Multi stage wheat yield estimation using multiple linear, neural network and penalised regression models

Question

1. How are weather variables used for wheat yield forecast?

2. What weather data was collected for wheat crop growing period?

3. What are simple weather indices?

4. What is the performance of SMLR model for wheat yield estimation at tillering stage?

Accepted Answer

Weather variables are used to estimate the relationship between weather and wheat yield through statistical methods. Models are calibrated and validated with historical data. Techniques like multiple linear regression, discriminant function analysis, and principal component analysis are employed to develop weather indices and forecast models. For instance, Garde et al. (2015) used MLR and discriminant function analysis for wheat productivity estimation in Varanasi, India. Similarly, Agrawal and Mehta (2007) developed multiple linear regression models using weather indices. These models help in predicting wheat yield based on weather conditions, aiding in effective crop yield forecasting.

Accepted Answer

Weather data collected includes maximum temperature, minimum temperature, rainfall, morning and evening relative humidity, and sunshine hours. Data was arranged for three stages: tillering, flowering, and grain filling. Daily weather data was converted into weighted and unweighted composite weather data for analysis. 70% of data was used for model calibration, and the remaining 30% for validation. Simple and weighted weather indices were developed for each district during different stages. The computation of these indices was based on specific formulas.

Accepted Answer

Simple weather indices are numerical values that represent specific weather variables, such as temperature, rainfall, and humidity, over a given period. These indices are calculated using formulas like Xiw Zij m w = 1, where Xiw represents the value of a weather variable in a specific week, and r j iw/r j ii'w represents the correlation coefficient of yield with the weather variable. Simple weather indices help researchers analyze the impact of weather conditions on agricultural productivity and other related fields.

Accepted Answer

The performance of the SMLR model for wheat yield estimation at tillering stage is presented in Table 3. The value of nRMSE during validation ranged between 4.08 to 16.14%. The model showed excellent performance for ICAR-IARI, New Delhi with nRMSE value 4.08%, and good performance for Hisar, Ludhiana, Amritsar, and Patiala with nRMSE values of 11.67%, 14.76%, 16.14%, and 12.09% respectively. The percentage deviation of estimated yield at tillering stage compared to observed yield was lowest for Patiala (-0.12%) and followed by ICAR-IARI, New Delhi (3.09%), Hisar (7.17%), Ludhiana (14.11%), and Amritsar (19.71%).

Accepted Answer

The performance of PCA-SMLR model for wheat yield estimation at flowering stage was excellent for ICAR-IARI, Delhi with nRMSE value 4.33%, good for Hisar, Patiala, Ludhiana, and Amritsar with nRMSE values of 12.0%, 12.02%, 13.56%, and 17.06% respectively. The percentage deviation of wheat yield estimation at flowering stage by observed yield was lowest for ICAR-IARI, New Delhi (0.91%) followed by Patiala (-4.58%), Hisar (7.84%), Ludhiana (22.7%), and Amritsar (22.86%).

Accepted Answer

The ANN model's performance for wheat yield estimation at different growth stages in various districts of north India was evaluated. The coefficient of determination R2 ranged from 0.81 to 0.96, with the optimum number of hidden neurons varying between 3 and 9. The lowest RMSE during calibration was observed in ICAR-IARI, New Delhi (144.4 kg/ha), followed by Patiala (200.5 kg/ha), Hisar (216.5 kg/ha), Amritsar (314.2 kg/ha), and Ludhiana (316.9 kg/ha). The ANN model showed good performance at tillering, flowering, and grain filling stages, with nRMSE values below 20% for all districts. The lowest percentage deviation of yield estimation was observed in ICAR-IARI, New Delhi (-1.26%) at the flowering stage, and in Hisar (-11.12%) and Patiala (-16.68%) at the grain filling stage. Overall, the ANN model demonstrated promising results for wheat yield estimation in different growth stages across various districts in north India.

Accepted Answer

The range of R2 values for PCA-ANN models during calibration was between 0.83 to 0.95. This indicates a good fit of the model to the data, with values closer to 1 indicating a better fit. The R2 value represents the proportion of the variance in the dependent variable (wheat yield) that is predictable from the independent variables (PCA factors and time). Higher R2 values suggest that the model can accurately estimate wheat yield based on the given factors and growth stages.

