IAPRS & SIS, Vol.34, Part 7, “Resource and Environmental Monitoring”, Hyderabad, India, 2002
stratification, with meteorological subdivision forming stage I
strata and rabi crop proportion in each 15x15 km sample
segment stage II strata. A sampling fraction of 10 percent
minimal segment in each sub-strata led to identification of 323
sample segments from sampling frame of 2807.
The consolidated final forecasts across the years are given in
Table 4. It can be seen that during last four seasons, RS based
forecasts have deviated from post-harvest estimates by +0.9,
+2.1, —4.4 and 40.1 Mt only. The difference of vegetation index
of comparable dates provides information about crop status vis-
à-vis last season. This type of analysis over multiple dates
during previous years has been useful in identifying crop shifts
(Rajak et al., 2002), changes in sowing dates and vigour of
crop. Similar analyses were carried out from time to time and
were communicated to Ministry of Agriculture, Government of
India, along with results. The crop assessments cover change in
extent and vigour of the crop and major observations in
previous years include (a) late sowing (Bihar 1998), (b) early
sowing (Haryana, 1999), (c) decrease in mustard area (East
Rajasthan 1999), (d) mustard wheat crop shift (2000), (e) wheat
decrease due to drought (western MP, 2001), etc.
5.0 PILOT FASAL IMPLEMENTATION - ORISSA
An activity on pilot implementation of FASAL for Orissa,
aiming at demonstration of techniques and their evaluation has
been in operation in Orissa since 1997. Prime objective is
multiple pre-harvest rice forecasts, combining results from
weather and RS inputs. Three pre-harvest forecasts (F1 through
F3) were planned, however as experience was gained, a fourth
forecast (F4) to include end-season impacts was also made.
First forecast, with
Table 3: Comparison of national wheat forecasts (NWPF) using
multi-date WiFS and weather data with Official
estimates (DES).
target date of August 1, uses only fortnightly rainfall and rainy
days information during June 1 to July 15, to make acreage
forecasts for each of 5 agro-climatic zones. The database of
blockwise rainfall and yield since 1987 is used as input. The
results for 3 seasons are given in Table 4 Dutta et al., 2001).
The seasons include drought as well as cyclone affects which
are incorporated in last forecast. Thus in two out of three years,
the desired accuracy goals are met.
MN] F2| BB] F4| DES
Year Auzi|Sepis| Oetis| Nov april ZPEY
Aa | 447] 4050 3591 3591 44181 7411
1998 [Yield -| n33| i180| 1L97| 11.70] 231
Prod -| 4so| 368] 426] 489] -1288
Ara ul tl ade a AY E 0D
1999 [Yield -| iL96| 1270| 10388| 1024| 137
Prod eer eh acd Minad d.i d0
TR I RT RY PTT TN BI 4473
2000 | Yield 75 41.6901 10.061 930] Lic
Prod I A821 ATIT 2| AL] ^ $00
Table-4: Summary of in-season multiple forecasts for rice
(kharif) in Orissa. (Acreage estimate of F4 is taken from
F3; * Drought, # Damaged by Cyclone, + Drought)
ACR(Mha) PROD(Mt) | ACR(Mha) PROD(MO
100596 [| 272600 1. —— 25.01 62.1
1996-97 | 26.003 64.979 25.89 69.35
1997-98 26.2 67.27 26.7 66.35
1998-99 | 26.603 72.876 274 70.78
70.10(Mar)
1999-00 | 26884 70.203 26.74
75.6 (Jun)
2000-01 | 24291 68.373 24.96 68.46
2001-02 | 26423 73.568 258; 22 p 7250 (APE
05, 2002)
6.0 NATIONAL RICE FORECAST USING SAR DATA
Preliminary analysis of ERS-1 SAR data for crops
discrimination in India showed good results for rice crop
identification (Patel et. al. 1995). District level rice acreage was
estimated using two and three date ERS data, early in the
growing season (Panigrahy et al., 1997). The multi-date SAR
also provides information about crop height/ stage and growing
environment (Chakraborty et al., 1997). Based on these studies,
rice acreage for the states of Assam, West Bengal and Orissa
was estimated using multi date ERS SAR data and sample
segment approach for state / district level during 1996-97
season. The rice temporal backscatter was unique for lowland
rice allowing inventory, however, due to large variation in ERS
backscatter from open water bodies induced by change in
surface roughness due to waves caused by wind, radar signature
of water bodies was found to mix with that of rice fields.
During 1997-98, RADARSAT ScanSAR Narrow (SNB) data
over West Bengal was successfully demonstrated investigated.
During next two seasons, 43 and 89 RADARSAT scenes were
analysed to estimate rice in 4 and 10 states, respectively, and
regular national-level rive inventory and production statistics
have been provided under cloudy conditions (Table 5,
Panigrahy and Parihar, 2002). District-level trend/time-series
and rainfall based agro-meteorological models have been used
for rice yield predictions. Two forecasts, first in October and
second in following January are issued.
Season Forecast Coverage Area Production
(Mha) (Mt)
2000-01 10 States 30.426 60.013
INDIA (Projected) 39.079 73.331
2001-02 7 States(@) 25.411 47.742
INDIA (Projected) 39.848 76.445
Table 5. Results for National kharif Rice forecast using multi-
date RADARSAT for acreage and yield-weather models
7.0 FUTURE
The main areas where improvements in techniques have been
initiated for better RS-based crop forecasts are as follows:
- With the availability of Advanced AWiFS from early
2003, the National Wheat Forecasting with WiFS will be
enlarged to cover multiple crops with AWIFS. The
additional crops planned to be covered include cotton,
sugarcane, mustard.
- The yield modeling effort will be directed for using crop
simulation models and a Crop Growth Monitoring System
