„

GIS APPLICATIONS IN AGRICULTRE AND SCM, RMSI

Remote sensing
Indian agriculture www.rmsi.com

RMSI has Introduction FASAL (Forecasting Agricultural output

In tropical countries like India with 127 using Space, Agrometeorological and Land
established a different agro-climatic zones, the impact of
global climate change is evidential through
based observations). Federal Agricultural
department generates the data by field
geospatial approach varied seasonal variances such as droughts
in Andhra Pradesh, Orissa, Tamil Nadu
sampling surveys. Industrial houses
send their own field team to assess the
to rationalise and flooding in places like Assam, Bihar,
Orissa, West Bengal, Uttar Pradesh, Andhra
acreage and production data. Agencies
like Agriwatch, CSE are also gathering
a competent Pradesh, etc. Coincidentally, these are also
the major agricultural states. Furthermore,
Agricultural Intelligence (AI) data from
multiple sources. However, a prudent
methodology for the dominance of middlemen increases
the extent of food insecurity. The end
examination reveals that all the above data
varies drastically.
Supply Chain result is that the government has to import
foodgrains from other countries.
Aiming to resolve such issues of vagaries
in the AI data, RMSI has established
Management which In India, we still predominantly use
traditional techniques such as field based
a geospatial approach to rationalise a
competent methodology for SCM (Supply
can combat food crop cutting experiments (CCE) to
assess the crop yield and acreage. It is
Chain Management) which can benefit
farmers, traders, exporters, industrial,
insecurity worthwhile to note that in India all crop
exports and import decisions are still based
government, and federal agencies to
combat the exports. The author, in
on historical production data (previous this paper highlights this through a
year’s production records), as against the specific case study conducted by RMSI for
growing international trend of basing estimating crop acreage estimation, crop
these decisions on more scientific and yield estimation and production estimation
accurate methods such as assessing the for various rice exporters.
current year’s yield and acreage much in
advance of the actual production by using Utility of Basmati Agricultural
remote sensing and GIS techniques. The Intelligence (AI) data
ramifications of taking crucial export and RMSI understands that AI data generated
import decisions based on historical data is is used in different ways:
that there could be a perceived shortage or 1. Industries - Use this data mainly
surplus. To cite an example, during FY-07, for procurement and supply chain
there was a bumper rubber production in management
India, as compared to previous few years. 2. Boards - Use this data to streamline
Dr K S Siva Subramanian
Assistant Vice President
Still the same was imported and the price supply chain as well as to fix the price
Agriculture and Natural Resources, RMSI of Indian rubber went down, all due to non in the market
[email protected] availability of timely data. 3. Insurance companies - Create
Yogesh Singh Agricultural data is currently generated an insurance product out of the
Project Lead - GIS, RMSI by multiple agencies in multifarious ways; agricultural yield data
[email protected]
both conventional field surveys based as well However, the conventional data does
Subrato Paul as advance information technology based. not suffice for many of the users. They
Project Manager
Agriculture and Natural Resources, RMSI
Some of the prominent agriculture data need agricultural data modeled in such as
[email protected] publishing programmes in India are: CAPE changes in the cropping pattern from the
(Crop Acreage and Production Estimation), last year, comparative analysis of the last
i4d | January 2009
30 Subrato Paul
year data vs. the current year, production pertaining to respective Rice and Bajra (Pearl
mandis, settlement packages for farmers without affecting the Millets) crops come
profit margin of the insurance companies, etc. To create such early in the market
intelligent data, RMSI followed two different aspects, namely, i.e., September–
1. Geospatial data validity of comparing Mandi data with October and Basmati
remote sensing based outputs in rice production and Sugarcane
2. Supply chain management methodology evolved from the varieties come later
above survey in November–
December months.
Study area To e n ve l o p t h i s Figure 2: Spatial Distribution of Basmati varieties
in Punjab
The study area covers major rice growing districts from the variation in time, a
Indian Ganges flood plains. This includes 13 districts each in thorough, periodic
Punjab and Haryana, 29 districts in Uttar Pradesh, 4 districts in and regular survey was carried out. Table 1 gives details of the
Uttarakhand and 2 districts of Jammu & Kashmir. Geographically, mandi data collected from a sample mandi survey in Amritsar
this spreads across 25° 83’ North to 33° 07’ North latitude and and Taran mandi in Punjab.
73° 87’East to 81° 86’ East longitude covering an area of about Details of crop varieties availability, quantity availability and
189,000 sq km. the market share in total purchase of produced crop was collected
from different mandi head offices. Agriculture crop mapping
Input data was carried out using strict scientifically programmed
The study entailed collection, procurement and analysis of primary algorithms of supervised and unsupervised classification in image
and secondary sources of information. The broad classification processing software. Training sets collected from the ground survey
of agricultural acreage over the entire region was carried out were used in this process to identify and delineate different crops.
