rence 
Phy. 
Mat. 
0.43 
(038 | 
qs | 
033 | 
0.38 
  
0.43 
  
  
  
  
  
  
  
  
  
  
. | Wax 
+ | -0.042 
2 | -0.303 
5 | -0.356 
7 | -0.481 
3 
IAPRS & SIS, Vol.34, Part 7, "Resource and Environmental Monitoring", Hyderabad, India,2002 
RELATING WHEAT SPECTRAL PROFILE PARAMETERS 
TO YIELD AND PHENOLOGY 
* 
D. R. Rajak , M. P. Oza, N. Bhagia and V. K. Dadhwal* 
Crop-Inventory and Modelling Division, 
Agricultural Resources Group, Remote Sensing Applications Area 
Space Applications Centre (ISRO), Ahmedabad-15, India. 
# Commission VII, Working Group VII/6 
Key Words: Wheat Yield Prediction, Crop Spectral Profile Modelling, Multi-date WiFS Data, Crop Phenology, Pre-anthesis Period. 
ABSTRACT 
This study reports spectral profile modelling with multi-date IRS WiFS, NDVI for wheat crop in 34 districts in the states of Haryana 
and Punjab (India), during 2000-2001 rabi (winter) season. The wheat spectral profiles were parameterised using the modified 
Gamma function suggested by Badhwar. The profiles were tested for goodness of fit and the profile parameters were found to be 
statistically significant. The profile derived peak NDVI (G,,,) was linearly related to district-wise yield and the yields for an 
independent set of 8 districts were predicted with a root mean square error of 4%. A delayed spectral emergence (T,) was observed to 
be associated with reduced pre-anthesis period matching with field observations. 
1. BACKGROUND 
Conventionally, the estimates of production of crops in India 
are based on area estimates from Timely Reporting Scheme 
(TRS) or complete enumeration by revenue agencies and yield 
estimates obtained from Crop Cutting Experiments (CCE). The 
final estimates, based on TRS or complete area enumeration as 
well as CCE are available much after the crops are harvested 
(about six months after the close of the agriculture year). 
The remote sensing (RS) technology has potential in estimating 
crop acreage and production at various spatial and temporal 
scales due to its multi-spectral, multi-temporal and synoptic 
coverage. Out of the two constituents of crop production, 
namely crop acreage and crop yield, the estimation of former 
using RS data has been demonstrated successfully. World over, 
several studies have shown the usefulness of RS data for crop 
yield estimation (Benedetti and Rossini, 1993; Doraiswami et 
al, 1995; Hochhiem and Barber, 1998; Sun, 2000). Multi- 
temporal RS observations add a new dimension to agricultural 
feature extraction and identification procedures. The temporal- 
spectral pattern of any crop summarises the phenology as well 
as growth pattern of the crop. Modelling of spectral crop 
growth profile is useful for crop identification (Odenweller and 
Johnson, 1984), feature extraction and classification (Badhwar 
et al., 1982; Badhwar, 1984), crop growth stage determination 
(Handerson. and Badhwar, 1984), crop emergence date 
determination (Badhwar, 1980) and crop yield modelling 
(Potdar, 1993; Dubey et al., 1991; Kaluburme et al., 1997). 
In India, attempts have been made to use RS data for crop 
growth monitoring and yield estimation. Sharma and 
Navalgund (1989) estimated crop growth stage using spectral 
data. Dadhwal and Ray (2000) have summarised the Indian 
experience in the field. While most of the work is based on 
empirical single date vegetation index (VI) — yield approach, 
multi-date NOAA-AVHRR derived spectral profiles have been 
used in developing yield relationships for wheat in Punjab and 
  
* 
Corresponding author (rajakdr@yahoo.com) 
363 
Haryana by Dubey et al. (1991) and Kaluburme et al. (1997), 
and for sorghum in Maharashtra by Potdar (1993). 
Wide Field Sensor (WiFS) data is very useful for regional 
monitoring due to its wide swath and enhanced revisit 
capability. WiFS was first launched on IRS-1C satellite and 
subsequently has been onboard IRS-1D and IRS-P3 also. The 
WiFS data has been used for national level crop inventory for 
wheat (Oza ef al., 1996; Oza et al., 2000) and cropping pattern 
change detection and mapping (Rajak et al., 2000; Rajak et al., 
2002). In the present study multiple acquisitions of WiFS data 
over Punjab and Haryana states in India were used for 
crop/wheat discrimination (Oza et al., 1996) and for developing 
wheat spectral profile for 2000-2001 rabi (winter) season. The 
crop spectral profiles were parameterised using Badhwar (1980) 
crop growth model on NDVI derived from WiFS data. A linear 
relationship was developed between crop spectral profile 
parameters and district-wise wheat yields for 2000-2001 rabi 
season. The relationship of profile derived time of peak NDVI — 
time of spectral emergence was evaluated as an indicator of pre- 
anthesis period. 
2. CROP SPECTRAL PROFILE 
The spectral profile of a crop can be described by the following 
functional form (Badhwar, 1980): 
Git) = G, forte Toi ios [1] 
G(0 = G, * (UT. * exp [-B (7-T,)] forts» Ty. 1... [2] 
where 
G(t) = wheat reflectivity at time t 
G, = soil reflectivity at the spectral emergence day (T,) 
o and B = crop specific constants