.23 K for day and nighttime, respectively. The daily mean LST by aggregating all-sky, hourly LST had an RMSE of 1.13 K and R 2 of 0.99. Overall, this product showed reliability under consistent cloud durations, although it was slightly affected by surface elevation. The diurnal temperature cycle climatology of major land cover types was also characterized. The product is freely available at: http://glass.umd.edu/allsky_LST/ABI/.
... Besides, they yield the "sub-surface temperature" instead of the "skin temperature" provided by TIR remote sensing. The TIR observations are consequently more suitable for obtaining high accuracy LSTs (Cook et al., 2014;Freitas et al., 2013;Malakar et al., 2018;Spacesystems and Team, 2001a;Wan, 2014;Wan et al., 2015a). ...
Thesis
Land surface temperature (LST) is an important parameter of the Earth surface. However, there are still some factors influencing the urban LST retrieval accuracy but have not been well addressed in existing LST retrieval algorithms: (1) the adjacency effect in the thermal infrared (TIR) spectral region; (2) the impact of the three-dimensional structures and their radiation on the TIR measurements; (3) the dependence of existing LST retrieval algorithms on the information of atmosphere and/or the Earth surface emissivity. In this thesis, two forward radiative transfer models and one new LST retrieval method have been developed to deal with these three factors, based on which, a preliminary exploration on developing an improved urban LST retrieval method from high spatial resolution satellite TIR measurements has been given. Results show that the urban LST retrieval accuracy could be improved by about 2.0 K, indicating the good performance of the proposed method.
... Therefore, satellite-based LST has been utilized more than weather station-measured LST in many studies [9][10][11][12][13]. Among various satellite-based LSTs, the moderate-resolution imaging spectroradiometer (MODIS) satellite-based LST has been widely utilized because of its 1 km spatial resolution and four-times-a-day temporal resolution. ...
Article
Full-text available
A high spatio-temporal resolution land surface temperature (LST) is necessary for various research fields because LST plays a crucial role in the energy exchange between the atmosphere and the ground surface. The moderate-resolution imaging spectroradiometer (MODIS) LST has been widely used, but it is not available under cloudy conditions. This study proposed a novel approach for reconstructing all-sky 1 km MODIS LST in South Korea during the summer seasons using various data sources, considering the cloud effects on LST. In South Korea, a Local Data Assimilation and Prediction System (LDAPS) with a relatively high spatial resolution of 1.5 km has been operated since 2013. The LDAPS model’s analysis data, binary MODIS cloud cover, and auxiliary data were used as input variables, while MODIS LST and cloudy-sky in situ LST were used together as target variables based on the light gradient boosting machine (LightGBM) approach. As a result of spatial five-fold cross-validation using MODIS LST, the proposed model had a coefficient of determination (R2) of 0.89–0.91 with a root mean square error (RMSE) of 1.11–1.39 °C during the daytime, and an R2 of 0.96–0.97 with an RMSE of 0.59–0.60 °C at nighttime. In addition, the reconstructed LST under the cloud was evaluated using leave-one-station-out cross-validation (LOSOCV) using 22 weather stations. From the LOSOCV results under cloudy conditions, the proposed LightGBM model had an R2 of 0.55–0.63 with an RMSE of 2.41–3.00 °C during the daytime, and an R2 of 0.70–0.74 with an RMSE of 1.31–1.36 °C at nighttime. These results indicated that the reconstructed LST has higher accuracy than the LDAPS model. This study also demonstrated that cloud cover information improved the cloudy-sky LST estimation accuracy by adequately reflecting the heterogeneity of the relationship between LST and input variables under clear and cloudy skies. The reconstructed all-sky LST can be used in a variety of research applications including weather monitoring and forecasting.
