Using NOAA AVHRR and SPOT VGT data to estimate surface parameters: application to a mesoscale meteorological model

Abstract

The mesoscale numerical weather prediction model MM5, the 5th generation Pennsylvania State University/NCAR Mesoscale Model, uses a global land-use map to set the physical parameters on the surface characteristics to model the soil-atmosphere processes. These parameters are albedo, emissivity, thermal inertia, roughness length and soil moisture. A new estimation of soil parameters is done for the north-east of the Iberian Peninsula from an AVHRR data set of year 2000. The new values are introduced into MM5 via a new land- use map, the recent NATLAN 2000-CORINE land-use map, in order to incorporate the last decade land-cover changes. The model is tested with the original and the CORINE land-use map to evaluate the sensitivity to land-use changes and new physical soil parameters definition. Results show clear local differences in some meteorological variables as wind fields or updraft move- ments, but comparisons with ground measurements do not lead to a clear improvement in the model general performance.

Full text

Using NOAA AVHRR and SPOT VGT data to estimate surface
parameters: application to a mesoscale meteorological model
N. PINEDA*{, O. JORBA{, J. JORGE{ and J. M. BALDASANO{
{Department of Applied Physics, Escola Universita`ria Polite`cnica de Manresa
(EUPM), Universitat Polite`cnica de Catalunya (UPC), Av. Bases Manresa
61-73, 08240 Manresa, Spain
{Laboratory of Environmental Modeling, Escola Te`cnica Superior
d’Enginyeria Industrial de Barcelona (ETSEIB), Universitat Polite`cnica de
Catalunya (UPC), Av. Diagonal 647, 08028 Barcelona, Spain
Abstract. The mesoscale numerical weather prediction model MM5, the 5
th
generation Pennsylvania State University/NCAR Mesoscale Model, uses a
global land-use map to set the physical parameters on the surface characteristics
to model the soil-atmosphere processes. These parameters are albedo, emissivity,
thermal inertia, roughness length and soil moisture. A new estimation of soil
parameters is done for the north-east of the Iberian Peninsula from an AVHRR
data set of year 2000. The new values are introduced into MM5 via a new land-
use map, the recent NATLAN 2000-CORINE land-use map, in order to
incorporate the last decade land-cover changes. The model is tested with the
original and the CORINE land-use map to evaluate the sensitivity to land-use
changes and new physical soil parameters definition. Results show clear local
differences in some meteorological variables as wind fields or updraft move-
ments, but comparisons with ground measurements do not lead to a clear
improvement in the model general performance.
1. Introduction
Correctly treating the land surface properties is becoming increasingly
important for meteorological models to be able to capture local mesoscale
circulation induced by land surface forcing (Chen and Dudhia 2001). Several
mesoscale models, like the fifth-generation Pennsylvania State University/NCAR
Mesoscale Model (MM5) used in this study, rely on albedo, emissivity, thermal
inertia, roughness length and soil moisture data sets derived from land-use maps.
The simple surface parameters scheme of the model specifies them according to land
use category and season. The accuracy of land-use information is important to
obtain accurate simulations.
MM5 (Dudhia 1993) uses the USGS global land-use map (Anderson et al. 1976)
to set the physical parameters on the surface characteristics, a 10-year old land-use
map that was created with multitemporal 1-km AVHRR NDVI data (1992–1993).
To incorporate the last decade land-cover changes, the recent CORINE land-use
International Journal of Remote Sensing
ISSN 0143-1161 print/ISSN 1366-5901 online # 2004 Taylor & Francis Ltd
http://www.tandf.co.uk/journals
*Corresponding author; e-mail: npineda@fa.upc.es
INT. J. REMOTE SENSING,10JANUARY, 2004,
VOL. 25, NO. 1, 129–143

