International Journal of Applied and

Natural Sciences (IJANS)

WEATHER BASED WHEAT YIELD FORECASTING IN EASTERN ZONE OF

Megha Goyal & Urmil Verma

ABSTRACT

Forecasting of crop production is needed by the national and state governments for various policy decisions
relating to storage, distribution, pricing, marketing, import-export etc. In this paper, a methodology for the
estimation of wheat yield, ahead of harvest time is developed, specifically for wheat growing districts of
eastern zone of Haryana (India). The eastern zone comprises Rohtak, Karnal, Jind, Sonipat, Panipat and Kaithal
districts. Zonal yield models, using time trend and weather predictors are generated using principal component analyses.
The estimated yield(s) from the selected models indicated good agreement with State Department of Agriculture (DOA) of
wheat yields, by showing 2-12 percent average absolute deviations.

KEYWORDS: Eigen Values, Principal Component Scores, Percent Deviation and Weather Parameters

Article History
Received: 14 May 2018 | Revised: 23 May 2018 | Accepted: 29 May 2018

INTRODUCTION

India has a well-established system for collecting agricultural statistics. In fact, as early as 1884, wheat
production estimation was done for the first time in India. The primary responsibility for collection of data regarding the
area under different crops and production of crops is that of the State Governments. The Ministry of Agriculture of the
Central Government has the responsibility of compilation of area and production of major crops at national level.
Area assessment is based on complete enumeration by revenue agencies. Yield estimates obtained through analysis of
scientifically designed crop-cutting experiments (Sukhatme and Panse, 1951) are routinely issued by the Directorate of
Economics and Statistics. However, the final estimates are given a few months after the actual harvest of the crop.
Thus, one of the limitations of the conventional method used is timeliness and quality of the statistics. This suggests that
improvement of these statistics for planning and management purposes is urgently needed. Thus, our study aims at
developing such an improved methodology for crop yield forecasting.

Various organizations in India and abroad are engaged in developing methodology for pre-harvest forecasting of
crop yield, using different approaches. The most commonly used models are based on empirical statistical models.
The main factors affecting the crop yield are agricultural input and weather parameters. Use of these factors forms a broad
category of models for forecasting crop yield. Hoogenboom (2000), Kandiannan et al. (2002), Bazgeer et al. (2007),
Esfandiary et al. (2009), Lobell and Burke (2010), Basso et al. (2012) etc. have used a series of weather predictors for crop

www.iaset.us [email protected]
20 Megha Goyal & Urmil Verma

yield forecasting. Goyal and Verma (2015) have used agromet/spectral indices in context of pre-harvest yield forecasting
of different crops in Haryana.

Crop Status and Study Region

Wheat is one of the most important cereal crops in India, as it forms a major constituent of the staple diet of a
large part of the population. India occupies second place in terms of acreage and production among wheat
growing countries of the world (Source:www.mapsofindia.com/indiaagriculture/). Haryana occupies third
place for wheat production (Source: www.agricoop.nic.in/statistics) among the various states in India.
Rohtak, Karnal, Jind, Sonipat, Panipat and Kaithal districts falling under the eastern zone of Haryana have
been considered for the model building.

The Department of Agriculture (DOA)’s wheat yield of the districts under consideration from 1980-81 to
2013-14 were used for time-trend analysis, and then computing the trend based yield i.e. Tr = a + bt, where Tr = Trend
yield, a = Intercept, b = Slope and t = Year. Weather data starting from 1st fortnight of November to 1 month before harvest
has been utilized for the model building (crop growth period: 1st November to 15th April). The average weather value
calculated over 1st to 15th November gives 1st fortnight weather parameter(s), average value calculated over 16th to 30th
November provides 2nd fortnight weather parameter(s) and proceeding in the next year, the average weather value
calculated over 16th to 31st March and 1st to 15th April gives the 10th and 11th fortnight weather. The fortnightly
parameter(s) form 1 st November to 1 month before harvest during the crop growth period viz., Tmx1, Tmx2, ...,Tmx9,
Tmn1, Tmn2,...,Tmn9 and Arf1, Arf2, ..., Arf9 have been used to develop the zonal yield models.

RESULTS AND DISCUSSIONS

Principal component method was used for the extraction of factors, which consists of finding the
eigen values and eigen vectors . Principal components P i (i=1, 2…) were obtained as P = kX , where P and X
are the column vectors of transformed and the original variables respectively, and k is the matrix with rows
as the characteristic vectors of the correlation matrix R. The variance of P i is the i th characteristic root λ i of
the correlation matrix R; λ s were obtained by solving the equation |R- λI| =0.

Zonal wheat yield models based on weather data of 1980-81 to 2009-10 were developed within the framework of
principal component analyses. The validity of the models was checked for the post sample period 2010-11 to 2013-14 for
eastern zone of Haryana. Under PC analysis, first ten PCs (Table 1) of correlation matrix of weather variables
explained about 82% of the variability, and thus the remaining components accounted for a smaller amount of total
variation, hence those components were not considered to be of much practical significance.

