|
Crop growth modelling - use of IT in agronomic research
Syed Aftab Wajid & Prof Dr Abid Hussain
Crop growth modelling, defined here as the dynamic simulation
of crop growth by numerical integration of constituent
processes with the aid of computer, is a technology used to
construct a relatively transparent surrogate (substitute)
for a real crop, one that can be analysed and manipulated
with far greater ease than the complex and cumbersome
original. The development of crop modelling, analogous to
biological life cycles, can be described as series of stages
from germination to maturity.
Pakistan is predominantly an agricultural country. In spite
of favourable conditions of soils, irrigation water and
climate, agriculture in the country suffers from
under-production both in terms of yield per hectare and
production per farm worker. The country is heavily dependant
on agriculture for food and fibre requirements of the
ever-increasing population. In order to cope with these
requirements, it is essential to increase food and fibre
production not only to attain self-sufficiency but also to
extent of exportable surplus for earning foreign exchange.
Agriculture sector of Pakistan mainly depends on the
production of field crops for arning foreign exchange
(cotton and rice) and daily diet of the people (wheat and
rice). As wheat is the major staple food, its fluctuating
production deeply affects the economy of the country.
Crop models have been used to quantify the yield gap between
actual and climatic potential yields of different field
crops. They can also be used to evaluate possible causes for
change in yield over time in a give region. Similarly models
can be used as a research tool to evaluate optimum
management for cultural practices, fertilizer use, water use
and pesticide use. Yield forecasting prior to harvest is of
interest to government agencies, commodity firms and
producers.
Finally, crop growth models can be used to evaluate
consequences of global climate change on agriculture
production, regional economies, and the like.
Conventional methods of analysis in agronomic research
usually produce results specific to the sites and seasons in
which experiments are conducted. Thus a coherent framework
of analysing the process of yield formation is lacking. The
old approaches of yield component analysis and growth
analysis have failed to provide this framework. Therefore,
the results from such studies provide few insights into the
causes of crop responses to agronomic treatments. To carry
the analysis of yield formation beyond above mentioned
descriptions, predictive models of crop growth and yield are
required.
Mechanistic computer simulation models, based on daily
weather data and initial soil conditions, can assist in
developing site-specific recommendations to optimise yield.
Furthermore, mechanistic models help in our understanding of
the complex interaction among plant growth, and
environmental conditions. The level of detail required
depends on the objectives and the data available to
construct and run the model. It is desirable to build a
model one or two hierarchical level below that required for
predictions. Crop models should incorporate process models
from the plant and organ levels. Such growth models, based
on radiation interception and radiation use efficiency have
been successfully developed for different crops.
Mathematical models have far long been used as standard
analytical tools in various branches of physical sciences
and engineering. But nowadays, with the development of
powerful computers and information sciences, it has become
possible to conveniently analyse and study even complex
real-world problems, with the help of mathematical
techniques, and the use of computational softwares. Today
the modelling approach is widely being used by planners and
policy-makers both in the advanced and developing countries,
to analyse and investigate complex socio-economic and other
developmental issues in a variety of disciplines such as
demography; agriculture, food and environment; climate
changes, etc. However, the relevant expertise is very much
lacking in Pakistan and the graduates and post-graduates
coming out of our universities are, in general, not familiar
with this important tool. It is in this background that a
Ph.D project on crop growth modelling was started by Dr.
Syed Aftab Wajid under the dynamic supervision of Prof. Dr
Abid Hussain, Department of Agronomy, University of
Agriculture, Faisalabad.
Mr Wajid collected data on growth, development, light
interception, biomass accumulation and grain yield from
field studies conducted at the Agronomic Research Area,
University of Agriculture, Faisalabad. A whole crop
mechanistic simulation model (WHEATGROW V.1) of wheat was
developed and written in Java (JDK 1.4) language. Model
simulates the growth, development and yield of wheat planted
on different sowing dates (10 November, 25 November, 10
December) under adequate husbandry management. The primary
objective of the model was to determine the potential grain
yield of wheat and to develop strategies for optimizing the
resources used under prevailing environmental conditions.
The model simulates daily canopy development, fraction of
intercepted radiation, cumulative intercepted PAR, crop
growth rate and total dry matter production.
Findings of the research revealed that observed duration
from sowing to emergence (S-E) phase varied between 6-10
days in 10 November sowing, 8-12 days in 25 November sowing
and 9-16 days in 10 December sowing in the two seasons.
Simulated duration for the S-E phase was between 7-8 days in
10 November, 9 days in 25 November and 11-15 days in 10
December sowing. Similarly, measured duration from emergence
to anthesis (E-A) phase differed between 85-89 days in 10
November sowing, 84-85 days in 25 November sowing, and 67-69
days in 10 December sowing. Simulated duration for the E-A
phase among different sowings varied between 78-79 days.