map of the European Environment Agency NATLAN information package (EEA
2000) was used in the simulations instead of the original map.
Surface parameters (albedo, emissivity and thermal inertia) were estimated for
the working region from an AVHRR data set of year 2000, and mean values for the
CORINE land-use classes were calculated for the four seasons. As soil moisture
and roughness can not be estimated with optical remote sensing sensors, a category
equivalence between land-use maps was established, in order to use the original
MM5 tabulated roughness and moisture values with the new land-use map.
2. Study location
The location of the present study is situated in the north-east of the Iberian
Peninsula (figure 1). This region is characterized with very complex topography, a
high mountain range to the north of the area, with extended mountains to the
south-west and an important river-valley canalisation going from the north-west to
the east flowing into the Mediterranean sea. The sea is another particularity of the
region, with a large coastline.
3. Surface parameter estimation
3.1. AVHRR processing
A selection of 70 afternoon pass and 55 morning pass NOAA-14 AVHRR L1B
images were used in this study. Such images were provided by the Centro de
Recepcio´n, Proceso, Archivo y Distribucio´n de Ima´genes de Observacio´n de la
Tierra (CREPAD). Before surface parameter estimation some corrections were
applied:
Step 1. Calibration. Because of sensor degradation after launch, it was necessary
to apply time-dependent calibration gains and offsets, obtained from NOAA
(1998), to derive the top of atmosphere (TOA) radiance for AVHRR 1 and 2
channels.
Step 2. TOA reflectance for AVHRR channels 1 and 2. A lambertian correction
(Teillet 1992) was used to transform radiance to reflectance.
Figure 1. Estimated albedo (%) for August 2000 in the D3 domain of MM5 over NE
Spain.
130 N. Pineda et al.

Step 3. Atmospheric correction. The SMAC algorithm (Rahman and Dedieu
1994) was employed to convert TOA reflectance to surface reflectance of each
image. SMAC input data, like water vapour content and columnar ozone content
was derived from the NOAA-14 TOVS sensor, data provided by the ATMOS User
Center of the German Remote Sensing Data Center (DFD). Aerosol optical depth
(at 550 nm) was obtained from the Aerosol Robotic Network (AERONET)
(Holben et al. 1998). For the thermal channels, a split-window technique was used.
Day and night land surface temperature were calculated with the Sobrino and
Raussoni (2000) algorithm. Input data are channels 4 and 5 brightness temperature,
surface emissivity and water vapour, obtained from the morning and afternoon
passes of the NOAA TOVS.
Step 4. Identification of cloud contaminated pixels and image compositing. In
order to create cloud-free monthly mean images, clouds were filtered using a
regional adaptation of the threshold procedure proposed by Derrien et al. (1993).
Once the clouds were masked, monthly composites were calculated, applying the
mean for the non-contaminated pixels, channel-by-channel. Surface parameters
were estimated from these monthly composites.
3.2. Geophysical surface parameter estimation
Albedo was estimated assuming lambertian reflection. To derive broad-band
albedo from AVHRR, a narrow-band to broad-band conversion method (Saunders
1990) that combines weighted AVHRR channels 1 and 2 was used. From the
channel weights proposed by different authors, the ones estimated for the Iberian
Peninsula by Valiente et al. (1995) were used.
Emissivity at 9 nm was calculated using the Valor and Caselles (1996) method,
which estimate emissivity from AVHRR NDVI images.
Thermal Inertia (TI) was estimated according to the simple formulation for
calculating TI from remote sensing data, given by Price (1977). The Sobrino and El
Kharraz (1999) Price’s model adaptation was used. Calculated TI values with
AVHRR were higher than USGS ones for all the land-use categories, especially in
coastal and high altitude regions as shown in figure 2. Low altitude flat and
continental regions are the most similar to MM5 global TI values. It seems that sea
and mountain effects in TI are not contemplated in the MM5 tabulated TI values.
4. Land-use category equivalence
Albedo, emissivity and TI were derived from the AVHRR data set, but
roughness and soil moisture can not be estimated with optical sensors. To run the
simulation in MM5 with the CORINE land-use map, it was necessary to obtain an
estimation of these surface parameters for the CORINE categories. In order to
relate current tabulated MM5 USGS roughness and moisture values to the new
land-use categories, equivalence between the land-use maps was carried out. The
NDVI vegetation index was used to establish category equivalencies, making the
assumption that surface roughness is mainly due to vegetation cover and that there
is a good relation between soil moisture and vegetation presence.
For the land-use category equivalence study, a set of 36 SPOT VGT S
10
(10-day
synthesis) NDVI images of the year 2000 was used. Although NDVI vegetation
index can be calculated with AVHRR images, the SPOT VGT S
10
data set has
more images (one every 10 days), better calibration and geographical quality than
Recent Advances in Quantitative Remote Sensing 131

