Impact of VTCI Features on Yield Predictions: A Shapley Value Analysis for 24 Counties

The provided graph illustrates the average Shapley values for 24 counties over an 11-year period, ranging from 2011 to 2021. The x-axis represents the years, while the left y-axis represents the average Shapley values for each year. The graph includes four distinct features, each represented by a separate line. Additionally, the right y-axis represents the predicted average yield for each year.

To analyze the graph, we will focus on three specific years: 2013, 2018, and 2021. These years represent low, medium, and high yield years, respectively, and will provide valuable insights into the observed trends and patterns in Shapley values and yield predictions.

In 2013, all four features exhibited relatively low Shapley values. The feature 'VTCI at the green-up stage' had a negative Shapley value, indicating a negative impact on yield. Conversely, the features 'VTCI at the jointing stage,' 'VTCI at the heading-filling stage,' and 'VTCI at the milk maturity stage' had lower positive Shapley values, suggesting a weaker positive impact on yield compared to other years. The yield prediction for 2013 was also relatively low, indicating a poor harvest.

Moving on to 2018, the Shapley values for all four features were higher compared to 2013. The feature 'VTCI at the green-up stage' had a significantly negative Shapley value, suggesting a strong negative impact on yield. The features 'VTCI at the jointing stage' and 'VTCI at the milk maturity stage' had positive Shapley values, indicating a positive contribution to yield. However, the feature 'VTCI at the heading-filling stage' had a slightly negative Shapley value. The yield prediction for 2018 was average, indicating a moderate harvest.

In 2021, the Shapley values for all four features were relatively high. The feature 'VTCI at the green-up stage' had the highest positive Shapley value, indicating a significant positive impact on yield. The features 'VTCI at the jointing stage' and 'VTCI at the milk maturity stage' also had positive Shapley values, suggesting a positive contribution to yield. The feature 'VTCI at the heading-filling stage' had a slightly negative Shapley value. The yield prediction for 2021 was the highest among all the years, indicating a bountiful harvest.

Overall, the analysis of the graph reveals that the different features represented by the Shapley values have varying impacts on yield. The feature 'VTCI at the green-up stage' consistently had a significant impact on yield, with its Shapley value being either strongly negative or strongly positive. The other features, 'VTCI at the jointing stage,' 'VTCI at the heading-filling stage,' and 'VTCI at the milk maturity stage,' had relatively weaker impacts on yield.

Analyzing the low yield year 2013, medium yield year 2018, and high yield year 2021 provides insights into the variations in Shapley values and yield predictions. It is evident that the fluctuation in Shapley values for different features directly influences yield predictions. This information is valuable for agricultural planning and decision-making as it allows for the identification of critical factors contributing to high or low yields in different years.

In conclusion, the dual y-axis graph visually represents the average Shapley values for different features and the corresponding yield predictions for 24 counties over an 11-year period. By analyzing specific years such as 2013, 2018, and 2021, we can gain insights into the impact of different features on yield and observe variations in yield predictions. This information is crucial for understanding the factors affecting agricultural productivity and making informed decisions for future crop management.