Automation of Agromet Advisory by IMD and
ICRISAT vs Bharat
Forecasting System (BFS) : Efforts to join the missing dots
The ICRISAT has launched artificial intelligence (AI) and
machine learning (ML) to deliver personalized, real-time climate advisory
services to farmers. Earlier, IMD, MoES had tried to developed e-advisory for
farmers and its automated generation utilizing weather forecast and remote
sensing inputs. But still are not in an operation mode at district level
advisories formulation. The two things one has to consider that
observed verses forecast pattern either it lies in weather or the agromet
advisory. Are observed weather has shown any pattern or in the agromet
advisories, if yes automation works precisely, if yes automation works precisely,
if not replication by automation is only refinement. If you have some patten than
AI will work better than any model.
Weather pattern could have forecast
probabilities, which are based on the number of ensemble members assigned to
each weather pattern. Robert et al. (2022) have reported that the winter dry period weather
patterns have the highest forecast skill, closely followed by retreating
monsoon weather patterns. In contrast, monsoon onset and break monsoon weather
patterns have the lowest forecast skill.
Though the bi-weekly
agromet advisory provides farmers with localized agricultural advice based on weather
forecasts, crop stage, soil conditions, pest/disease risks, and agricultural
best practices, typically issued every 4-5 days.
To have
more precise weather forecast the GoI had launched the BFS which may provide highly
accurate, real-time weather predictions at an unprecedented 6-kilometer
resolution, a significant improvement over the previous 12-kilometer system.
This allows for village-level forecasts up to 10 days in advance.
Comments are given by
many scholars that one has to consider weather parameters rather than rather
variables, use of AI in weather forecasting and agromet advisory. To define a
model or define the state parameters are more suitable than variables. The
other know dots or difficulties involves in weather forecasting are tropical weather is inherently more
chaotic, complex interactions with weather systems, data gaps and randomness. Hence, we need to connect all
these dots.
AI
and forecasting have limitations
AI cannot reliably
predict the toss of a One Day International (ODI) cricket match or any coin toss
because the outcome is random by nature. Although the
probability is 50 50 just like rain or no rain.
Where chaos and complex
interaction exits, precise quantitative forecasting may be not has 95-100
probability. Randomness in output restrict the use of AI, Hopeful for
e-advisory and AI forecast would be made automation more precise in this
decade.