AVHRR data, thus it is more suitable for annual vegetation cycle variation
analysis.
Calculated parameters for the CORINE categories are shown in table 1. Winter
and summer periods correspond to three-month means. The equivalence between
land-use categories is shown on the right part of table 1. The final column
corresponds to Gower metric statistics (Gower 1971), used to compare annual
NDVI cycle similarities and to establish category equivalencies. As Gower statistics
value gets lower, annual NDVI patterns between categories are more similar.
In figure 3 annual NDVI cycles of some CORINE and USGS land-use
categories are represented. Cropland categories are represented in figure 3(a).
CORINE ‘Non-Irrigated Arable Land’ (cat.12) has a similar NDVI annual pattern
with its comparable class in USGS, dry crops (cat.102), but with higher NDVI
values. The same occurs for the irrigated crops (CORINE cat.13 and USGS
cat.103), but rice fields (CORINE cat.14) have a completely different NDVI annual
pattern. According to Gower statistics roughness and moisture values for rice fields
were taken from USGS cat.103 (Gower: 21.7). Figure 3(b) shows a good
equivalence between USGS cat.111 ‘Deciduous Broadleaf Forest’ and CORINE
cat.23 ‘Broad-Leaved Forest’ (Gower: 5.7) and a less good equivalence between
USGS cat.114 ‘Evergreen Needleleaf Forest’ and CORINE-24 ‘Coniferous Forest’
(Gower: 20,4). Shrubland categories are represented in figure 3(c). CORINE
‘Sclerophyllous Vegetation’ (cat.28) should theoretically correspond to USGS
‘Shrubland’ (cat.108), but according to the NDVI pattern and Gower metrics (21.3
for cat.109 and 45.3 for cat 108), it seems more appropriate to take values of
roughness and moisture from class ‘Mix Shrubland/Grassland’ (cat.109) for the
‘Sclerophyllous Vegetation’ category. Finally, besides Shrubland categories, grass-
land (USGS cat.107) is represented in figure 3(c). The NDVI is quite different
between summer and winter periods in this category, due to the snow presence
between November and April.
Figure 2. Differences (CORINE – USGS) in thermal inertia W m
{2
k
{1
s
1
=
2

over land for
August 2000.
132 N. Pineda et al.

Table 1. Surface parameters for CORINE land-use categories for winter (W) and summer (S) and equivalencies with USGS land-use map. The third
column is percent of category presence in the land surface of D3 domain.
CORINE
Category
description %
Albedo
(%)
Moisture
avail. (%)
Emissiv.
(% at 9 mm)
Roughness
length (cm)
Thermal Inertia
Wm
{2
k
{1
s
1=2

