temporal trends of microwave emission from forest areas observed from satellite.ppt
TRANSCRIPT
Simonetta Paloscia, Emanuele Santi, Simone Pettinato, Marco Brogioni
CNR-IFAC, FlorencePaolo Ferrazzoli, Rachid Rahmoune
DISP, Tor Vergata University, Rome(Italy)
Microwave satellites demonstrated to be good sensors for investigating land surface features, and in particular soil moisture and vegetation cover, at both global and regional scales.
The retrieval of information on forests is crucial for all studies concerning global changes and carbon balance.
The temporal trends microwave emission measured by AMSR-E (Advanced Microwave Scanning Radiometer onboard Aqua) and ESA/SMOS (Soil Moisture Ocean Salinity) satellites were analyzed on some forest plots in Russia, China and Italy.
AMSR-E data (55°) at C (6.8GHz), X (10GHz), Ku (19GHz), and Ka (37GHz) bands, were collected during one year from May 2007 to April 2008
SMOS LC1 data al L (1.4GHz) band were collected from January to December 2010 and averaged between 37.5° and 47.5°. Samples affected by RFI were removed.
Seasonal trends of brightness temperatures (Tb) at different frequencies, in both H and V polarizations, were analyzed on the 3 test areas, together with the following microwave indexes:
Polarization Index: PI=(Tbv-Tbh)/0.5*(Tbv+Tbh) at both X- and Ku-bands;
Frequency Index: FI = [(TbvKu - TbvKa)+ (TbhKu + TbhKa)]/2;
Normalized Temperature: Tn=Tbh(C)/Tbv(Ka) or Tb(L)/Ts
The following 3 forest areas, have been studied by using the AMSR-E & SMOS sensors:
A Needle-leaved deciduous forest of Larix (Jiagedaqi) in China, characterized by cold winter with snowfalls (123°E/49.8°N);
A boreal Evergreen Spruce forest in Russia, with cold winters and snowfalls (60°E/50.5°N)
The Foreste Casentinesi in Italy, a mixed forest located in Central Italy and characterized by mild weather conditions (11.8°E/43.8°N)
The first 2 areas have already been selected in the past for investigations carried out by using SSM/I data
1
3 2
1. Russian forest (Evergreen)
2. Jagedaqi forest (China)
3. Foreste Casentinesi (Italy)
PIPI (X & Ku) shows a decreasing behavior in summer, due to the increase in leaf biomass, and an increasing trend in winter, due to the simultaneous decrease of biomass and presence of snow.
The trend of LAILAI has an opposite trend with respect to these curves.
The FI (Ku-Ka) shows 2 peaks, one in agreement with the development of tree LAILAI in summer, and the second one with snowfall in winter.
Jagedaqui
-2.00
-1.00
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
FI, L
AI
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
0.02
PI
-2.00-1.00
0.001.00
2.003.00
4.005.00
6.007.00
8.00
09
/03
/20
07
28
/04
/20
07
17
/06
/20
07
06
/08
/20
07
25
/09
/20
07
14
/11
/20
07
03
/01
/20
08
22
/02
/20
08
12
/04
/20
08
01
/06
/20
08
Date
LA
I, P
(cm
)
PI
LAILAI
FIFI
y = 0.7307x - 1.3418
R2 = 0.5985
-1
-0.5
0
0.5
1
1.5
2
2.5
3
3.5
0 1 2 3 4 5 6
LAI
FI(
Ku
-Ka)
LAILAI
FI
y = -0.0015x + 0.0102
R2 = 0.5894
0.00000
0.00200
0.00400
0.00600
0.00800
0.01000
0.01200
0.01400
0 1 2 3 4 5 6
LAI
PIK
u
PI(Ku)=0.01-0.0015 LAI (R2=0.59)
FI(Ku-Ka)=0.73-1.34 (R2=0.6) Winter data (snow) were not
considered
PI
LAILAILate snowfall
Tn=0.986-0.0023 R (R2=0.79) Monthly rainfall data were recorded at a nearby meteo
station and compared to averaged Tb data Winter data (snow) were not considered
Jagedaqui
y = -0.0023x + 0.9855
R2 = 0.7943
0.95
0.96
0.97
0.98
0.99
0 2 4 6 8 10 12
Rainfall (cm)
Tb
hC
/Tb
vK
a
SMOS Tb, normalized to surface temperatures estimated by ECMWF, was transformed into surface emissivity (Tn)
In winter (until DoY 80) the soil is frozen and covered by snow, with low permittivity and then emissivity is high.
Between DoY 90 and 120 there is a clear decreasing trend, associated to snow melting.
This effect is due to the strong variation of soil properties, from frozen to wet.
After this date, Tn increases again and shows variations partially related to soil moisture effects.
