Impact of future Climate on Agriculture and Modeling the Crop Water requirement using Machine Learning approach: A Case Study in a Semi-Arid Climatic Zone of Karnataka, India
- Paper number
GLOC-2023,T,IP,x74640
- Author
Dr. Vinay Kellengere Shankarnarayan, RV Institute of Management, India
- Year
2023
- Abstract
The potential of remote sensing data in weather forecasts and agricultural practices has been widely acknowledged. However, in reality, operational applications of remote sensing and machine learning techniques in irrigation management are few. The applicability of remote sensing technologies to evaluate the potential requirement of crop water in a pilot region in India was investigated in this study. Understanding global crop yield production is key to meeting food security challenges and reducing the impacts of climate change, which indirectly helps India achieve zero hunger, one of the United Nations’ Sustainable Development Goals 2030. Accurate weather forecasting is not an easy task due to the non-stationarity nature of the weather. Traditional time series data often have relatively low dimensionality. While data volumes continue to grow, traditional statistics techniques can no longer deal with massive amounts of data. Many methods have been developed to model the evolution of meteorological simulations. Compared to parametric methods like ARIMA, nonparametric methods are more effective in prediction performance, even for stochastic weather data. We describe a performance modeling method that uses a K-Nearest Neighbor (KNN), a non-parametric version of a machine learning approach. We calculated the Normalized Difference Vegetation Index at each station using kriging. We use the CROPWAT, a Crop Water Requirement application algorithm, to estimate the crop water requirement for annual water allocation planning. This model can predict crops with a high spatial resolution months before harvest, using only globally available covariates, and help make agricultural decisions. This approach shows that it surpasses both classical statistical methods and a parametric time series model in predicting the output of years held during model training. This research helps determine crop water requirements by using Quantitative analysis and Remote Sensing at the regional level.
- Abstract document
- Manuscript document
(absent)