Table 1: Eigen Values and Variance (%) Explained by Different Principal Components
Components 1 2 3 4 5 6 7 8 9 10
Eigen value
5.10 3.74 3.02 2.81 2.07 1.89 1.48 1.12 1.06 0.90
(% variance
(18.90) (13.85) (11.19) (10.39) (07.67) (07.02) (05.48) (04.15) (03.92) (03.32)
explained)

Zonal wheat yield models were fitted by taking weather (PC) scores and trend yield as regressors and DOA crop
yield as regress and. The best subsets of predictors were selected using the stepwise regression (Draper and Smith, 2003)
method, in which all variables were first included in the model and eliminated one at a time, with decisions at any

Impact Factor (JCC): 5.0273 NAAS Rating 3.73

particular step, conditioned by the result of the previous step. The selected zonal yield model along with R2 and standard
error is given below:

Yield est = 0.31 + 0.99 *Tr - 0.73 * PC 1 + 0.81 * PC 2 -0.51* PC 7

R 2 = 0.911, adj.R 2 =0.909 & SE= 2.14

Where Yield est - Model predicted yield

Tr - Linear time trend based yield

PCi - ith principal component score (i = 1,2,…,10)

R2 - Coefficient of determination

SE - Standard error of yield estimate

The model based yield(s) showing a good agreement with the real time data along with percent relative deviations
are given in Table 2.

Table 2: District-Specific Wheat Yield Forecast along with Percent Deviations from
Real-Time Yield(s) Using the Fitted Models
Rohtak Karnal Jind
Districts/
Obs. Fitted Obs. Fitted Obs. Fitted
Years
Yield Yield RD(%) Yield Yield RD(%) Yield Yield RD(%)
(q/ha) (q/ha) (q/ha) (q/ha) (q/ha) (q/ha)
2010-11 45.52 43.28 4.92 44.47 49.36 -10.99 45.45 48.26 -6.19
2011-12 50.2 43.08 14.18 56.7 49.24 13.15 52.35 48.23 7.88
2012-13 37.38 44.38 -18.72 46.7 50.62 -8.39 42.7 49.69 -16.36
2013-14 39.24 44.52 -13.44 49.12 50.84 -3.51 45.49 49.99 -9.89
Av. abs. Dev 12.81 9.01 10.08

Table 3
Sonipat Panipat Kaithal
Obs. Fitted Obs. Fitted Obs. Fitted
Districts/Years
Yield Yield RD(%) Yield Yield RD(%) Yield Yield RD(%)
(q/ha) (q/ha) (q/ha) (q/ha) (q/ha) (q/ha)
2010-11 46.43 47.99 -3.35 40.88 43.21 -5.70 47.2 47.36 -0.34
2011-12 55.21 47.94 13.17 39.17 43.48 -11.00 54.51 47.06 13.66
2012-13 45.2 49.40 -9.29 42.47 45.24 -6.53 46.84 48.26 -3.04
2013-14 46.5 49.69 -6.86 42.83 45.85 -7.06 48.15 48.31 -0.33
Av. abs. Dev 8.17 7.57 4.34

Percent Relative Deviation (RD%) = 100 × [(observed (obs.) yield – fitted yield) / observed yield]

REFERENCES

1. Basso, B., C. Fiorentino, D. Cammarano, G. Cafiero and J. Dardanelli (2012). Analysis of rainfall distribution on
spatial and temporal patterns of wheat yield in Mediterranean environment. European Journal of Agronomy,
41, 52– 65.

2. Bazgeer, S., Gh. Kamali and A. Mortazavi (2007). Wheat yield prediction through agro meteorological indices for
Hamedan, Iran. Biaban, 12, 33-38.

www.iaset.us [email protected]
22 Megha Goyal & Urmil Verma

3. Draper, N.R. and Smith, H. 2003. Applied Regression Analysis. 3rd edition, John Wiley & Sons. New York.

4. Esfandiary, F., G. Aghaie and A. D. Mehr (2009). Wheat yield prediction through agro meteorological indices for
Ardebil district. World Academy of Science: Engineering and Technology, 49, 32-35.

5. Goyal, M. and Verma, U. 2015. Development of weather-spectral models for pre-harvest wheat yield prediction
on agro-climatic zone basis in Haryana, International J. of Agricultural and Statistical Sciences, 11(1): 73-79.

6. Hoogenboom, G. (2000). Contribution of agro meteorology to the simulation of crop production and its
applications. Agricultural and Forest Meteorology, 103, 137-157.

7. Najah Ali Abidalkarem, Effect of Farm Possession on the Economic Efficiency of Wheat Yield Production in the
Irrigated Regions of Iraq (Al-Azezia, Khan Bne-Saad), Case - Study, International Journal of Humanities and
Social Sciences (IJHSS), Volume 6, Issue 3, April-May 2017, pp. 47-54

8. Kandiannan, K., K. K. Chandaragiri, N. Sankaran, T. N. Balasubramanian and C. Kailasam (2002).

9. Lobell, D. B and M. Burke (2010). On the use of statistical models to predict crop yield responses to climate
change. Agricultural and Forest Meteorology, 150, 1443–1452.

10. Sukhatme, P.V. and Panse, V.G. (1951). Crop surveys in India-II. J. Ind. Soc. Ag. Statistics, 2, 95-168.

Impact Factor (JCC): 5.0273 NAAS Rating 3.73