Tn
SMC
Melting
The snowfalls in winter affect both PI and FI.
FIFI shows a great sensitivity to snow but even to the variations of LAILAI in summer and spring time.
The variations of PI at XX and Ku band are similar to those in Jagedaqui.
0
1
2
3
4
5
6
09/0
3/20
07
28/0
4/20
07
17/0
6/20
07
06/0
8/20
07
25/0
9/20
07
14/1
1/20
07
03/0
1/20
08
22/0
2/20
08
12/0
4/20
08
01/0
6/20
08
Date
LAI,
P(cm
)
PIFI
LAILAI
y = -0.0012x + 0.0072
R2 = 0.5589
0
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0 1 2 3 4 5 6
LAI
PIK
u
PI(Ku)=0.007-0.0012 LAI (R2=0.56)
Winter data (snow) were not considered
PI
LAILAI
Russia
y = -0.0003x + 0.9899R2 = 0.5738
0.95
0.96
0.97
0.98
0.99
0 10 20 30 40 50 60 70
Rainfall (cm)
Tb
hC
/Tb
vKa
Tn=0.99-0.0003 R (R2=0.57) Monthly rainfall data were recorded at a nearby
meteo station and compared to averaged Tb data Winter data (snow) were not considered
In winter (until DoY 80) SMOS surface emissivity, Tn, shows values close to 1, when the soil is frozen and covered by snow, with low permittivity.
Between DoY 80 and 120 there is a clear decreasing trend, associated to snow melting.
This effect is due to the strong variation of soil properties, from frozen to wet.
However, the emissivity remains > 0.9 and does not show further variations related to soil moisture effects, due to the high forest density.
Tn
SMC
Melting
A mixed dense forest located in Tuscany, was selected as a temperate test area, where snowfalls are rather exceptional.
Due to the small dimensions and the heterogeneity of the area, a preliminary analysis was carried out by using a RGB Landsat image in order to better identify and geolocate the forest site.
The dimensions of the image are 40kmx40km. In the image, the area of about 20 km x 20km, corresponding to the AMSR-E acquisition, was indicated.
RGB Landsat image in the visible bands: R= Band 3 (0.63-0.69 µm)G= Band 2 (0.53-0.61 µm)B= Band 1 (0.45-0.52 µm)
Seasonal trends of the PI(Ku), FI, and LAILAI from 2006 to 2008. The annual trend of FI is in phase with the forest LAI, whereas the PI(Ku) is inversely related to it.
The X-band values were not used, since they were affected by strong RFI, probably originated by the radio transmitters close to this area.
FI, LAILAI PIKu
PI(Ku)=0.012-0.0009 LAI (R2=0.4)
FI(Ku-Ka)=0.98-1.44 (R2=0.65)
y = -0.0009x + 0.0125
R2 = 0.40590
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0 1 2 3 4 5 6
LAI
PIK
u
LAILAI
PI
y = 0.9855x - 1.4443
R2 = 0.6524
-1
0
1
2
3
4
5
0 1 2 3 4 5 6
LAI
FI(
Ku
-Ka)
LAILAI
FI
Emissivity data at L band collected with an airborne sensor on some dense forests in Tuscany showed a fairly high sensitivity to SMC at both H and V pol.
These trends have been confirmed by model simulations (Della Vecchia et al. 2010)
0.8
0.82
0.84
0.86
0.88
0.9
0.92
0.94
0 0.1 0.2 0.3 0.4 0.5
Soil Moisture Content (cm3/cm3)
e/Tn
ev
eh
TnV
TnH
Temporal trends of brightness temperature and related microwave indexes from AMSR-E & SMOS satellites were analyzed over three forest areas characterized by different climatic conditions and tree species.
At the higher frequencies, the frequency index between Ku and Ka bands is sensitive to the snow cycle, whereas the polarization index at both X and Ku bands is sensitive to the leaf cycle. Direct relationships between PI(Ku) and LAI, derived from ECOCLIMAP database, confirmed a high correlation between these two quantities.
Looking at SMOS data, the emissivity, obtained normalizing L band (1.4 GHz) emission to the surface temperature derived from ECMWF, shows a clear decrease, at both polarizations, which can be associated to the snow melting process and therefore to a soil moisture increase.
SMOS Tb, normalized to surface temperatures estimated by ECMWF, was transformed into surface emissivity
In winter (until DoY 80) the soil is frozen and covered by snow, with low permittivity and then emissivity is close to 1.
Between DoY 80 and 120 there is a clear decreasing trend, associated to snow melting.
This effect is due to the strong variation of soil properties, from frozen to wet.
However, the emissivity remains > 0.9 and does not show further variations related to soil moisture effects.
Soil moisture