using IRS P6 AWiFS satellite images with spatial resolution of Figure 2 shows the Basmati crop variety map for selected districts
56m. Information, from regional to local, were extracted using of Punjab. Analysis was carried out using the information collected
medium and high resolution satellite images of IRS series 1C, 1D, from remote sensing interpretations and mandi survey from
P6 LISS III and LISS IV with spatial resolution of 23.5m and various sources. Final production estimation was carried out using
5.6m, respectively. Secondary sources of information like Survey combined analysis of remote sensing outputs and mandi data.
of India Toposheet
on 1:50,000 scale, Geo-spatial data validity
district maps for In conventional ways, after the estimation, second level sample
the study area were surveys are carried out (crop cutting experiments) or remote
used as reference sensing-based ground verification is done. However, the data
maps. RMSI also produced, often does not sync with the final output data produced
collected primary from mandis or markets or final government figure. This, in
information for turn, leads to the question of authenticity of data produced from
ground truth, field remote sensing.
validation, sample RMSI undertook a hybrid approach to compare and assess
based farmer the accuracy of remote sensing-based production estimation of
survey and Mandi Basmati rice for kharif 2005 against the mandis arrivals and other
data through field related sources at the end of the Kharif season-2005. Mandi arrival
Figure 1: The study area based-surveys in data of rice and Basmati was collected for Haryana and Punjab,
all the districts. whereas in Uttar Pradesh and Uttaranchal only total rice data
Secondary information was also collected from district agriculture was available and the same was collected. The survey experts also
and state agriculture boards/offices for reference. collected data from rice mills, state marketing boards of respective
states and the Agents (Arthias). Analysis was carried out based on
Methodology information gathered from various sources like discussions with
Major components of compiling a methodology for SCM included respective mandi officials, mandi agents (Arthias), and rice millers
Market survey and assessment and Agriculture and Land Resource to arrive at the conclusion. Based on comparative analysis between
Mapping. While market survey is done through direct field Kharif 2005 estimates derived using remote sensing approach
authenticated data, agriculture and land use map was envisaged and Mandi arrivals (as well as allied sources) as on 31st March
through interpretation of remote sensing satellite images. 2006 and considering certain calculated and logical assumptions
With a cursory look at mandi information, it is apparent that and limitations, it is concluded that the remote sensing based
both quantity and expected time of arrival information are vital. estimated results for Kharif 2005 is matching up to an accuracy
RMSI selected a sample district mandi in each of the Basmati of 90% to 94% in the states of the study area. As a by-product,
growing states. Mandi arrival data (quantity) and time of arrival was an interesting result on RS data being always higher than total
collected from important mandis and market board at block/tehsil supply chain sources confirms the reliability of this data. Table 2
level and mandi board at district and state level. It was observed that gives the results so produced and the comparison analysis.
January 2009 | www.i4donline.net
31
Discussion Name of Basmati Sources of Rice Arrived in
In recent years, remote sensing has been accepted as an indispensable tool in the Mandi Varieties Material Mandi (in %)
the field of agriculture. It helps in crop acreage and yield estimation; health (District)
monitoring, agriculture information system, supply chain management,
Basmati 386 60-70%
etc. by trimming the cost and human effort. It is increasingly being used
to take vital decisions for crop marketing and export, crop inventory and Basmati 370 60-70%
commodity trading. The major activities in SCM that are answered through HBC 19 Amritsar 20-30%
this RS & GIS approach are: Sharbati 50-60%
1. Satellite based crop mapping and acreage estimation at district/tehsil Pusa 1121
level 60-65%
Super
2. Crop health monitoring using temporal satellite images and derived 50-60%
Basmati
vegetation indices
Basmati 386 30-35%*
3. Satellite derived indices and weather parameter based yield estimation
4. Comparative analysis on remote sensing based production and actual PB 1 15-20%*
Amritsar Patiala
arrival in the market & Tarn Super 15-20%*
5. Spatial database creation of agri-market location and proximity analysis Taran Basmati
20-25%*
from the produce Mandi, Sharbati
6. Vehicle routing from farm to retail outlet using network analyses Punjab
Basmati 386 30-35%*
7. Decision support system for supply-chain design and management Hoshiarpur
Pusa 1121 10-15%*
It is evident from here that AI data produced using geospatial technology
is authentic. However, the data in raw format may not yield results. Hence a Basmati 386 Ludhiana 25-30%*
methodology is discussed to use the above data for supply chain management. Sharbati 20-25%*
Gurdaspur
This approach uses a statistical model developed to use the AI data for making PB 1 15-20%*
price forecast of different Basmati varieties, a robust statistical model was PB 1 Kaithal 15-20%*
used as given in the following equation:
PB 1 Sirsa 10-15%*
Equation 1 Basmati 386 Saharanpur 20-25%*
PB1 8-10%*
Price = {ƒ(ma, ta, cd, pv)}/ƒprice
* Depend on Mandi rate
Where,
ma – Mandi Arrival Table-1: Gives details of mandi data collected from a sample mandi
survey in Amritsar and Taran mandi in Punjab.

