Estimating NPP at an Ecosystem Scale based on fAPAR from Remotely Sensed Information A Case Study

A comparative study on the relationship between the fAPAR of various plant canopies and seven vegetation indices was conducted based on airborne remote sensing data (Inoue et al., 2001). During this experiment, well-calibrated airborne optical measurements at fine spatial resolution were obtained simultaneously with ground-based hyperspectral and detailed plant measurements. This may be a useful case study that investigates the realistic relationship between vegetation indices and fAPAR, since there are few experimental studies at an ecosystem or landscape scales based on accurate ground-based data and airborne remote sensing data.

4.2.1 Experimental Site

An experimental study was conducted in agricultural fields at the National Agricultural Research Center (Tsukuba, Japan; 36°010N, 140°070E, 25 m above sea level) and surrounding farmland areas. The size of the whole area was approximately 60 km2, where a wide range of crop fields (rice, soybean, corn, peanut, etc.) and natural vegetation areas (tree, turf grass, bush, etc.) were included.

4.2.2 Airborne and Ground-Based Remote Sensing Measurements

An airborne multispectral scanner (AZM, Nakanihon-kouku, Co. Inc., Japan) was used for remote sensing of the study area. Airborne spectral images were taken at eleven optical bands (474, 553, 656, 675, 848, 1,089, 1,193, 1,623, 2,044, 2,135, and 2,207 nm) and one thermal band (9.3 mm) from the altitude of 1,000 m around midday. The spatial resolution on the ground was 1.25 m. The digital images were converted to reflectance images using calibration sources. No atmospheric correction was made because the sky condition was perfectly clear and humidity was low (49%). Ground-based hyperspectral reflectance and surface temperatures were measured using a portable radiometer (FS-FR1000, ASD, USA) and infrared thermometers (Model4000, Everest Inc., USA), respectively. The reflectance spectra (380-2,500 nm, 1 nm resolution) were taken over more than 50 different uniform target areas including rice, soybean, corn, grassland, turf, vegetables, bare soil, asphalt, concrete, water pond, river, etc. Infrared surface temperatures were measured for 10 different uniform targets (bare soils, rice canopies, water pond, asphalt, concrete, etc.). Emissivity was assumed to be 0.98 for all targets. Temperature data for the time window of airborne observation were averaged and used for calibration.

4.2.3 Measurements of fAPAR and Plant Parameters

Plant data such as LAI and biomass were estimated from destructive measurements based on plant sampling. The LAI was measured using an optical leaf area meter (AAM8, Hayashi-denkoh Co. Ltd., Japan) for green leaves only. Wet and dry biomass was measured for green leaves, stems, roots, and senescent plant parts, respectively.

The fAPAR was estimated from the budget of four components of the photosynthetic photon flux density (PPFD) measurements, i.e., downward and upward over and below canopies. These PPFD values were measured using a meter-long PPFD sensor (LI-191SA, Li-Cor Inc., USA) several times, and were averaged so that they were representative for each canopy. The fAPAR data were taken at near the time of remotely sensing measurements; these data were not daily average values of fAPAR, but the instantaneous values near midday. It is well known that the value of fAPAR is not consistent during a day but relatively high near sunrise and sunset periods; nevertheless, the midday fAPAR may be representative enough for the daily average of fAPAR, because they are highly correlated with each other due to extremely low solar radiation during those periods of low sun elevation.

4.2.4 Estimating fAPAR and Plant Productivity from Spectral Indices

From the data for 107 different canopies, a close linear relationship was obtained between the fraction of intercepted radiation by a plant canopy (fIPAR) and fAPAR:

These two parameters are both often used in simple process models, but it should be noted that the fIPAR is smaller than fAPAR by 5%.

The relationship of fAPAR with the seven different vegetation indices, NDVI, RVI, SAVI (Huete, 1988), MSAVI (Qi et al., 1994), WDVI (Clevers, 1989), GEMI (Pinty and Verstraete, 1992), and EVI (Huete et al., 1999) was compared. Definitions for VIs are as follows:


with a optimized value of L


Figure 3: Relationship between fAPAR of various plant canopies and vegetation index NDVI derived from airborne remote sensing images (For colour version, see Colour Plate Section).


Figure 3: Relationship between fAPAR of various plant canopies and vegetation index NDVI derived from airborne remote sensing images (For colour version, see Colour Plate Section).

The fAPAR was best correlated with NDVI (Fig. 3), which was of better linearity than the other VIs; RVI (r2 - 0.55), SAVI (r2 - 0.73), MSAVI (r2 - 0.62), WDVI (r2 - 0.61), GEMI (r2 - 0.67), and EVI (r2 - 0.68). The regression equation for NDVI, fAPAR - 1.176NDVI - 0.145 r2 - 0.84 (13)

agreed well with the following theoretically derived equation (Myneni and Williams, 1994), fAPAR - 1.164NDVI - 0.143. (14)

Since the regression equation (13) was derived from a data set for 2 years and for a wide range of plant canopies including crops, vegetables, and woods, it will be a useful basis for various applications. The deviation of individual data points from the regression line, i.e., error of the estimations is large in some cases, but this is the inherent limitation to this simple approach, which uses only two spectral wavelengths.

On the basis of this equation, a fine resolution map of fAPAR was generated (Inoue et al., 2001), from which plant productivity could be estimated at the landscape scale using measured RUE and PAR data. Since the daily global solar radiation can be estimated from satellite imagery (GMS) at 5-km resolution, it is possible to estimate daily productivity for specific biomes or areas. During the midsummer growing season, typical values of the fAPAR for uniform areas of rice-paddies, upland-crops, and urban zone were 0.3, 0.25, and 0.1, respectively. It is noticeable that the regional average of fAPAR for the typical and uniform rice areas was not more than 0.3, while it was 0.8-0.9 for rice canopies with nearly maximum LAI on the same day. The amount of carbon fixed into the area of rice-paddies, upland-crops, and urban areas was estimated to be 2.8, 2.4, and 1.0Cgm-2day-1. For example, on a clear-sky day with the global solar radiation (RAD) of 25 MJ m-2 day-1, assuming that RUE is 2.66 gDM MJ^PAR, PAR is 0.45 of RAD, and carbon content is 0.449 of plant DM, respectively. The values for cropped areas are comparable with typical values for crop species such as 300-400 gC m-2 y-1 (Goudriaan et al., 2001).

During the same experiment, the soil carbon content in agricultural fields was found to be highly correlated with spectral reflectance at 480 and 560 nm (r2 - 0.63; range of carbon content: 0.20-0.61% DM) based on the same data set. This result suggests that the annual change in the soil carbon content may be estimated from remotely sensed spectral reflectance. Nevertheless, further research is needed for more accurate and robust estimation.

Was this article helpful?

0 0

Post a comment