Performance of the Synergy

The information on vegetation coverage or LAI is essential for estimating both energy and CO2 fluxes from a plant canopy. We derived an intimate relationship between LAI and NDVI from our experimental data for over 6 years (Fig. 9). The relationship was consistent and robust for the years. The effects of the other plant conditions (leaf angle, etc.) and measuring configurations (sun elevation, etc.) on the relationship were also examined using a canopy reflectance model SAIL+PROSPECT (Jacquemoud, 1993), which indicated the reliability of the relationship. We incorporated this relationship between NDVI and LAI into the SVAT model so that remote sensing measurements can be used in the parameterization process. It was confirmed that NDVI values could be accurately simulated as long as the SVAT model was well calibrated. Consequently, the coupled model was able

Figure 9: Relationship between leaf area index and spectral index NDVI in soybean canopies.

to simulate the surface temperature, NDVI as well as all components of ecosystem CO2 fluxes and biomass growth.

As discussed in previous Section 6.1, the parameterization, that is, tuning of a process model with remotely sensed data is probably the most effective approach to utilize infrequent and multisource remote sensing data for estimating the dynamics of various ecosystem components (Olioso et al., 2002; Inoue, 2003). The optimization procedures of the SVAT model using the remotely sensed data are shown in Fig. 10. The structure of the SVAT model is also schematically indicated. Remotely sensed signatures in optical, thermal, and microwave domains are to be used to drive this parameterization procedure. Fig. 11 shows an example result of parameterization where the error between measured and simulated NDVI values was minimized through iterative optimization of the initial soil moisture content; the NDVI is the key parameter for optimization in this case. The initial soil moisture content was assumed to be unknown since it is not routinely available but one of the most important input data to properly run the SVAT model. After parameterization, the NDVI was best simulated with the initial soil moisture of 0.456, which was nearly equal to the independently measured value of soil moisture, 0.46. The dynamic changes in SSFCO2, net photosynthesis, and their sum ESFCO2, as well as the plant growth were also simulated for the entire growth period. Seasonal changes of LAI and DM production agreed well with those by destructive measurements. Simulated values of net photosynthesis showed reasonable diurnal and seasonal changes during the entire period. The simulated ESFCO2 was in good agreement with the independent measurements of ESFCO2 by the eddy covariance method during both early and

Figure 9: Relationship between leaf area index and spectral index NDVI in soybean canopies.

Meteorological forcing plant parameters soil parameters

SVAT model

SVAT model

Model for RS signature

Simulated

> Surface temperature

> Spectral reflectance

Model for RS signature

Simulated

> Surface temperature

> Spectral reflectance minimization of error tuning i

Remotely-sensed

> Surface temperature

> Spectral reflectance

Ecophysiological Variables

Figure 10: Structure of the SVAT model and the procedure for parameterization of the model using remotely sensed signature. The final output results (bottom) can be derived after tuning through iterative optimization of parameter or state variables. LAI: leaf area index; gs: stomatal conductance; An: leaf net photosynthesis; Anc: canopy net photosynthesis; Tr: transpiration; ET: Evapotranspiration; RS: remote sensing.

late growth stages. Simulation results of canopy transpiration and evapotranspiration also agreed well with those measurements by stem heat balance method and eddy covariance method, respectively. Furthermore, the simulated and measured canopy surface temperatures agreed quite well each other, which suggests that remotely sensed canopy temperature will also be used for calibration of the SVAT model instead of optical measurements. Optimization trials with various sets of remote sensing measurements suggested that a set of only a few remote sensing measurements could have a significant effect on improvement of simulation results.

220 230 2402 Day of year

220 230 2402 Day of year

200 210 220 230 240 250 260 270 Day of year

Iterative optimization of the initial soil water content (SWCi) using NDVI.

SWCi=0.456

0.5

a

0.4

0.3

p

02

©

ft

0.1

a

0.0

— simulated dry biomass • measured dry biomass

m/

| • /

200 210 220 230 240 250 260 270 Day of year

f sim-Soil CO2 flux 1 I measured-Ecosystem ^---|_CO2 flux by Eddy C.

jfsim-Ecosystem C02 flU^'fs^-Vegetation C02 fluXJ

jfsim-Ecosystem C02 flU^'fs^-Vegetation C02 fluXJ

200 210 220 230 240 250 260 270 Day of year

[9 2

20 2

12 2

Day of year

Late growth stage

- sim NVFc02

- sim SMFCO2 sim ESFCO2

- ESFCO2 by eddy covariance

Late growth stage

- sim NVFc02

- sim SMFCO2 sim ESFCO2

- ESFCO2 by eddy covariance

60

50

1>

40

30

¡3

20

H

10

measured surface temperature (C) - simulated surface temperature (C)

234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262

Day of year

Figure 11: Dynamic change of ecosystem CO2 (ESFCO2) flux estimated by the synergy of remote sensing and process model - a case study for soybean field. NVFCO2: net vegetation CO2 flux; SMFCO2: soil microbial CO2 flux (For colour version, see Colour Plate Section).

Was this article helpful?

0 0

Post a comment