Land-atmosphere feedbacks contribute to crop failure in global rainfed breadbaskets

Ensuring global food security is a major challenge, particularly given the increasing global population and the vulnerability of agriculture to climate variability. Dry and hot spells frequently lead to crop failures, a problem potentially exacerbated by climate change. A significant portion of global crop yield variability (approximately one-third) is attributed to weather and climate fluctuations. While irrigation can mitigate some of these effects, it's not a sustainable solution for all regions due to freshwater resource constraints. Rainfed agriculture, responsible for a substantial portion of global croplands and food production, remains a vital, sustainable approach. Understanding the climate vulnerability of global rainfed breadbaskets is critical for developing effective adaptation strategies. Heat stress can directly damage plants and contribute to soil moisture deficits and atmospheric aridity. Land-atmosphere feedbacks amplify these effects: as soils dry, temperatures increase, exacerbating water deficits and heat stress, ultimately reducing crop yields. These local feedbacks are influenced by large-scale synoptic systems and climate oscillations (like ENSO and NAO), which can lead to widespread synchronous crop losses. Large-scale atmospheric patterns affect heat and moisture flow, leading to anomalies in breadbasket regions, further modified by local land-atmosphere feedbacks. These feedbacks can propagate dry and hot anomalies downwind, impacting agricultural productivity across broader areas. While the influence of local land-atmosphere feedbacks on crops has been studied, the role of upwind climate conditions remains less understood. This study investigates 75 global rainfed breadbaskets (maize, wheat, soybean, and rice) to analyze crop yield variability, focusing on low-yield years and their relationship to local and upwind climate anomalies. A novel atmospheric Lagrangian modeling framework, combined with satellite observations, is used to trace the sources of heat and moisture during low-yield years and disentangle the contributions of local and upwind land-atmosphere interactions.

The existing literature highlights the sensitivity of crop growth to climate variations, with a substantial portion of interannual yield variability attributed to weather and climate fluctuations. Studies have shown the significant impact of extreme heat and drought on maize, wheat, and soybean yields in various regions, including the United States and Europe. The role of local land-atmosphere feedbacks in exacerbating these impacts has been increasingly recognized. However, there is a knowledge gap in understanding the dependency of agricultural yields on upwind climate conditions and the combined effects of both local and upwind land-atmosphere feedbacks on crop productivity deficits. This study aims to address this gap by analyzing the influence of both local and upwind land-atmosphere interactions on crop yields in global breadbaskets.

This study identified 75 global rainfed breadbaskets for four major crops (maize, wheat, soybean, and rice) based on criteria including minimum area, rainfed fraction, and consistent sowing/harvest months. Crop yield data was obtained from the Global Gridded Crop Yield dataset (GDHY), covering 1981-2016. Yield data was detrended using LOWESS to remove long-term trends associated with technological improvements, resulting in relative yield anomalies. Crop failure events were defined as years with yields below the 25th percentile (1983-2015). Local climate anomalies (precipitation and temperature) were analyzed using datasets such as MSWEP, ERA-5, and GLEAM, considering an extended growing season to account for soil moisture memory effects. The aridity index (Ep/P) was used to differentiate between water-limited and energy-limited breadbaskets. To determine the origins of atmospheric moisture and heat, a novel Heat And MoiSture Tracking (HAMSTER) framework was employed. This framework, based on the Lagrangian particle dispersion model FLEXPART, tracked two million air parcels globally, tracing their backward trajectories for up to 15 days to identify source regions of moisture and heat (ocean, upwind land, and local land). The resulting source-receptor relationships were bias-corrected using precipitation from MSWEP, evaporation from GLEAM and OAFlux, and sensible heat from ERA-Interim. The contributions from these three sources (ocean, upwind land, and local land) were analyzed separately for each breadbasket during crop failure events. Finally, yield anomalies were mapped as a function of heat and moisture imports from the three sources to assess the relative importance of each contribution during crop failure events. The Mann-Whitney U test was used to assess the statistical significance of differences between water- and energy-limited regions and between events with and without concurrent land-atmosphere feedbacks.

The study revealed significant spatial variation in average crop yield deficits during failure events, with wheat showing the largest average loss (~15%) and rice the lowest (~7%). Water-limited breadbaskets generally experienced higher losses than energy-limited ones. Precipitation deficits (low P) and high temperatures (high T) were frequently associated with crop failure, particularly in water-limited regions. However, some energy-limited regions experienced higher-than-usual P during low-yield years, possibly due to waterlogging or reduced radiation. Analysis of moisture and heat sources highlighted the importance of upwind ocean and land regions. During crop failure events, moisture imports from nearby oceans were often below normal, while heat imports were often above normal. These anomalies were particularly pronounced within the breadbaskets themselves and in nearby land regions. In most breadbaskets (especially water-limited ones), the trend was clear: lower-than-usual moisture and higher-than-usual heat contributions during crop failure events, with larger anomalies closer to the breadbaskets. Analysis of air parcel trajectories revealed an intensification of negative moisture and positive heat anomalies as air parcels traveled from the ocean towards the breadbaskets, indicating ‘en route feedbacks’. These feedbacks often started over remote land regions, suggesting a role for upwind land-atmosphere feedbacks. The study further examined the combined effects of ocean, upwind land, and local land contributions on yield during crop failure events. For maize, wheat, and soybean, most events were associated with moisture deficits originating in the ocean, particularly significant for water-limited breadbaskets. Upwind land regions tended to contribute less moisture and more heat during crop failure events, especially in water-limited regions. Local land-atmosphere feedbacks further aggravated the dry and hot conditions within the breadbaskets themselves. Finally, the study compared yield deficits with and without concurrent local and upwind land-atmosphere feedbacks (CLF). For all crops, yield deficits were significantly higher when both upwind and local land regions contributed to lower moisture and higher heat, particularly in water-limited breadbaskets. On average, crop failures were 42% larger when both upwind and local land-atmosphere feedbacks caused anomalously low moisture transport and high heat transport.

This study's findings highlight the previously understudied role of upwind land-atmosphere feedbacks in influencing crop yields. The significant amplification of yield deficits when both upwind and local feedbacks concur demonstrates the interconnectedness of climate conditions across larger spatial scales. This is particularly crucial in water-limited regions, where crop productivity is strongly dependent on moisture supply from upwind areas. The results have implications for improving seasonal crop forecasts and developing early-warning systems, particularly by incorporating data on upwind conditions. The study also implies the importance of upwind land management practices in mitigating crop failure risks. Deforestation upwind, for example, may reduce moisture recycling and decrease precipitation downwind.

This research demonstrates a clear link between upwind and local land-atmosphere feedbacks and the severity of crop failures in global rainfed breadbaskets. The findings highlight the critical importance of considering broader spatial scales and interactions in understanding and mitigating agricultural risks in a changing climate. Future research should focus on improving the resolution of data and models, expanding the analysis to other crops and regions, and integrating these findings into climate adaptation strategies and land management practices.

The study's results are conditioned on uncertainties inherent in the evaluation of Lagrangian trajectories and the bias-correction methods used. The sparsity of ground observations poses a challenge in validating and bias-correcting the sources of precipitation and fluxes, and the statistical relationships found are not necessarily causal, despite being consistent with physical interpretations. Additionally, the study assumes static breadbasket extents and does not explicitly account for inter-annual technological changes or land use management impacts on crop yields.