Equivalencies
W S W S W S W S W S CORINE USGS
USGS
category
description GOWER
1–11 Urban 0.43 16.8 17.8 10 10 93.5 93.7 50 50 2283 2895 1–11 101 Urban and
Built-Up Land
6.2
12 Non-Irrigated
Arable Land
21.77 18.4 20.7 60 30 96.0 95.8 5 15 2064 2332 12 102 Dryland Cropland
and Pasture
15.2
13 Permanently
Irrigated Land
3.23 19.7 20.5 50 50 95.4 96.5 5 15 1582 2322 13 103
Irrigated Cropland
and Pasture
11.2
14 Rice Fields 0.18 15.7 17.0 50 50 92.8 97.5 5 15 2618 3977 14 103 21.7
15 Vineyards 3.91 18.5 19.9 60 35 95.2 95.6 20 20 1906 2321 15 106
Crops/Wood
mosaic
23.5
16 Fruit Trees &
Berry Plantations
1.97 16.5 17.1 60 35 96.4 95.9 20 20 2097 2831 16 106 27
17 Olive Groves 0.51 15.8 17.1 60 35 96.4 95.3 20 20 2010 2537 17 106 15
18 Pastures 2.32 16.0 17.1 60 30 97.6 98.9 5 15 2636 3348 18 102 Dryland Cropland
and Pasture
36.1
19 Annual Crops &
Permanent Crops
0.50 18.5 20.4 60 35 97.1 93.7 20 20 2093 2410 19 106
Crops/Wood
mosaic
22.6
20 Complex Cultivation
Patterns
9.21 17.4 18.8 60 35 96.5 96.7 20 20 2219 2619 20 106 33.9
21 Mixed Agriculture &
Natural Vegetation
5.94 17.8 19.4 60 35 95.7 95.0 20 20 1655 2211 21 106 24.5
22 Agro-Forestry Areas 0.08 17.8 19.4 60 35 95.7 95.0 20 20 1655 2211 22 106 19
23 Broad-Leaved
Forest
10.03 14.4 16.1 60 30 97.1 98.5 50 50 2771 3269 23 111 Deciduous
Broadleaf
Forest
5.7
24 Coniferous Forest 12.56 14.1 14.3 60 30 97.3 97.4 50 50 2394 2983 24 114 Evergreen
Needleleaf
Forest
20.4
25 Mixed Forest 3.15 14.1 14.8 60 30 97.4 98.4 50 50 2659 3283 25 115 Mixed Forest 4.8
Recent Advances in Quantitative Remote Sensing 133

Table 1. (Continued)
CORINE
Category
description %
Albedo
(%)
Moisture
avail. (%)
Emissiv.
(% at 9 mm)
Roughness
length (cm)
Thermal Inertia
Wm
{2
k
{1
s
1=2

Equivalencies
W S W S W S W S W S CORINE USGS
USGS
category
description GOWER
26 Natural Grassland 4.30 17.7 17.2 30 15 96.1 98.0 0.10 0.12 2505 2889 26 107 Grassland 42
27 Moors & Heathland 1.19 16.0 16.4 25 15 97.4 98.4 10 11 2500 3216 27 109
Mix Shrubland/
Grassland
28.4
28 Sclerophyllous
Vegetation
6.41 15.6 15.8 25 15 96.4 96.0 10 11 2135 2575 28 109 21.3
29 Transitional
Woodland-Shrub
6.06 15.4 15.6 25 15 96.7 96.5 10 11 2112 2657 29 109 11.5
30 Beaches, Dunes &
Sand Plains
0.06 16.8 17.1 5 2 94.5 96.9 10 10 2752 2778 30 119
Barren or
Sparsely
Vegetated
66.4
31 Bare Rock 0.62 17.3 16.9 5 2 96.2 96.5 10 10 2619 2948 31 119 55.8
32 Sparsely Vegetated
Areas
0.98 19.6 21.1 5 2 95.1 94.3 10 10 1631 2023 32 119 72
33 Burnt Areas 0.19 13.3 13.8 5 2 97.0 96.4 10 10 1966 2600 33 119 113.2
34 Glaciers & Perpetual
Snow
0.00 51.3 41.5 95 95 99.7 96.1 5 5 359 418 34 124 Snow or Ice –
35–38 Inland Marshes
Peatbogs, Salines
0.06 14.7 15.5 75 60 94.6 95.8 20 20 2611 3329 35–38 117
Herbaceous
Wetlands
–
39 Intertidal Flats 0.01 15.1 15.8 75 80 93.7 95.6 20 20 2802 3901 39 117 –
40–43 Inland Water 0.16 7.8 7.7 100 100 97.8 97.8 0.01 0.01 5916 7116 40–43 116
Water Bodies
–
44 Sea & Ocean – 7.8 7.7 100 100 97.8 97.8 0.01 0.01 7055 7829 44 116 –
134 N. Pineda et al.