... Cette haute résolution temporelle permet de bien représenter la dynamique du cycle diurne ainsi que de pouvoir bénéficier au mieux des fenêtres sans nuages. Cependant, les observations des satellites géostationnaires sont limitées par une couverture réduite en latitude (±60 • N ), et la nécessité de combiner plusieurs instruments géostationnaires pour observer toutes les longitudes (Freitas et al., 2013). ...
Thesis
La température de surface des continents est une variable essentielle dans la modélisation du bilan énergétique global. Les mesures effectuées par les imageurs microondes passifs en orbite permettent d’estimer la température de la surface quelle que soit la couverture nuageuse contrairement aux mesures dans l'infrarouge qui ne sont pas utilisables sous les nuages. Néanmoins, plusieurs difficultés surviennent lors de l'utilisation de ces observations. Parmi celles-ci, la correction des écarts entre des observations provenant d’instruments différents par un inter-étalonnage des mesures, la détection des observations contaminées par des nuages en fonction de la fréquence d'observation et du type de nuage sont abordées dans cette thèse. Ou encore la création de cartes recensant les zones arides dans lesquelles la température en microondes correspond à celle de la sub-surface qui peut être différente de la température à la surface. Une méthode pour augmenter la résolution spatiale des températures de surface en se basant sur des relations statistiques avec d’autres variables est aussi présentée.
... A number of geostationary satellites (GOES, MSG, MTSAT) carry instruments that capture thermal radiation or aerosol properties at very high temporal resolutions (e.g. every 15 minutes), although they have lower spatial resolutions (2 km to 5 km at nadir), a constant field of view, and possibly greater bias and inaccuracy (Freitas et al., 2013). Two recent studies used LST from the SEVIRI instrument on the MSG satellites to estimate hourly Ta over Germany via multiple linear regression (Bechtel et al., 2017) or via inverse distance weighted interpolation combined with a regional climate model (Krähenmann et al., 2018). ...
Thesis
Studies suggest that exposure to heat, cold, or particulate matter (PM) during pregnancy may decrease birth weight and shorten the duration of gestation, but we do not fully understand these associations. This is concerning because low birth weight and preterm birth increase risk for numerous morbidities and chronic diseases throughout childhood and into adulthood. The health burdens of temperature and PM are expected to increase with climate change and global urbanization. Improving our understanding requires accurate exposure assessment because error can bias associations with health outcomes. A scalable way to assess exposure over large areas and long time periods is to model the relationship between monitored temperature or PM and satellite-derived variables such as land surface temperature (LST) or aerosol optical depth (AOD). Improved calibration methods might increase accuracy, and finer spatial resolutions could be useful in urban areas.This PhD developed novel approaches to increase the accuracy and spatial resolution of satellite-based exposure models, applied them to reconstruct daily temperature and PM in France over the past two decades, and used the temperatures to identify periods during pregnancy when exposure to cold or heat may increase the risk of preterm birth.First, we found that allowing the relationship between monitored ambient temperature and satellite-derived LST to vary over space as well as time improves the accuracy of mixed models. We used this approach to model diurnal air temperature (mean [Tmean], minimum [Tmin], and maximum [Tmax]) over France at 1 km resolution from 2000 to 2018. We also showed that an ensemble of random forests and gradient boosting machines can downscale daily temperature over urban areas by incorporating thermal satellite data at a high spatial but low temporal resolution. We used this approach to model daily Tmean, Tmin, and Tmax at 200 m resolution over 103 cities.Second, we found that Gaussian Markov random fields can reconstruct PM based on gap-filled AOD more accurately than widely used mixed models or random forests, and that an ensemble of all three algorithms performs even better. We used this approach to model daily mean PM2.5 and PM10 (PM <2.5 and <10 microns in diameter, respectively) concentration over France at 1 km resolution from 2000 to 2019 and showed that alleviating the sparsity of PM2.5 monitors by imputing PM2.5 at more common PM10 monitors increased accuracy.Third, we conducted a survival analysis using Cox proportional hazards models with distributed lags to assess the time-varying association between residence-based ambient temperature estimated by our exposure model and the risk of preterm birth among 5347 singleton live births (4.3% preterm) from prospective pregnancy cohorts in France. We simultaneously examined chronic exposure (weekly from conception) and acute exposure (daily during the 30 days before delivery), and we explored the effects of temperature variability and acclimation to location and season during heatwaves. Cold and night-time heat increased the risk of preterm birth, with susceptibility starting as early as conception and continuing through portions of trimesters 1 and 2. Cold around 7 days before delivery also increased the risk of preterm birth.Our approaches could improve satellite-based exposure models in other areas and our multi-decadal dataset will be of interest to epidemiologists, climatologists, planners, policymakers, and the public. In the context of rising temperatures and more frequent weather hazards, our findings on the risks that heat and cold pose to pregnant women and their infants should inform public health policies to reduce the growing burden of preterm birth.
