Drought is a complex natural hazard that affects ecological and human systems, often resulting in substantial environmental and economic losses. Recent increases in drought severity, frequency, and duration underscore the need for effective monitoring and mitigation strategies. Predicting drought impacts rather than drought conditions alone offers opportunities to support early warning systems and proactive decision-making. This study applies machine learning techniques to link drought indices with historical drought impact records (2005:2024) to generate short-term impact forecasts. By addressing key conceptual and data-driven challenges regarding temporal scale and impact quantification, the study aims to improve the predictability of drought impacts at actionable lead times. The Drought Severity and Coverage Index (DSCI) and the Evaporative Stress Index (ESI) were combined with impact data from the Drought Impact Reporter (DIR) to model and forecast weekly drought impacts. Results indicate that Fire and Relief impacts were predicted with the highest accuracy, followed by Agriculture and Water, while forecasts for Plants and Society impacts showed greater variability. County and state level forecasts for New Mexico were produced using an eXtreme Gradient Boosting (XGBoost) model that incorporated both DSCI and ESI. The model successfully generated forecasts up to eight weeks in advance using the preceding eight weeks of data for most impact categories. This work supports the development of an Ecological Drought Information Communication System (EcoDri) for New Mexico and demonstrates the potential for broader application in similar drought-prone regions. The findings can aid stakeholders, land managers, and decision-makers in developing and implementing more effective drought mitigation and adaptation strategies.
Drought is a complex natural hazard that develops gradually over time, posing serious threats to both human and natural ecosystems (Wilhite et al 2007). Growing concern surrounding drought events stems from their widespread and often severe impacts, which include loss of life, economic damage, and long-term environmental degradation. Since 1980, the United States has experienced 31 major drought events, resulting in an estimated $362 billion in damages, making drought the third costliest disaster type and accounting for approximately 13% of total disaster-related costs (NOAA NCEI, 2025). Unlike other natural hazards, drought lacks visually apparent markers, yet its consequences can be profound. Therefore, effective drought preparedness and mitigation require not only monitoring current conditions but also anticipating potential impacts (UNDRR, 2021;WMO, 2025). Predicting these impacts is essential for developing early warning systems, such as NOAA's Drought Early Warning Systems (NOAA NIDIS 2025), and for empowering resource managers, government agencies, and other stakeholders to take proactive and informed action.
Drought is conceptually defined as a prolonged precipitation deficit (Wilhite, 2000) and can manifest in multiple forms, including meteorological, hydrological, agricultural, and socioeconomic droughts, as well as recently recognized flash and ecological droughts (Crausbay et al., 2017;Wilhite & Glantz, 1985;Otkin et al., 2018). Each type affects physical and social systems differently, often producing compounding impacts such as water shortages, wildfires, crops and livestock losses, vegetation decline, and livelihood disruptions. Linking drought types to their impacts is critical for effective management but remains challenging (Bachmair et al., 2015;2016;2017;Sandholt et al., 2002;Zhang et al., 2023). Drought monitoring relies on indicators such as precipitation and temperature, summarized through indices like the Evaporative Stress Index (ESI), which describe the onset, duration, severity, and extent (WMO, GWP 2016). These indices can be associated with text-based impact reports, providing opportunities for groundtruthing and verification. However, conceptual challenges persist: First, no single index can capture all types of droughts. Second, the effectiveness of drought indices in representing specific drought conditions relies heavily on a convergence of evidence since drought lacks direct observable indicators (Van Loon et al., 2016a;2016b;Noel et al., 2020). Third, impacts often result from cascading effects. Recent studies combine multiple indices to predict impacts (Zhang et al., 2023;Noel et al., 2020;Sutanto et al., 2019), but gaps remain. This study approaches these challenges conceptually by examining how drought impacts are represented and how well existing drought indices capture the temporal and spatial dimensions of those impacts.
To address key conceptual challenges and establish functional relationships between drought indices and impacts, this analysis focuses on two main aspects: (1) how the temporal and spatial characteristics of drought impacts are numerically represented, and (2) how existing drought indices reflect those impacts. Developing meaningful relationships between drought indices and their associated impacts requires well-represented, high-quality drought impact data. However, limited and imbalanced datasets have posed significant obstacles, particularly in tackling the second and third challenges outlined earlier. Some studies have attempted to mitigate this issue by aggregating data. For example, Sutanto et al. (2019) and Bachmair et al. (2017) aggregated data across broad geographic regions and impact categories within the European Drought Impact Report Inventory (EDII). Similarly, Noel et al. (2020) and Zhang et al. (2023) aggregated U.S. Drought Impact Reporter (DIR) data from counties to states or by time and impact category at the national level. Although aggregation and cost-sensitive learning approaches offer partial solutions, they do not fully address the underlying imbalance (Zhang et al., 2023). Moreover, aggregation can obscure critical information: temporal aggregation limits fine-scale (e.g., weekly) analysis; spatial aggregation reduces insight into localized impacts and spatial dependencies; and categorical aggregation masks relationships between specific impacts and drought events. To overcome these limitations, this study employs an advanced data augmentation technique known as Synthetic Minority Over-sampling (SMOTE), available through the Imbalanced-Learn library. Data augmentation was essential in this analysis, given the goal of obtaining higher resolution predictions (i.e., weekly and county-level), which exacerbates the imbalance in drought impact datasets.
Moreover, temporal aggregation presents a challenge in capturing the rapid development of certain physical processes that lead to drought events and their associated impacts. Th
This content is AI-processed based on open access ArXiv data.
