RESUMEN
Drought is a common and costly natural disaster with broad social, economic, and environmental impacts. Machine learning (ML) has been widely applied in scientific research because of its outstanding performance on predictive tasks. However, for practical applications like disaster monitoring and assessment, the cost of the models failure, especially false negative predictions, might significantly affect society. Stakeholders are not satisfied with or do not "trust" the predictions from a so-called black box. The explainability of ML models becomes progressively crucial in studying drought and its impacts. In this work, we propose an explainable ML pipeline using the XGBoost model and SHAP model based on a comprehensive database of drought impacts in the U.S. The XGBoost models significantly outperformed the baseline models in predicting the occurrence of multi-dimensional drought impacts derived from the text-based Drought Impact Reporter, attaining an average F2 score of 0.883 at the national level and 0.942 at the state level. The interpretation of the models at the state scale indicates that the Standardized Precipitation Index (SPI) and Standardized Temperature Index (STI) contribute significantly to predicting multi-dimensional drought impacts. The time scalar, importance, and relationships of the SPI and STI vary depending on the types of drought impacts and locations. The patterns between the SPI variables and drought impacts indicated by the SHAP values reveal an expected relationship in which negative SPI values positively contribute to complex drought impacts. The explainability based on the SPI variables improves the trustworthiness of the XGBoost models. Overall, this study reveals promising results in accurately predicting complex drought impacts and rendering the relationships between the impacts and indicators more interpretable. This study also reveals the potential of utilizing explainable ML for the general social good to help stakeholders better understand the multi-dimensional drought impacts at the regional level and motivate appropriate responses.
RESUMEN
We used monthly precipitation and temperature data to give early warning of years with higher West Nile Virus (WNV) risk in Nebraska. We used generalized additive models with a negative binomial distribution and smoothing curves to identify combinations of extremes and timing that had the most influence, experimenting with all combinations of temperature and drought data, lagged by 12, 18, 24, 30, and 36 months. We fit models on data from 2002 through 2011, used Akaike's Information Criterion (AIC) to select the best-fitting model, and used 2012 as out-of-sample data for prediction, and repeated this process for each successive year, ending with fitting models on 2002-2017 data and using 2018 for out-of-sample prediction. We found that warm temperatures and a dry year preceded by a wet year were the strongest predictors of cases of WNV. Our models did significantly better than random chance and better than an annual persistence naïve model at predicting which counties would have cases. Exploring different scenarios, the model predicted that without drought, there would have been 26% fewer cases of WNV in Nebraska through 2018; without warm temperatures, 29% fewer; and with neither drought nor warmth, 45% fewer. This method for assessing the influence of different combinations of extremes at different time intervals is likely applicable to diseases other than West Nile, and to other annual outcome variables such as crop yield.