Comparison of Kharif 2005-06 Satellite-based estimated production and Mandi arrival

for Punjab, Haryana, UP and Uttaranchal
ta – Time of Arrival
Description Punjab Haryana Uttar Uttaranchal cd – Competitor details
Pradesh
pv – Production of variety
Satellite Based-Estimated 100.00 100.00 100.00 100.00 ƒprice = ƒ(last year low, last year high, Month, Mandi
Production of Basmati 2005 [a]
arrival)
Total Basmati Arrival in Mandis 49.16 63.36 68.97 65.45
[b] Conclusion
Personal Use and Seed Purpose Conventional methods of generating data such as
15.00 10.00 10.00 10.00
[c] undertaking extensive surveys are time consuming and
Contract 6.67 1.00 2.00 2.00 expensive. At times the government takes important
Farming decisions on importing wheat, rubber, sugar, etc with
Sold through Millers 6.67 2.00 2.00 2.00 insufficient and pseudo-geospatial data. Similarly,
other channels traders rely on the information provided by the
d] Agents 6.67 7.00 1.00 1.00 midd lemen who in turn get this information from
Unorganised 10.00 10.00 the farmers.
10.00 10.00
Sector This methodology clearly depicts the advantage of
Sub-Total Geospatial based supply chain management techniques.
45.01 30.00 25.00 25.00
[e] = [c +d] However, the study has a few limitations such as paucity
of secondary data and cloud free satellite data especially
Total Basmati Accounted by
Satellite Based Study 94.17 93.36 93.97 90.45 during Kharif season. However, if the industrialists and
[f ] =[b +e] federal government jointly produce such data, future
generations will have easy utility of this supply chain
Table-2: Gives the results so produced and comparison analysis management model. „
i4d | January 2009
32