Figure 3. Annual SPOT VGT S
10
NDVI cycles for year 2000 of some CORINE and USGS
land-use categories: (a) Cropland categories, (b) Forest categories and (c) Shrubland
categories.
Recent Advances in Quantitative Remote Sensing 135

5. Meteorological model
The mesoscale meteorological model used in this study is MM5. It is a
community mesoscale model widely used for numerical weather prediction, air
quality studies, and hydrological studies. On the smaller meso-beta and meso-
gamma scales (2–200 km), MM5 can be applied to studies involving mesoscale
convective systems, fronts, land-sea breeze, mountain-valley circulations, and urban
heat islands (MMD/NCAR 2001).
Two major changes were introduced into the model in order to evaluate some
improvements in its performance. A new set of physical soil properties was
introduced to MM5 via a more accurate land-use map. Some modifications were
carried out to adapt the new information to the model. These new parameters are
tabulated in table 1 for winter and summer cases.
To evaluate the behaviour of MM5 with these modifications, two simulations
were performed. A basic case with the default values of physical soil properties and
the land-use map of the USGS, and a simulation with the new parameters and the
CORINE land-use map. Hereinafter the simulation with the CORINE land-use
map and the new physical soil properties will be named CORINE simulation, and
the simulation with the default values of MM5, USGS simulation.
5.1. Model configuration
Four nested domains were selected (figure 4), which essentially covered Europe
(Domain 1, D1), the Iberian Peninsula (Domain 2, D2), NE of the Iberian
Peninsula (Domain 3, D3) and Catalonia (Domain 4, D4). D1 has 50635 grid
points in the horizontal with 72 km grid-point spacing. D2 has 61649 24 km cells.
D3 has 93693 6 km cells. D4 has 1516151 2 km cells. The vertical resolution
consisted for all domains of 23 s-layers, with the lowest one situated approximately
at 36 m AGL.
The model uses the Mellor-Yamada scheme as used in the ETA model (Janjic
1994) for the boundary layer parameterization, the Anthens-Kuo and Kain-Fritsch
(Kain and Fritsch 1993) cumulus scheme for domains 1 and 2, and no cumulus
parameterization for domains 3 and 4. A simple ice explicit moisture scheme, a
cloud-radiation scheme, and the five-layer soil model are the other physical
parameterizations used in these simulations.
Figure 4. MM5 domain definition over Southern Europe.
136 N. Pineda et al.

Initialization and boundary conditions for the mesoscale model were introduced
with analysis data of the European Center for Medium-Range Weather Forecasts
(ECMWF) global model. Data were available at a 1‡ resolution (100 km approx. at
the working latitude) at the standard pressure levels every 6 hours.
5.2. Meteorological situation
A synoptic situation was studied in order to evaluate the performance of the
model working with a new land use map and physical parameters obtained with the
methodology explained before. The meteorological case was 14 August 2000. A
situation with weak synoptic forcing was chosen, so that mesoscale phenomena
induced by the particular topography of the region, and the physical properties of
the soil would be dominant.
Synoptic situation of 14 August corresponds to a typical summertime
barometric swamp over the Iberian Peninsula. At surface, the high pressure area
is centred over the south Atlantic ocean, with the anticyclonic wedge affecting most
parts of the Iberian Peninsula, producing a typical barometric swamp along the
easterly part of the Peninsula. Surface winds are low. This fact, and the strong
daily solar heating, produced the development of mesoscale phenomena. These
phenomena in the region are mainly sea breezes, up-slope and down-slope winds
and valley channelled winds. The heating during 14 August was so intense that a
thermal low started to develop in the south-east of the Iberian Peninsula. In height,
a zonal flow blows aloft the Peninsula veering to the south-east having north-
westerly winds affecting the north-east of the Iberian Peninsula. The winds aloft
exhibit the maximum velocities north-west of the Peninsula.
6. Model results
The new land-use map and the physical soil parameters were introduced to
MM5. Three regions of domain 4 are described in order to illustrate the differences
produced by those changes. The differences in moisture availability between the two
model simulations are displayed in figure 5, with the location of three test regions
Figure 5. Differences in Moisture Availability (%) between the two MM5 simulations
(CORINE-USGS). Test regions: (1) Barcelona Region, (2) Garrotxa Region and (3)
Pyrenees.
Recent Advances in Quantitative Remote Sensing 137