... Its estimation further depends on the albedo, the vegetation cover and the soil moisture. Further details about the temperature retrieval can be found in Freitas et al. (2013). The LST V2 product user manual and validation reports can be found at https://land.copernicus.eu/global/products/lst? qt-lst_characteristics=5#qt-lst_characteristics (last access: 7 February 2022). ...
Article
Full-text available
This paper is the companion paper of Córdoba-Jabonero et al. (2021). It deals with the estimation of the long-wave (LW) and net dust direct radiative effect (DRE) during the dust episode that occurred between 23 and 30 June 2019 and was paired with a mega-heatwave. The analysis is performed at two European sites where polarized micro-pulse lidars ran continuously to retrieve the vertical distribution of the dust optical properties: Barcelona, Spain, 23–30 June, and Leipzig, Germany, 29–30 June. The radiative effect is computed with the GAME (global atmospheric model) radiative transfer model separately for the fine- and coarse-mode dust. The instantaneous and daily radiative effect and radiative efficiency (DREff) are provided for the fine-mode, coarse-mode and total dust at the surface, top of the atmosphere (TOA) and in the atmosphere. The fine-mode daily LW DRE is low (
Article
The urban heat island effect (UHI) has been widely observed globally, causing climate, health, and energy impacts in cities. The UHI intensities have been found to largely depend on background climate and the properties of the urban fabric. Yet, a complete mechanistic understanding of how UHIs develop at a global scale is still missing. Using an urban ecohydrological and land-surface model (urban Tethys-Chloris) in combination with multi-source remote sensing data, we performed simulations for 49 large urban clusters across the Northern Hemisphere in 2009–2019 and analysed how surface and canopy air UHIs (SUHI and CUHI, respectively) develop during day and night. Biophysical drivers triggering the development of SUHIs and CUHIs have similar dependencies on background climate, but with different magnitudes. In humid regions daytime UHIs can be largely explained by the urban-rural difference in evapotranspiration, whereas heat convection and conduction are important in arid areas. Plant irrigation can largely promote daytime urban evapotranspiration only in arid and semi-arid climates. During night, heat conduction from the urban fabric to the environment creates large UHIs mostly in warm arid regions. Overall, this study presents a mechanistic quantification of how UHIs develop worldwide and proposes viable solutions for sustainable climate-sensitive mitigation strategies.
Article
Full-text available
Land surface temperature is linked to a wide range of surface processes. Given the increased development of earth observation systems, a large effort has been put into advancing land surface temperature retrieval algorithms from remote sensors. Due to the very limited number of reliable in situ observations matching the spatial scales of satellite observations, algorithm development relies on synthetic databases, which then constitute a crucial part of algorithm development. Here we provide a database of atmospheric profiles and respective surface conditions that can be used to train and verify algorithms for land surface temperature retrieval, including machine learning techniques. The database was built from ERA5 data resampled through a dissimilarity criterion applied to the temperature and specific humidity profiles. This criterion aims to obtain regular distributions of these variables, ensuring a good representation of all atmospheric conditions. The corresponding vertical profiles of ozone and relevant surface and vertically integrated variables are also included in the dataset. Information on the surface conditions (i.e., temperature and emissivity) was complemented with data from a wide array of satellite products, enabling a more realistic surface representation. The dataset is freely available online at Zenodo.