Both observed and simulated durations for E-A and anthesis
to physiological maturity (A-PM) phase showed a difference
of 1-3 days (1998-99) and 2-7 days (1999-2000) among various
sowing dates. Thermal duration (above a base temperature of
5 _C) sowing to maturity was at 1290-1362 _C days in
November sowing, 1089-1195 _C days for crops sown in
December. This could be due to dail y temperature variations
during the seasons.
Generally, wheat varieties sown in Punjab (Pakistan) belong
to spring group of wheat, and thus temperature only was used
to simulate the development of various phases. Generally
there is a linear relationship between the rate of crop
development and temperature.
Researcher further explained that, in all the sowings, there
was a tendency to underestimate LAI in the model. This
indicates the need for the development of algorithms of
different processes involved in canopy expansion,
partitioning of assimilate to leaves, specific leaf area
etc. to refine the model. The pattern of simulated LAI
values for the two seasons was quite close to measured
values during the life cycle. When values of simulated and
observed LAI development, up to maximum, were regressed, a
highly close correlation was noted and regression accounted
for variability ranging from 0.90 to 0.97 among different
sowing dates during both the seasons.
Researcher also showed the importance of the amount of
intercepted PAR in maximising the yield of wheat. Primarily,
the growth rate of a crop is a function of incident PAR, and
the proportion that is intercepted by the leaf canopy.
Interception of PAR was modelled with reasonable accuracy
for different sowings by the model. Final measured and
simulated amount of cumulative intercepted PAR was higher
(10-39 per cent) in 10 November sowing than 25 November or
10 December sowing. The cumulative intercepted PAR showed
significant correlation between simulated and observed
values among different sowing dates. The regression
accounted for variability in the data ranging from 87.5 per
cent to 99.7 per cent among different treatments.
Total biomass (TDM) production in wheat is dependant upon
its growth rate times total duration. Syed Aftab Wajid,
while talking to the correspondent, told that the model
estimated production of TDM in both November sowings
accurately, but it was underestimated in the December
sowing, especially in the later season growth. A significant
and positive correlation between simulated and observed TDM
was observed among different sowing dates in both the
seasons. The percentage variation accounted for (R2) in the
data was ranged from 84.3 per cent to 98.3 per cent among
different treatments. Model estimates of TDM were high as
simulated TDM was higher as much as 50- 62 per cent than
observed, indicating the potential of biomass productivity
under the prevailing environmental conditions.
Present research highlighted the gap between the attained vs
potential productivity of wheat under the semi-arid
conditions. Data indicated a reduction of 55.5 per cent to
60.1 per cent in 10 November 49.2 per cent to 53.7 per cent
in 25 November and 38.2 per cent to 47.2 per cent in 10
December sowing from the potential productivity of wheat
under the prevailing conditions. Thus, in this environment,
the primary determinants of crop growth and yield are
temperature and radiation. Early (10 November) sowing showed
its superiority over 25 November or 10 December sowing in
producing higher yield at 5495 kg to 8544 kg ha-1 in the two
seasons. These higher yields were primarily due to higher
amount of intercepted PAR in the former than in the latter.
The observed values may be an inadequate representation of
the large variation normally associated with the field
experiments. Nevertheless, it is clear that the model
simulates growth and yield under well defined agronomic
conditions, and can be used in variable en vironments of the
country where temperature and radiation receipts fluctuates
considerably to determine the potential crop yield. The
model can also be employed for the assessment of impact of
climatic changes on crop production characterized by
increasing temperature and Co2.
WHEATGROW (V.1) presented here is based on sound physical
and physiological processes. The model describes the growth
and potential yield of wheat mechanistically under
non-limiting conditions. However the model needs further
developments in the field of nitrogen and drought effects
under the hot semi-arid areas. Subroutines can be built into
the model for nitrogen and drought effects on growth,
development and yield of wheat. This preliminary version of
the model needs further testing and validation for a wide
range of cultivars and environments. This will certainly
lead to further refinement of the model. Detailed crop
growth and light interception data from any source other
than these experiments is not available in the country.
For this, there is urgent need for more observations on a
range of cultivars and over sites spread throughout the
Punjab province, as this can be used as a practical
management tool.
The principal objective of developing WHEATGROW was to have
management tool to determine climatically potential yields
as well as to analyse the effect of climatic variables and
management on productivity of wheat. The performance of this
model has been evaluated using a small database assembled
from the experimental studies conducted by the author. In
general, the model was able to simulate well the trends in
light interception dry matter growth and productivity. In
particular, the productivity as affected by various climatic
constraints and management treatments was satisfactorily
simulated.
In conclusion, the model shows that to achieve high grain
yield (~6 t ha-1), a wheat crop must intercept 600 Mega
Joules of photosynthetically active radiation. This would
require an LAI of 4.0 to 6.0 from mid December to early
March during a typical season of wheat in the Punjab
(Pakistan).
|
Pakissan.com;
|