(1. Barcelona, 2. Garrotxa, 3. Pyrenees). A first area with an introduction of several
urban zones around Barcelona city, a second one with a large extension of a land-
use homogeneous change, and a third one comprising a mountain chain ranging
from 1000 to 3000 m.
6.1. Barcelona region
In figure 6(a) differences in ground temperature (CORINE-USGS) are
displayed. The values are percentages relative to USGS ground temperature. The
regions with a new reclassification of the land-use to urban soil appear to be
warmer in CORINE simulation. The difference range varies from 25% to 225%.
These differences in ground temperature affect air temperature at the first layer of
the model in a lower way, with variations at 14 UTC ranging from 2.5% to 21.5%
(figure 6(b)). Such variations are due to the change of soil physical parameters,
changes that are sufficiently important to produce variations in cloud develop-
ments, and as a result, precipitation patterns are slightly different.
The variation of surface budgets between the two simulations is able to change
the surface and aloft wind fields in several regions of domain 4. Figure 7(a) shows
the surface wind field of the geographical area of Barcelona obtained with the
USGS land-use map at 14 UTC. A sea breeze is well established, blowing inland
over the entire domain, with southerly winds veering south-easterly leeward the
coast mountain range. Figure 7(b) shows the differences between the two surface
wind fields simulated by CORINE and USGS land-use map. The arrows are the
wind field difference between CORINE and USGS, and the coloured areas
represent the differences in wind velocity. With the modifications introduced to
MM5, the sea breeze is weaker inland, and over the sea has a more south-westerly
component. The differences in the velocity magnitude are lower in comparison to
the direction differences. It is important to note that in other regions where
convective flows are simulated, surface winds can vary in 5 ms
21
. Such differences
are due to the modifications of the soil physical parameters, which in turn modify
the placement of the vertical thermals that produce the cumulus clouds.
Figure 6. Barcelona Region temperature (‡C) differences (CORINE-USGS): (a) ground
temperature and (b) first level temperature.
138 N. Pineda et al.

6.2. Garrotxa region
The change of the land-use map produces a general increase of thermal inertia,
roughness length, emissivity and albedo, and a decrease of moisture availability in
the Garrotxa Region. In the USGS map, the land-use is cropland-woodland
mosaic, and with the implementation of the CORINE land-use map, that region is
considered basically as broad-leaved forest. The values of the physical soil
parameters with the two maps are tabulated in table 2.
The more important differences in physical parameters reside on thermal inertia
and roughness length. With these values, a warmer layer near the ground should be
simulated during night with CORINE map, and colder during daytime. However,
the model seems to be more sensitive to moisture availability than to the rest of the
other parameters.
In figure 8, the evolution of temperature at a location within the region is
shown. At night, the ground temperature (a) is higher with the CORINE simulation
due to the high values of thermal inertia associated to the land-use class of that
point, nevertheless, during daytime the temperature with CORINE continues over
the USGS one. This performance can be explained due to the lower values of
moisture availability, producing a decrease of the latent heat flux. In the evolution
of the first layer temperature (b) the two simulations have a similar behaviour, with
a faster response to a decrease of the incident short wave downward radiation by
USGS case, because of the lower value of TI. The decrease of incident radiation is
caused by clouds, with slight variations in extension dimension between
simulations.
(a)
(b)
Figure 7. Barcelona region differences (CORINE-USGS) in (a) surface wind field of the
basic case and (b) surface wind field.
Table 2. Physical parameters for the dominant category in the Garrotxa Region (2) for the
two land-use maps.
Land-use category
Albedo
(%)
Moisture
avail. (%)
Emissivity
(% at 9 mm)
Roughness
length (cm)
Thermal inertia
Wm
{2
k
{1
s
1
=
2

USGS Crop./Woodland
mosaic (cat.106)
16 35 93 20 1672
CORINE Broad-leaved
forest (cat.23)
16.1 30 98.5 50 3269
Recent Advances in Quantitative Remote Sensing 139