Preprint
Daily mean land surface temperatures (LSTs) acquired from polar-orbiters are crucial for various applications such as global and regional climate change analysis. However, thermal sensors from polar-orbiters can only sample the surface effectively with very limited times per day under cloud-free conditions. These limitations have produced a systematic sampling bias (ΔTsb) on the daily mean LST (Tdm) estimated with the traditional method, which uses the averages of clear-sky LST observations directly as the Tdm. Several methods have been proposed for the estimation of the Tdm, yet they become less capable of generating spatiotemporally seamless Tdm across the globe. Based on MODIS and reanalysis data, here we proposed an improved annual and diurnal temperature cycle-based framework (termed the IADTC framework) to generate global spatiotemporally seamless Tdm products ranging from 2003 to 2019 (named as the GADTC products). The validations show that the IADTC framework reduces the systematic ΔTsb significantly. When validated only with in situ data, the assessments show that the mean absolute errors (MAEs) of the IADTC framework are 1.4 K and 1.1 K for SURFRAD and FLUXNET data, respectively; and the mean biases are both close to zero. Direct comparisons between the GADTC products and in situ measurements indicate that the MAEs are 2.2 K and 3.1 K for the SURFRAD and FLUXNET datasets, respectively; and the mean biases are −1.6 K and −1.5 K for these two datasets, respectively. By taking the GADTC products as references, further analysis reveals that the Tdm estimated with the traditional averaging method yields a positive systematic ΔTsb of greater than 2.0 K in low- and mid-latitude regions while of a relatively small value in high-latitude regions. Although the global mean LST trend (2003 to 2019) calculated with the traditional method and the IADTC framework is relatively close (both between 0.025 to 0.029 K/year), regional discrepancies in LST trend does occur – the pixel-based MAE in LST trend between these two methods reaches 0.012 K/year. We consider the IADTC framework can guide the further optimization of Tdm estimation across the globe; and the generated GADTC products should be valuable in various applications such as global and regional warming analysis. The GADTC products are freely available at https://doi.org/10.5281/zenodo.6287052 (Hong et al., 2022).
Article
Full-text available
One mechanism for climate change is the collected impact of changes in land cover or land use. Such changes are especially significant in urban areas where much of the world's population lives. Satellite observations provide a basis for characterizing the physical modifications that result from urbanization. In particular, the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument on the National Aeronautics and Space Administration (NASA) Terra satellite measures surface spectral albedos, thermal emissivities, and radiative temperatures. A better understanding of these measurements should improve our knowledge of the climate impact of urbanization as well as our ability to specify the parameters needed by climate models to compute the impacts of urbanization. For this purpose, it is useful to contrast urban areas with neighboring nonurban surfaces with regard to their radiative surface temperatures, emissivities, and albedos. Among these properties, surface temperatures have been most extensively studied previously in the context of the ``urban heat island'' (UHI). Nevertheless, except for a few detailed studies, the UHI has mostly been characterized in terms of surface air temperatures.To provide a global analysis, the zonal average of these properties are presented here measured over urban areas versus neighboring nonurban areas. Furthermore, individual cities are examined to illustrate the variations of these variables with land cover under different climate conditions [e.g., in Beijing, New York, and Phoenix (a desert city of the United States)]. Satellite-measured skin temperatures are related to the surface air temperatures but do not necessarily have the same seasonal and diurnal variations, since they are more coupled to surface energy exchange processes and less to the overlying atmospheric column. Consequently, the UHI effects from skin temperature are shown to be pronounced at both daytime and nighttime, rather than at night as previously suggested from surface air temperature measurements. In addition, urban areas are characterized by albedos much lower than those of croplands and deciduous forests in summer but similar to those of forests in winter. Thus, urban surfaces can be distinguished from nonurban surfaces through use of a proposed index formed by multiplying skin temperature by albedo.