Accepted Answer

The performance of the LASSO model for wheat yield estimation at different growth stages in various districts of north India was evaluated. The coefficient of determination (R2) during calibration ranged between 0.86 to 0.96. The root mean square error (RMSE) during validation ranged between 174.6 to 546.9 kg/ha. The LASSO model showed excellent performance for ICAR-IARI, New Delhi and Hisar with nRMSE values of 4.13 and 6.78% respectively. Good performance was observed for Ludhiana, Amritsar, and Patiala with nRMSE values of 10.34, 12.32, and 10.54% respectively. The percentage deviation of estimated yield by LASSO model at tillering stage was lowest for Patiala (-2.38%) followed by Hisar (-2.85%), ICAR-IARI, New Delhi (4.02%), Ludhiana (10.65%), and Amritsar (14.81%). At flowering stage, the lowest percentage deviation was observed for Hisar (-1.64%) followed by Patiala (-2.06%), ICAR-IARI, New Delhi (9.07%), Ludhiana (9.78%), and Amritsar (12.09%). At grain filling stage, the lowest percentage deviation was observed for Patiala (-3.06%) followed by ICAR-IARI, New Delhi (3.99%), Hisar (-6.47%), Ludhiana (6.78%), and Amritsar (10.84%). The LASSO model showed excellent performance for ICAR-IARI, New Delhi and Hisar at flowering stage with nRMSE values of 5.22 and 6.63% respectively. Good performance was observed for Ludhiana, Amritsar, and Patiala with nRMSE values of 10.45, 11.87, and 12.48% respectively. The LASSO model's performance was influenced by the Z variates, with positive influence on yield except for Z151, Z261, Z271, and Z671 in Hisar, Z360 in Ludhiana, and Z41, Z141, and Z231 in Patiala. Overall, the LASSO model demonstrated excellent to good performance in wheat yield estimation at different growth stages in various districts of north India.

Accepted Answer

The LASSO and Elastic Net models were found to be the best for multistage wheat yield estimation. These models showed higher estimation accuracy at different growth stages compared to the SMLR model. The study suggests that Elastic Net and LASSO models can be effectively used for district-level wheat yield forecasting at various crop growth stages. The research highlights the importance of using these models for accurate yield estimation, which can aid in better agricultural planning and decision-making. It is essential to note that the results are based on the evaluation of different empirical models and their performance in the context of five major wheat-growing districts in north India. The findings contribute to the understanding of the effectiveness of LASSO and Elastic Net models in predicting wheat yield, providing valuable insights for researchers and practitioners in the field of agricultural modeling and forecasting.

Multi stage wheat yield estimation using multiple linear, neural network and penalised regression models

Chat with Paper

AI Agents for this Paper

Most frequently asked questions

1. How are weather variables used for wheat yield forecast?

2. What weather data was collected for wheat crop growing period?

3. What are simple weather indices?

4. What is the performance of SMLR model for wheat yield estimation at tillering stage?

5. What is the performance of PCA-SMLR model for wheat yield estimation at flowering stage?

6. ANN model performance in north India?

7. What is the range of R2 values for PCA-ANN models during calibration?

8. LASSO model performance in wheat yield estimation at different growth stages in north India districts?

9. Which models performed best for multistage wheat yield estimation?

References

Evaluation of multiple linear, neural network and penalised regression models for prediction of rice yield based on weather parameters for west coast of India

Different approaches on pre harvest forecasting of wheat yield

District wise yield models of rice in Bihar based on water requirement and meteorological data

Related Papers (5)

ELASTICREGRESS: Stata module to perform elastic net regression, lasso regression, ridge regression

Applying "Lasso" Regression to Predict Future Glaucomatous Visual Field Progression in the Central 10 Degrees.

Prediction of compressive strength of GGBFS and Flyash-based geopolymer composite by linear regression, lasso regression, and ridge regression

Linear Regression Modelling on Epigallocatechin-3-gallate Sensor Data for Green Tea

A Comparison of the Efficiency of Using a Deep CNN Approach with Other Common Regression Methods for the Prediction of EGFR Expression in Glioblastoma Patients