Identification and Mapping of Irrigated Vegetation
using NDVI-Climatological Modeling
Ramesh S. Hooda and Dennis G. Dye
Global Engineering Laboratory,
Institute of Industrial Science,
The University of Tokyo,
Japan
Abstract
Distinguishing artificially managed vegetation (mainly agricultural crops and termed here as irrigated vegetation) from the natural vegetation is an important step in modeling and monitoring terrestrial primary production and carbon cycling through satellite remote sensing. Spectral indices such as the normalized difference vegetation index (NDVI) provide information about the vigor of the vegetation which in turn is influenced by climatic factors. Therefore, NDVI of the region is expected to have a relationship with the climatic parameters. NDVI values for irrigated vegetation ( mainly agricultural crops), however, would show a less strong relationship to climatic factors compared to the natural vegetation which is totally dependent on climatic conditions. This differential behavior of the of the r two types of vegetation can be exploited to identify and map the irrigated and non-irrigated vegetation zones. The irrigated vegetation pixels a re screened out as outliers in the NDVI-climatological relationship. The present paper reports on work in progress in which we are examining the possibility of identifying irrigated vegetation based upon above relationship. NOM A VHRR 10 day composite NDVI data with 8 km spatial resolution was used in combination with monthly average climatic data of the region for the study. The Indian territory having substantial mix of r irrigated and natural vegetation, was chosen as the case study area
1. Introduction
Biomass and productivity estimates from spectral data in the visible and
near-infrared wavelengths originally used highly empirical relationships to assess the important biological variables of vegetation (Emori et al 1978 ) . Recent work on this subject is more analytical and assumes that r- productivity results from photosynthesis through which a fraction of the incident solar energy that is intercepted by vegetation canopy is converted r into biomass (Prince, 1991; Ruimy 1994). This involves decomposition of productivity into several independent variables such as incoming solar radiation, fraction of radiation absorbed and conversion efficiency of absorbed radiation into dry matter (E).
A major challenge in the NPP estimation is finding representative values of E for various vegetation types as it changes with the types of vegetation and climatic conditions (Stockle and Kiniry, 1990). Estimating productivity based on vegetation indices, therefore, requires a land cover classification that distinguishes natural and irrigated vegetation (mainly agricultural crops) to account for differences in E. In this paper we examine the potential of an NDVI -climatological modeling approach to distinguish natural and irrigated vegetation.
2. Background
The normalized difference vegetation index (NDVI) is calculated from
surface reflectance in red (R) and near infra-red (NIR) regions obtained through sensors onboard many remote sensing satellites as
NIR - R
------------
NIR+R
NDVI is an effective indicator of the amount of green vegetation present in an observed landscape. Goward ~~ (1985) showed that vegetation indices such as NDVI are related to net primary production (NPP, gm-2 year-I). Kumar and Monteith (1981) showed that the fraction of photosynthetically active radiation (PAR) absorbed by the vegetation cover is related to the ratio of red reflectance (R) to near-infrared reflectance (NIR). Asrar ~ ~ (1984) subsequently related the NDVI to the fraction of PAR absorbed. These and related studies led to development of the "production efficiency model" (PEM). The simple form of the PEM is:
NPP = eå(AP AR) = eå(NDVI * PAR)
where L(APAR) is the annual sum of APAR and E is the PAR conversion efficiency (g MJ-I). To account for differences in E for different types of vegetation, first a classification among natural and managed vegetation is required.
Because climate influences the condition and growth of vegetation, the NDVI of a region may be expected to be related to the climatic variables of the region. The NDVI for artificially managed (irrigated) vegetation, however , should show a less strong relationship to the climatic factors compared to the natural vegetation which are entirely dependent on climatic conditions. This differential behavior of the of the two types of vegetation may be exploited to separate them. If NDVI is correlated with climatic variables that are known to influence plant growth, particularly temperature and rainfall, the pixels associated with irrigated vegetation can be separated as outliers. The values of these pixels then can be used in automated classification which distinguishes the irrigated vegetation from the natural vegetation.
3. Study Area
The study area consists of the Indian subcontinent and adjoining areas, which have a substantial mix of irrigated and natural vegetation. In India, .the role of agro-ecosystem should be significant, since 45 percent of the geographical area of the country is under agriculture compared to 10-11 percent globally (Dhadwa1 1994). Because the area has large deviations in climatic conditions and therefore, variety of vegetation, it provides a suitable study area in which to develop and evaluate our approach.