Article
Full-text available
A comparison of clear-sky radiances of the Meteosat window channel with the European Centre for Medium-Range Weather Forecasts (ECMWF) model results is presented, aiming to assess both the model's performance and the quality of the observations. The comparison is made for four periods covering the seasonal cycle and for the Meteosat-7 (Africa and southern Europe) and Meteosat-5 (Middle East and the Indian subcontinent) disks. Results show 1) an underestimation of the diurnal cycle of model brightness temperature in clear-sky conditions with a cool bias during daytime of up to 7-8 K and 2) a much smaller and less widespread warm bias at night. Although such discrepancies are sensitive to surface characteristics (e.g., vegetation type, terrain elevation, and slope), problems in the model surface-to-boundary layer coupling are identified as the most likely model deficiency associated with the underestimation of diurnal amplitudes of skin temperature. A conditional error analysis also reveals strong error stratification with the model cloud cover and with the percent of clear-sky pixels associated with the Meteosat data, suggesting great care should be taken when using Meteosat clear-sky observations, particularly in the ITCZ and adjacent areas.
Conference Paper
The algorithm for retrieving atmospheric temperature, moisture, and total column ozone using the Moderate Resolution Imaging Spectroradiometer (MODIS) longwave infrared radiances is presented. The operational MODIS algorithm performs clear sky retrievals globally over land and ocean for both day and night. The algorithm is based on a regression and has an option to follow the statistical retrieval with a nonlinear physical retrieval. The regression coefficients are determined from an extension of the NOAA-88 data set containing more than 8400 global radiosonde measurements of atmospheric temperature, moisture and ozone profiles. Evaluation of atmospheric products is performed by a comparison with data from ground-based instrumentation, geostationary infrared sounders, and polar orbiting microwave sounders. MODIS moisture products are in general agreement with the gradients and distributions from the other satellites, while MODIS depicts more detailed structure with its improved spatial resolution.
Article
In an initial investigation, remotely sensed surface temperature is assimilated into a coupled atmosphere/land global data assimilation system, with explicit accounting for biases in the model state. In this scheme, a incremental bias correction term is introduced in the model's surface energy budget. In its simplest form, the algorithm estimates and corrects a constant time mean bias for each gridpoint; additional benefits are attained with a refined version of the algorithm which allows for a correction of the mean diurnal cycle. The method is validated against the assimilated observations, as well as independent near-surface air temperature observations. In many regions, not accounting for the diurnal cycle of bias caused degradation of the diurnal amplitude of background model air temperature. Energy fluxes collected through the Coordinated Enhanced Observing Period (CEOP) are used to more closely inspect the surface energy budget. In general, sensible heat flux is improved with the surface temperature assimilation, and two stations show a reduction of bias by as much as 30 W m(-2). At the Rondonia station in Amazonia, the Bowen ratio changes direction in an improvement related to the temperature assimilation. However, at many stations the monthly latent heat flux bias is slightly increased. These results show the impact of univariate assimilation of surface temperature observations on the surface energy budget, and suggest the need for multivariate land data assimilation. The results also show the need for independent validation data, especially flux stations in varied climate regimes.
