Extreme weather events, including high-intensity rainfall, heat waves, and droughts, pose significant threats to global food security. The increasing frequency of these events necessitates quantifying their impact on crop yields to develop effective risk management and adaptation strategies. While crop modeling is frequently used to assess climate change impacts, it simulates potential yields rather than actual yields and has limitations in assessing extreme event effects. Empirical methods applied directly to farmers' fields are crucial for accurate impact assessment. However, disentangling the effects of weather extremes from those of soil conditions, management practices, technological advancements, market forces, and policy changes is challenging. This requires detailed data on weather extremes, soil quality, and management decisions, which are often scarce. This study leverages a unique dataset collected by a farmer in the southern Netherlands, encompassing potato production on over 160 fields spanning 600 ha from 2015 to 2020. The dataset includes detailed information on weather, soil, and management practices, allowing for a causal analysis of the impact of extreme weather events on potato yields. The six-year period included a very wet year (2016), dry and hot years (2018 and 2020), and years with favorable growing conditions (2015 and 2017), providing a valuable opportunity to analyze the effects of extreme weather on yield variations observed both between years and within a single year.

Existing studies on climate change impacts on crop yields often rely on crop modeling, which has limitations in accurately capturing the effects of extreme weather events. Empirical studies attempting to assess the impacts of extreme weather on crop yields often struggle to separate the effects of weather from other factors such as soil and management practices. The lack of detailed data at the farm level further hinders accurate causal inference. Several studies have shown substantial year-to-year differences in crop yields, but the extent to which these differences are attributable to weather extremes remains unclear. This highlights the need for detailed observational data and rigorous causal methods to quantify the impacts of extreme weather on crop yields in farmers' fields.

This study employs a causal approach using a matching strategy to estimate the effect of extreme weather on potato yields. The dataset comprised six years (2015-2020) of data from a single farm in the southern Netherlands, encompassing detailed information on weather conditions (from a nearby meteorological station), soil quality, management practices, and yields across more than 160 fields. The methodology involved several key steps: 1. **Defining Extreme Weather Events:** The study utilized definitions of extreme weather events from the Agro Climate Calendar for potato, supplemented by a drought definition. These definitions incorporate rainfall amount, temperature, and duration to capture the intensity of each event. 2. **Propensity Score Matching:** To isolate the effect of weather, the researchers used propensity score matching. This involved creating pairs of fields across different years that were similar in terms of management and soil variables (using logistic regression with soil N, soil K, soil S, seed potato size, and irrigation). This approach helps to control for confounding factors and create a quasi-experimental setup. 3. **Estimating the Impact of Extreme Weather:** Pairwise yield differences between matched fields were then analyzed using Spearman's Rank Correlation to determine the relationship between yield differences and differences in the frequency of extreme weather events. Linear regression was then used, with high-intensity rainfall combined with sustained wet weather and drought as independent variables, to quantify the impact of these extreme weather events on yield. The R-squared value assessed the goodness of fit. 4. **Analyzing Within-Farm Variability:** The study investigated how the impact of extreme weather differed across field categories based on drought sensitivity, irrigation possibilities, and nutrient richness. For each category, the same correlation and regression analysis was performed to determine the impact of the extreme weather events. 5. **Projecting Future Frequencies:** The study used the KNMI'23 climate scenarios (downscaled from CMIP6) to project the future frequency of extreme weather events and estimate the potential impact on future potato yields. This involved calculating probabilities of extreme weather events occurring and combining them with estimated yield losses.

The analysis revealed that high-intensity rainfall combined with sustained wet weather and drought were the primary causes of yield loss in potatoes. High-intensity rainfall resulted in a 36% yield reduction, while drought led to a 13% reduction (doubling if the drought period was twice as long). Other extreme weather events, such as heat waves, had a negligible impact. The study found that 98% of the between-year variability in yield could be explained by these two major extreme weather events. The impact of drought and high-intensity rainfall varied across different field characteristics. Drought had the greatest impact on wet fields and the least impact on dry fields. Irrigation mitigated the impact of drought, but not completely. Nutrient-rich soils also showed a reduced impact from drought. High-intensity rainfall had the smallest impact on dry fields, likely due to earlier planting and irrigation. Analysis of the KNMI'23 climate scenarios projected an increase in the frequency of extreme weather events in the future, particularly under high emission scenarios. The projected increase in extreme weather events is estimated to increase the average yearly yield loss from 4% to 6.3%.

The findings demonstrate the significant negative impact of extreme weather events on potato yields in the Netherlands. The causal approach used in this study, employing propensity score matching, provides a more precise estimation of the impact of extreme weather compared to previous studies that often lack the rigorous control of confounding factors. The results underscore the critical need for adaptation strategies to mitigate the adverse effects of climate change on potato production. The findings' relevance extends to other regions with similar climates and soil conditions, suggesting that the observed impacts may be generalizable to other potato-producing areas. The study’s robust methodology in isolating the effects of weather on yield contributes significantly to the understanding of climate change's impact on agriculture. The variation in the impact of extreme weather events across fields with varying characteristics highlights the importance of targeted adaptation strategies based on specific field conditions.

This study provides a comprehensive analysis of the impact of extreme weather events on potato yields, using a robust causal methodology. High-intensity rainfall and drought are identified as major drivers of yield loss. The results emphasize the need for adaptation strategies tailored to specific field conditions, such as improved irrigation management and soil improvement to enhance water retention. Future research could explore the development of drought- and rainfall-resistant potato varieties, optimized planting schedules, and integrated data-driven and process-based models to account for both gradual changes and extreme events.

The study's primary limitation is its reliance on data from a single farm. While the farm is representative of potato production in the region, results may not be generalizable to all potato-producing areas in the Netherlands or other regions. The use of a single farm also limits the assessment of regional variations in the impacts of extreme weather. Furthermore, the study acknowledges that the effects of elevated CO2 levels, which are typically positive for potato growth, are not captured by the empirical approach employed.