Location: Ergun City, Hulunbuir, Inner Mongolia, China (Photo by Russ Lee on Unsplash)

Authors: Xiaojie Wen, Philipp Mennig and Johannes Sauer

Since droughts, floods, hail, frost — extreme weather is an ever-bigger threat to farms, governments invest heavily in infrastructure to help agriculture cope: irrigation systems, reservoirs, drainage networks, soil conservation projects (Rezaei et al., 2023). But how much of this is actually needed before it makes a real difference? A new study using two decades of data from Chinese cropland suggests that risk mitigation measures don’t work gradually — many of them need to reach a certain scale before they meaningfully alleviate the damage caused by climate disasters.

Agriculture is one of the sectors most exposed to climate variability (Lesk et al., 2022; World Meteorological Organization, 2023). China illustrates the scale of the problem: between 1978 and 2018, agrometeorological disasters — droughts, floods, hail, low temperatures and frost — led to a predicted yield loss on more than 38 million hectares of the sown area. On around 22.4 million hectares of the sown area, crop yield losses of over 30 % were estimated (National Bureau of Statistics of China, 2018; Xu and Tang, 2021). Irrigation, reservoirs, soil conservation and drainage systems are widely promoted as ways to build resilience against such shocks. Yet whether — and how much — these measures actually change the relationship between climate risk and crop yields has rarely been tested rigorously.

In our recent study (Wen et al., 2025), we examine this question using data from 31 Chinese regions between 2000 and 2021. Rather than relying on a composite index, we measure agricultural vulnerability in a simple, intuitive way: the share of climate-damaged cropland that suffered severe losses (30% or more) out of all cropland affected by a disaster in the first place. Mapping agricultural vulnerability across 31 regions over 22 years reveals substantial variation across both time and space. Because hazardous weather events are inherently unpredictable, vulnerability fluctuates considerably from year to year rather than following a smooth trend. Nevertheless, vulnerability tends to be somewhat higher in the earlier years than in more recent years, suggesting a gradual improvement in China’s capacity to cope with weather-related agricultural risks. At the same time, substantial differences across regions remain, highlighting the spatial heterogeneity of agricultural vulnerability across China.

Not all regions are equally at risk

We find that the link between vulnerability and crop yields is not constant across China. In regions where agriculture plays a smaller role in the local economy, vulnerability has no statistically significant effect on yields. In regions where agriculture matters more, however, higher vulnerability translates into significantly lower yields — and the model shows these two “regimes” are persistent, with regions rarely switching back and forth from one year to the next. In short, the same climate shock is far more costly in places that depend on farming, which is exactly where risk mitigation should be prioritized first.

This raises the central question of our study: can mitigation measures actually move a region out of the “high-risk” regime — and if so, how much of them is needed? To answer this question: we test four mitigation measures — irrigation, reservoir capacity, soil loss control (e.g. afforestation and land conversion), and drainage systems — to see whether crossing a certain scale changes how strongly vulnerability drags down yields.

Risk mitigation measures do not work in the same way

Our results show that the four risk mitigation measures differ substantially in how they shape the relationship between agricultural vulnerability and crop yields.

Irrigation provides the clearest example of a tipping point. Expanding effectively irrigated area is associated with higher yields, but only up to a certain level. More importantly, when effectively irrigated area exceeds about 51% of total sown area, the negative relationship between agricultural vulnerability and yields is no longer statistically significant. Below this threshold, vulnerability is significantly associated with lower yields. Most observations in our sample remain below this threshold, suggesting considerable scope for irrigation development to strengthen the capacity of agricultural production to cope with weather-related risks.

Reservoirs tell a different story. Greater reservoir capacity substantially weakens the negative relationship between vulnerability and crop yields once a critical level is reached. In contrast to irrigation, most observations in our sample are already above this threshold. This suggests that reservoir infrastructure is already operating at a scale that helps buffer agricultural production against weather-related risks in many regions.

Soil conservation measures show a more complex pattern. Measures such as converting degraded or sloping farmland into forest or grassland do not necessarily translate into higher crop yields due to reduced cultivated land for production. We also find a stronger negative relationship between vulnerability and yields at higher levels of soil conservation. Rather than implying that conservation increases vulnerability, this pattern may reflect where these measures are implemented: more intensive conservation efforts are often needed in areas already facing greater agricultural venerability challenges.

Drainage, meanwhile, shows no statistically significant threshold effect in our analysis. Compared with the other measures, drainage infrastructure operates at a relatively limited scale in our sample. The absence of a detectable threshold should therefore not be interpreted as evidence that drainage is ineffective; rather, its effects may not yet be observable at the scale captured by our data.

Taken together, these results paint a nuanced picture of climate adaptation in agriculture: mitigation measures are not automatically effective just because they exist — many need to reach a critical mass before they change the game. For policymakers, this means two things. First, priority should go to regions where agriculture is economically important and vulnerability is high, rather than spreading investment thinly and evenly. Second, especially for irrigation, it is worth being deliberate about scale: under-investment leaves regions exposed, but pushing irrigation too far can also become counterproductive for yields directly. Designing effective adaptation policy is therefore not just about which measures to fund, but how much of them a region actually needs.

Xiaojie Wen is at the Production Economics Group, University of Bonn, and was previously at the Chair of Agricultural Production and Resource Economics, Technical University of Munich. Philipp Mennig and Johannes Sauer are at the Chair of Agricultural Production and Resource Economics, Technical University of Munich.

References

Lesk, C., Anderson, W., Rigden, A., Coast, O., Jägermeyr, J., McDermid, S., … & Konar, M. (2022). Compound heat and moisture extreme impacts on global crop yields under climate change. Nature Reviews Earth & Environment, 3(12), 872-889.

National Bureau of Statistics of China. (2018). Chinese Statistics Yearbook. China Statistics Press, Beijing.

Rezaei, E. E., Webber, H., Asseng, S., Boote, K., Durand, J. L., Ewert, F., … & MacCarthy, D. S. (2023). Climate change impacts on crop yields. Nature Reviews Earth & Environment, 4(12), 831-846.

Wen, X., Mennig, P., & Sauer, J. (2025). Assessing the regime-switching role of risk mitigation measures on agricultural vulnerability: A threshold analysis. Ecological Economics, 227, 108360.

World Meteorological Organization. (2023). WMO Statement on the State of the Global Climate in 2022. World Meteorological Organization (WMO).

Xu, X., & Tang, Q. (2021). Spatiotemporal variations in damages to cropland from agrometeorological disasters in mainland China during 1978–2018. Science of the Total Environment, 785, 147247.

Acknowledgments:

ChatGPT (OpenAI, GPT-5.6) was used for minor language improvements. The author reviewed and revised the final text and takes full responsibility for its content.