Article
A combined land surface temperature-emissivity retrieval algorithm is developed and tested for Geostationary Operational Environmental Satellite (GOES)-Imager and National Oceanic and Atmospheric Administration Advanced Very High Resolution Radiometer (AVHRR) split-window channels. By assuming that the spectral emissivities are constant over a short time period (12-24 h), two sets of split-window radiance measurements taken at two different times are used to retrieve two spectral emissivities and two land surface temperatures (LSTs) simultaneously. The algorithm employs an optimization scheme rather than a direct solver for a system of equations because of constraint requirements. The retrieved variables minimize the rms differences between measure satellite radiances and those predicted by a spectrally detailed radiative transfer model. A GOES-8 version of the algorithm is validated with in situ radiometer measurements from the Department of Energy's Atmospheric Radiation Measurement Program Cloud and Radiation Testbed (ARM CART) site. In addition, an AVHRR version is validated with in situ measurements from the First ISLSCP Field Experiment (FIFE) site, the Hydrological Atmospheric Pilot Experiment-Sahel (HAPEX-Sahel) site, and an LST validation site operated near Melbourne, Australia. The biases of the retrieved LSTs for the validation sites in the Australian. FIFE, and ARM CART study areas are approximately 0.08°, 1.7°, and 1.4°C, respectively, yielding an overall bias error of better than half the current expected accuracy limit of some ±3°C. The associated bias-adjusted rmse differences are approximately 0.78°, 4.8°, and 4.5°C, respectively, mostly driven by intercomparing in situ point measurements to area-integrated satellite pixel retrievals. The bias-adjusted rmse differences for HAPEX-Sahel are larger (5°and 11°C), resulting from incomplete characterization of site heterogeneity, insufficient radiosonde launch frequency, and poor data quality of the temperature-moisture soundings, rather than intrinsic algorithm problems. Notably, the averaged retrieved emissivities for the trouble-free sites are within the expected range of emissivities for vegetated surfaces. The GOES-8 retrieved LSTs exhibit small amplitude, high-frequency noise, and a daily error cycle when compared to in situ measurements. The noise is attributed to random detector errors in the satellite observations for which the channel 4 noise-equivalent temperature difference is larger than that of channel 5. The systematic differences between validation measurements and retrievals are near zero during nighttime but exhibit a small semidiurnal oscillation during daytime. Notwithstanding a possible semidiurnal bias in the pyrgeometer validation measurements associated with imperfect solar dome heating corrections, plus unaccounted-for attenuation between the surface and pyrgeometer, the latter error cycle is attributed to a too-coarse sampling of the nonlinear diurnal evolution of the thermodynamic structure of the atmospheric boundary layer, particularly near the sunrise and sunset transition times. Thus, sounding frequency determines the error characteristics of the nonlinearly evolving split-window weighting functions.
Article
Data from the Spinning Enhanced Visible and Infrared Imager (SEVIRI) on board the first Meteosat Second Generation (MSG) satellite have been available since February 2003. Four MSG satellites are planned to ensure an operational service until at least 2018. A software package, which derives from MSG/SEVIRI imagery a set of 12 products useful for nowcasting purposes, has been developed cooperatively by the Satellite Application Facility for supporting NoWCasting and very short range forecasting (SAFNWC) and is distributed by EUMETSAT.This paper describes the cloud mask (CMa) and type (CT) algorithms implemented in this SAFNWC/MSG software package. A multispectral thresholding technique has been used: the test sequence depends on illumination conditions and geographical location whereas most thresholds are dynamically computed from ancillary data (atlas, climatology maps, numerical weather prediction (NWP) model forecast fields) using radiative transfer models. These algorithms have been prototyped using GOES‐8 and MODIS imagery before being applied to MSG‐1/SEVIRI. The cloud mask and type can be extracted in any area inside the MSG full disk. Preliminary validation results obtained from a comparison with surface observations using a few months of MSG‐1/SEVIRI data show good performances.
Article
This study presents an intercalibration of Meteosat‐5 11 µm channel and NOAA‐14 10.8 µm and 12.0 µm channels, and their comparison for sea and land pixels. The intercalibration empirical relation is derived for clear‐sky sea measurements, with similar zenith viewing angles. The root mean square difference between NOAA‐14 and Meteosat‐5 intercalibrated brightness temperatures is about 1.4 K (4.7 K) for all clear‐sky sea (land) pixels. The discrepancies over land are analysed in terms of viewing angle, surface type, terrain elevation and exposure to sunlight. The satellite viewing geometry is responsible for two major impacts, namely: the obstruction by neighbouring clouds towards one of the satellites; and differences in surface solar illumination viewed by each sensor. It is also shown that the higher discrepancies between intercalibrated temperatures occur for the most heterogeneous surfaces (e.g. Open Shrublands). The effect of terrain variability is not linear and depends strongly on the surface type.