6.3. Pyrenees
With the CORINE land-use map and the new values of the physical soil
parameters the model simulates higher ground temperatures in the Pyrenees
mountain range during all the simulated period. These differences can reach values
up to 90% of the value simulated with USGS. That behaviour can be explained due
to the low values of moisture availability associated to the high peaks of the
Pyrenees with the new reclassification. Complex variations in the wind fields are
produced within this region due to the high complexity of the topography and the
variations introduced with the new parameters. Note that depending on the
location of cumulus development, the aloft winds can vary in an important way. A
cross section of the wind field and the mixing ratio is displayed in figure 9, the top
panel shows the cross section for the results of the CORINE simulation, and the
bottom panel the cross section for the USGS simulation.
The differences are clearly evident, with an important updraft over the middle of
Figure 8. Daily evolution of (a) ground temperature (‡ C) and (b) first layer temperature (‡C)
at 42‡8’24@ N2‡30’ E in Garrotxa Region (2) (dashed line CORINE, solid line USGS).
Figure 9. Cross section from north (left) to south (right) of the wind field (white arrows)
and the mixing ratio (kg/kg) (coloured scale with topography in black) along 1.55‡ E
at 15 UTC (CORINE upper, USGS bottom panel).
140 N. Pineda et al.

the region injecting moist air aloft in the CORINE simulation. This updraft is less
important in the USGS case, with a boundary layer warmer and dryer in
comparison with the CORINE one.
6.4. Measurement comparison
An extensive net of ground measurements of wind and temperature from the
Catalan Meteorological Service is available at the area of domain 4. Comparisons
with these data are done by calculating the root-mean square error (RMSE), the
BIAS and for the vector wind, the root-mean square vector error (RMSVE).
Temperature at 2 m and wind at 10 m are evaluated.
Figure 10 shows the evolution of the RMSE and the BIAS for temperature at
2 m. Both statistics are calculated in basis to the observations. The errors in both
simulations are nearly equal. The CORINE simulation is slightly warmer during
night-time, as can be appreciated with the bias evolution, and slightly colder during
daytime. The model is not able to reproduce the daily variations of the temperature,
with an overestimation at night and underestimation during daytime.
The evolution of the RMSVE (figure 11) is similar with small differences.
CORINE simulation gives little better results in RMSVE, but the RMSE of the
magnitude of the velocity is slightly higher than in USGS results. The surface winds
appear to be lower in CORINE simulation than in USGS one, and in both cases
are lower than the observations during daytime, and higher at night. The evolution
of RMSE and RMSVE for the wind shows a little improvement in wind direction
with the CORINE land-use map.
Figure 10. Daily evolution of the RMSE (upper curves) and the BIAS (bottom curves) for
the temperature (‡C) at 2 m (dashed line CORINE, solid line USGS).
Figure 11. Daily evolution of the RMSVE for the winds (ms
21
) at 10 m (dashed line CORINE,
solid line USGS).
Recent Advances in Quantitative Remote Sensing 141

7. Summary
An updating of the local physical parameters used in MM5 was intended, in
order to get a better performance of the model and to have better regional
simulations. Remotely sensed data of the year 2000 was used, as well as a recent
land-use map. Results show that differences in surface parameters basically rely on
thermal inertia. Besides, land-use map comparison had shown important differences
between classifications that also affect the final composition of surface parameters
that get into the model.
New values of the physical soil parameters have been introduced to MM5 with a
new land-use map. These modifications have been sufficiently significant to produce
variations in the performance of the model. The cloud development differs basically
in the location and dimensions of the clouds, that drives to a different superficial
radiative budget affecting the evolution of air temperature at low levels. The
different results in cumulus simulation produced important differences in the
surface wind field and the updrafts. The changes introduced are sufficiently
significant to obtain slight variations in the pattern of accumulated precipitation for
the simulated period.
Comparisons with ground measurements of wind and temperature have been
done in the test regions. Similar errors are obtained with the two land-use maps and
physical parameters, without a clear improvement in the performance of the
meteorological model.
This work contributes evidence to the high influence of surface scheme in
applications of mesoscale models at high horizontal resolution. In the context of a
dialogue between remote sensing scientists and numerical climate modellers, it is
expected that more research should be done to investigate the sensibility of
mesoscale models to improvements in the surface properties characterization.
Acknowledgments
The authors wish to thank the Environmental European Agency for providing
the EEA NATLAN-2000 land-use information package; the CREPAD-INTA
(Instituto Nacional de Te`cnica Aerospacial, Spain) for the NOAA-14 AVHRR
data; the AUC-DLR Center (German Aerospace Center) for the NOAA-14 TOVS
data; the VITO (Flemish Inst. Technological Research, Belgium) for providing the
SPOT VGT S
10
NDVI images; the Spanish Meteorological Institute (INM) for
providing data from the ECMWF; and the Catalan Meteorological Service (SMC)
for providing surface station data for validation. Simulations were run on an HP
Exemplar V2500 belonging to CESCA (Centre de Supercomputacio´ de Catalunya).
This work was developed under projects IMMPACTE and CICYT REN2000-1754-
C02-01/CLI.
References
ANDERSON, J. R., HARDY, E. E., ROACH, J. T., and WITMER, R. E., 1976, A land use and
land cover classification system for use with remote sensor data: U.S. Geological
Survey Professional Paper 964.
C
HEN, F., and DUDHIA, J., 2001, Coupling an advanced land surface-hydrology model with
the Penn State-NCAR MM5 modelling system. Part I: Model implementation and
sensitivity. Monthly Weather Review, 129, 569–585.
D
ERRIEN, M., FARKI, B., HARANG, L., LEGLE
´
AU, H., NOYALET, A., POCHIC, D., and
142 N. Pineda et al.

