Models in Irrigation andWater Management - Practices of Irrigation and On-farm Water Management - page 413

technological trend. They developed a simple linear regression equation between
the two indices.
However, because yield anomalies can vary considerably over scales of hundreds
of kilometers, national or international yield fluctuations are generally not simple
extrapolations of local or single-state values. The spatial complexities inherent in
the analysis of large-scale yield anomalies are due largely to fluctuations in mete-
orological variables over scales of hundreds to thousands kilometers. So, location
specific model is useful.
10.5.3 Methods of Formulation of Weather-Based
Prediction Model
Crop-production statistics are composed of two main components:
- the area under the crop (e.g., hectare, ha)
- the production rate (t/ha) and/or total production ( t )
The following steps may be followed to construct a regression based prediction
model (Ali and Amin, 2006 ) :
(i) Zoning of crop area
The total land area of the location under consideration is to be divided into a
number of zones considering agro-climatic regions and/or soil resource based
regions. But for all these zones, the yield and weather data may not be read-
ily available. In that case, for model development, the total land area is to be
divided into a number of crop zones keeping in mind the agro-climatic or soil
resource based zones and available crop-reporting districts.
(ii) Removing trend component
The technological trend component is resulted from the use of high yielding
cultivars, improved fertilization, management, etc. It is depended only on time.
The trend curve of yield (of the crop under consideration) is to be examined
for technological trend/advancement through time series analysis of the data.
For this purpose, the yield rate of the crop (t/ha) is plotted against the relative
year value (e.g., for 1980-1997 years data, relative year values or year ranks
are 1-18), yield being the dependent variable. The slope of the plot represents
the technological trend (t/ha/yr). The trend component is subtracted from the
original yield series.
(iii) Choice of weather variables
The agro-climatic variables are selected based on the cause and effect rela-
tionship. The principal weather elements affecting yield of crops are rainfall,
temperature (day-time and/or night temperature), light (photo period and/or
bright sunshine hour), humidity, solar radiation and wind speed. For a partic-
ular crop variety, all of the factors or elements may not contribute to yield
anomaly. To test a weather element whether it is attributable to yield, perform
regression analysis of the yield series with the corresponding weather series.
If the regression coefficient is significant at 5% level, consider this element for