Article
The satellite application facility on land surface analysis (Land SAF) generates, archives, and disseminates land surface temperature (LST) in an operational basis. LST is estimated from the spinning enhanced visible and infrared imager (SEVIRI) onboard Meteosat, making use of a generalized split-windows algorithm. Here SEVIRI LST is compared with retrievals from the moderate resolution imaging spectroradiometer (MODIS), collocated in space and time, for three 10° × 10° areas (Iberian Peninsula, Central Africa, and the Kalahari), and for six 7-day periods between July 2005 and May 2006. Overall, SEVIRI LSTs are warmer than MODIS values, with maximum discrepancies generally observed for daytime. The mismatches between the two satellite retrievals are then analyzed in terms of (1) satellite viewing angle differences, (2) surface orography, and (3) surface type. Daytime discrepancies are strongly impacted by differential heating rates of elements within a pixel (e.g., vegetation types, bareground), leading to a relatively wide range of MODIS-SEVIRI LST differences, with strong dependency on the MODIS view zenith angle. In contrast, average nighttime discrepancies are generally below 2°C. The intercomparison between MODIS and SEVIRI LST is complemented with in situ observations taken at Evora ground station (southwestern part of the Iberian Peninsula). The differences between ground and satellite-derived values show high variability for daytime for both sensors, with a systematic overestimation of in situ values by SEVIRI LST. In the case of nighttime observations, both sensors tend to underestimate local measurements, with estimated bias over all events under study of -1.7°C and -2.6°C for SEVIRI and MODIS LST, respectively.
Article
A split-window algorithm for deriving land surface temperatures (LSTs) from advanced very high resolution radiometer (AVHRR) channels 4 and 5 is proposed and validated with in situ measured temperatures. On the basis of the radiative transfer theory the algorithm defines a set of surface-independent coefficients which are equivalent to the classical split-window coefficients for sea surface temperature (SST). These coefficients are calculated using SST matchups (coincident AVHRR and buoy measurements) provided by the National Oceanic and Atmospheric Administration (NOAA)-NASA Pathfinder Database of worldwide measurements. Thus calibration of the split-window coefficients is done using real data. The variability of atmospheric attenuation is represented in the proposed algorithm by a quadratic dependence on the brightness temperature difference. For LST determination the emissivity effect is modeled through an additive coefficient which depends on surface emissivity in the AVHRR channels 4 and 5. The algorithm is validated for both SST and LST by using independent ground-based and AVHRR data. The database used in the validation of LST was obtained for a wide range of surface types in a semiarid environment. The same databases are used to compare the accuracies of other published split-window algorithms. The proposed algorithm yields standard errors of temperature estimate between +/-1.0 and +/-1.5K, and no significant biases are observed. Although results are encouraging, more validation is required principally for moist atmospheric conditions.
Article
The most advanced climate models are now incorporating effects of vegetation into their surface energy and hydrological balance formulations. These efforts are reviewed here with emphasis on the canopy resistance component through development of a simple generic canopy model that includes the relatively well-established processes common to several canopy submodels. We also review where the land surface treatments diverge, in particular in their consideration of water stress, partial vegetation, within-canopy resistances and computational algorithms for the determination of canopy temperature. The inclusion of canopy submodels in climate models may be of questionable utility without improvements in some aspects of atmospheric simulations, such as the modeled distributions of precipitation and incident surface radiation. Improved treatments of the heterogeneous nature of precipitation and land surface properties are also needed.