SAIROUNI, A., 1993, Automatic cloud detection applied to NOAA-11/AVHRR
imagery. Remote Sensing of Environment, 46, 246–267.
D
UDHIA, J., 1993, A non-hydrostatic version of the Penn State-NCAR mesoscale model:
validation tests and simulation of an Atlantic cyclone and cold front. Monthly
Weather Review, 121, 1493–1513.
E
UROPEAN ENVIRONMENT AGENCY, 2000, CORINE Land Cover (1 : 100 000), NATure/
LANd Cover information package. http://natlan.eea.eu.int
G
OWER, J. C., 1971, A general coefficient of similarity and its properties. Biometrics, 27,
857–871.
H
OLBEN, B. N., ECK, T. F., SLUTSKER, I., TANRE, D., BUIS, J. P., SETZER, A., VERMOTE, E.,
R
EAGAN, J. A., KAUFMAN, Y., NAKAJIMA, T., LAVENU, F., JANKOWIAK, I., and
S
MIRNOV, A., 1998, AERONET – a federated instrument network and data archive
for aerosol characterization. Remote Sensing of Environment, 66, 1–16.
M
ESOSCALE AND MICROSCALE METEOROLOGY DIVISION,NATIONAL CENTER FOR ATMO-
SPHERIC
RESEARCH, 2001, PSU/NCAR Mesoscale modeling system tutorial class
notes and user’s guide: MM5 Modeling system version 3, June 2001.
NOAA, 1998, Polar Orbiter Data User’s Guide (November 1998 version) http://
www2.ncdc.noaa.gov/docs/podug/index.htm
P
RICE, J. C., 1977, Thermal inertia mapping: a new view of the earth. Journal of Geophysical
Research, 82, 2582–2590.
S
AUNDERS, R. W., 1990, The determination of broad band surface albedo from AVHRR
visible and near-infrared radiances. International Journal of Remote Sensing, 11,
49–67.
S
OBRINO, J. A., and EL KHARRAZ, M. H., 1999, Combining afternoon and morning NOAA
satellites for thermal inertia estimation. 2. Methodology and application. Journal of
Geophysical Research, 104, 9455–9465.
S
OBRINO, J. A., and RAISSOUNI, N., 2000, Toward remote sensing methods for land cover
dynamic monitoring. Application to Morocco. International Journal of Remote
Sensing, 20, 353–366.
T
EILLET, P. M., 1992, An algorithm for the radiometric and atmospheric correction of
AVHRR data in the solar reflective channels. Remote Sensing of Environment, 41,
185–195.
V
ALIENTE, J. A., NUNEZ, M., LOPEZ-BAEZA, E., and MORENO, J., 1995, Narrow-band to
broad-band conversion for Meteosat-visible channel and broad-band albedo using
both AVHRR-1 and -2 channels. International Journal of Remote Sensing, 16,
1147–1166.
V
ALOR, E., and CASELLES, V., 1996, Mapping land surface emissivity from NDVI:
application to European, African, and South American areas. Remote Sensing of
Environment, 57, 167–184.
Recent Advances in Quantitative Remote Sensing 143