JAPRS & SIS, Vol.34, Part 7, “Resource and Environmental Monitoring”, Hyderabad, India, 2002
MODELISATION OF LST-NDVI RELATION AND SURFACE CHARACTERISTICS
USING NOAA AVHRR
NR Patel*” and LM Pande?
"Indian Institute of Remote Sensing, Dehradun — 248001 (India)
KEY WORDS: NOAA AVHRR, NDVI, Land Surface Temperature, Validation
ABSTRACT:
The importance of mapping, monitoring and quantifying changes in eco-physical and agro-hydrological environments has received
greater attention by scientific community in the study of global change. Land Surface Temperature (LST) and its relation with
Normalised Difference Vegetation Index (NDVI) can be used as an indicator to quantify changes in the physical surface
characteristics on a regional scale. The present study was aimed to model temporal evolution of LST and its relation with NDVI and
ground temperatures during rabi crop growing season over semi-arid to arid Gujarat region, India.The estimated LSTs were
regressed against NDVI for assessing dynamic response of Surface Temperature / Vegetation Index for different districts in Gujarat.
The results shows that a strong dynamic negative correlation exists between land surface temperature and NDVI. The steepness of
slope of LST/NDVI relation was found less during mid growth stages (i.e. peak vegetative period) compared to early and maturity
stages of rabi crop growing season. Scatterogram of Banaskantha district over different dates depicts the narrower spread during
vegetative growth period than early and late growth stages. However, width and spread of scatterogram during peak vegetative
growth period (i.e. January) vary among districts in response to vegetation cover and hydric deficits. A good agreement was
observed between satellite retrieved surface temperature with ground estimates of mid day surface soil temperature and near surface
air temperatures. The result show that average bias in retrieved surface temperature compared with ground based surface soil
temperatures observations over three meteorological stations is approximately 2°C or less for four dates of satellite over pass. The
results again confirm that mid day near surface air temperature is easier to model in period of good vegetation cover when extreme
temperatures are not present and no important hydric deficit exists
1. INTRODUCTION
The use of satellite remote sensing to monitor changes in
physical surface characteristics governing biospheric processes
have received a greater attention by scientific community in the
study of global climate change. The derivation of large-scale
continuous fields of surface characteristics is possible by the
use of high resolution satellite imagery. Multispectral
measurements from Advanced Very High Resolution
(AVHRR) on board the operational Polar Orbiter NOAA
Satellite provide information on surface reflectivity, land cover
(including snow), state and/or amount of vegetation, surface
temperature, daily temperature range etc. A Land Surface
Temperature (LST) and its relation with SVIs (Spectral
Vegetation Indices) have often been used as an indicator to
quantify changes in the physical surface characteristics on a
regional scale.
Since seventies, several attempts have been made to use
relationship between SVIs and LST for a number of
applications, including the estimation of evapotranspiration
(Hope, 1988; Nemani and Running, 1989; Price, 1990;
Lagouarde, 1991) energy balance components (Pierce and
Congalton, 1988,Carlson et al., 1990), Surface Moisture Status
(Schmugge, 1978; Nemani et al. 1993; Goward et al. 1994) and
more recently, air temperature (Goward et al., 1994; Prihodiko
et al., 1997) and land cover classification (De Fries et al. 1995;
Lambin and Ehrlich, 1996). It is thus important to be able to
characterize and to understand sources of variation in SVI —
LST relationships at variety of scales, both temporal and
spatial. The observed linear decrease in LST with increase in
SVI has generally been explained in terms of the increase in
latest heat flux associated with greater amount of
transpirationally active vegetation. In recent times, LST and its
relation with NDVI has also been explored to estimate soil and
593
air temperatures and validated with actual ground observations
(Gupta et al. 1995; Chada et al., 2000). Moreover, few studies
looked at dynamic response of LST to vegetation cover and
other surface environment variables. Therefore, this study aims
to examine variation in the relationship between LST and
NDVI with respect to spatial and temporal properties of
observations and to analyze variation of estimated LST in
relation to actual ground temperatures.
2. STUDY AREA AND DATA SET
2.1 Study Area
The study was carried out over Gujarat State in India. The state
of Gujarat is situated between 20?01' to 24?07" north latitudes
and 68"04' to 74^04" east longitudes. The topography includes
a central high land in the northeastern part, western hills in the
southeastern part and West Coast in the central part, comprising
Gujarat plain, Kathiawar Peninsula and Kachchh Peninsula
which covers major portion of the state. The climate represents
a wide variability ranging from arid, through semi-arid, to sub-
humid tropical monsoonic type. The annual normal rainfall
varies from about 400 mm at North West end of the state to
about 2500 mm at South-East end of the state. The amount of
rain and its distribution is highly erratic over time and space. In
general, districts located in North, North-East and North-West
parts of the state suffer from drought or scarcity with a re-
occurrence interval of 3-4 years. The major land use is
agriculture (50%). Of the total cropped area, food crops like
cereals and pulses account for about 50%, while the remaining
area is under oil seed, fibre and fodder crops. The Forest area
is 10% and distributed all along the eastern border and hilly
parts of Kathiawar Peninsula.
