AVHRRBased Vegetation Indices

Radiances measured by the AVHRR instrument onboard National Oceanic Atmospheric Administration (NOAA) polar-orbiting satellites can be used to monitor drought conditions because of their sensitivity to changes in leaf chlorophyll, moisture content, and thermal conditions (Gates, 1970; Myers, 1970). Over the last 20 years, these radiances were converted into indices that were used as proxies for estimating various vegetation conditions

(Kogan, 1997, 2001, 2002). The indices became indispensable sources of information in the absence of in situ data, whose measurements and delivery are affected by telecommunication problems, difficult access to environmentally marginal areas, economic disturbances, and political or military conflicts. In addition, indices have advantage over in situ data in terms of better spatial and temporal coverage and faster data availability.

The AVHRR-based indices used for monitoring vegetation can be divided into two groups: two-channel indices, and three-channel indices. The normalized differences vegetation index (NDVI) is derived from two-channel data, the visible (VIS, 0.58-0.68 ¡m), and near infrared (NIR, 0.72-1.1 ¡m) and is defined as NDVI = (NIR — VIS)/(NIR + VIS). The NDVI has been widely used for characterizing distribution of vegetation (Tarpley et al., 1984; Justice et al., 1985; Tucker and Sellers, 1986; Boken and Shaykewich, 2002) and for monitoring vegetation conditions (Kogan, 1995). However, this approach is insufficient for monitoring crops specifically during drought periods because crop health depends not only on the water stress but also on thermal conditions. Therefore, three-channel indices were introduced to monitor impacts of moisture and thermal stresses on the vegetation conditions.

New Method and Data

The new numerical method, introduced in the late 1980s, is based on the combination of VIS, NIR, and thermal (10.3-11.3 ¡m) channels (Kogan, 1997, 2001). This method is built on three basic environmental laws: law of minimum (LOM), law of tolerance (LOT), and the principle of carrying capacity (CC). The Leibig's LOM postulates that primary production is proportional to the amount of the most limiting resource contributing to growth and is at its lowest when one of the factors affecting it is at the extreme minimum. The Shelford's LOT states that the effect of each environmental factor on an organism or ecosystem is maximum or minimum when the environmental factor ranges between the limits of tolerance. With regard to these laws, the CC is defined as the maximal population size of a given species that resources of a habitat can support (Ehrlich et al., 1977; Orians, 1990).

The new method was applied to the NOAA global vegetation index (GVI) data set issued routinely since 1985 (Kidwell,1997). The GVI is produced by sampling the AVHRR-based 4-km (global area coverage format; GAC) daily radiances in the VIS, NIR, and IR (10.3-11.3 and 11.5-12.5 ¡m), which were truncated to 8-bit precision and mapped to a 16-km2 latitude/longitude grid. To minimize the cloud effects, these maps were composited over a seven-day period by saving radiances for the day that had the largest difference between NIR and VIS channels.

Because AVHRR-based radiances have both interannual and intra-annual noise (varying illumination and viewing, sensor degradation, satellite navigation and orbital drift, atmospheric and surface conditions, methods of data sampling and processing, communication and random errors), their removal is crucial for improving data interpretation. Therefore, the initial processing included postlaunch calibration of VIS and NIR, calculation of NDVI, and converting IR radiance to brightness temperature (BT), which was corrected for nonlinear behavior of the sensor (Rao and Chen, 1995, 1999). The three-channel algorithm routines included a complete removal of high-frequency noise from NDVI and BT values, stratification of world ecosystems, and detection of medium-to-low frequency fluctuations in vegetation condition associated with weather variations (Ko-gan, 1997). These steps were crucial in order to use AVHRR-based indices as a proxy for temporal and spatial analysis and interpretation of weather-related vegetation condition and health.

Finally, three indices characterizing moisture (VCI), thermal (TCI), and vegetation health (VT) conditions were constructed following the principle of comparing a particular year NDVI and BT with the entire range of their variation during the extreme (favorable/unfavorable) conditions. Based on the LOM, LOT, and CC, the extreme conditions were derived by calculating the maximum (max) and minimum (min) NDVI and BT values using 14-year satellite data. The maximum/minimum criteria were used to classify carrying capacity of ecosystems in response to climate and weather variations. The VCI, TCI, and VT were formulated as:

where NDVI, NDVImax, and NDVImin are the smoothed weekly NDVI, its multiyear absolute maximum and minimum, respectively; BT, BTmax, and BTmin are similar values for brightness temperature; and a is a coefficient that quantifies a share of VCI and TCI contribution to the vegetation condition. For example, if other conditions are near normal, vegetation is more sensitive to moisture during canopy formation (leaf appearance) and to temperature during flowering. Therefore, the share of moisture contribution into the total vegetation condition is higher than temperature during leaf canopy formation and lower during flowering. Because moisture and temperature contributions during a vegetation cycle are currently not known, the share of weekly VCI and TCI can be assumed to be equal.

Table 6.1 explains the algorithm development. As seen in the first row, both NDVI and BT data fluctuate considerably, primarily due to clouds, sun-sensor position, bidirectional reflectance, and random noise. In June, for example, clouds triggered considerable reduction in NDVI and BT; but such a reduction in August was smaller. The smoothing procedure (second row) eliminates outliers and emphasizes seasonal cycle. Since the smoothed NDVI was close to multiyear maximum (third row) values and BT was close to minimum (fourth row) values, VCI, TCI, and VT are >60,

TCI = [(BTmax - BT)/(BTmax - BTmm)]100 VT = a (VCI) + (1 - a)TCI

max max

Table 6.1 Coefficient used for algorithm development for various vegetation indices

Normalized difference vegetation index Brightness temperature

Table 6.1 Coefficient used for algorithm development for various vegetation indices

Normalized difference vegetation index Brightness temperature

Parameters

May

June

July

August

May

June

July

August

Raw

0.60

0.22

0.45

0.39

35.3

9.9

31.7

15.4

Smooth

0.39

0.41

0.44

0.45

27.8

28.5

29.3

26.7

Max

0.42

0.44

0.45

0.47

31.0

31.3

31.7

31.1

Min

0.31

0.32

0.34

0.35

27.0

27.5

27.8

26.4

V(T)CI

73

75

91

83

80

74

62

94

VT

77

74

76

88

indicating good vegetation health. In contrast, during drought years, these indices will be <40, indicating vegetation stress.

0 0

Post a comment