TL;DR
This paper uses process-based agricultural models to quantify the impact of recent large-scale US droughts (1988 and 2012) on maize production. It demonstrates that technological and management advancements over the past few decades have significantly reduced the system's sensitivity to drought, mitigating economic losses.
What Problem Does This Paper Address?
Droughts are a major threat to agricultural productivity and food security, with significant economic consequences. While historical events like the 1988 and 2012 US droughts have been costly, understanding how changes in agricultural technology and management practices over time influence drought impacts is crucial for future adaptation strategies. Previous studies often relied on statistical models or lacked the ability to disentangle the effects of climate from technological advancements. This paper aims to address this by using process-based crop models to:
- Accurately reproduce observed drought impacts on agricultural production at various scales.
- Quantify the extent to which technological and management changes have reduced system-level drought sensitivity.
- Provide a robust method for comparing the economic costs of drought events separated by decades.
Methods
The study employs the CERES-Maize model, a well-established process-based crop simulation model, integrated within the Parallel System for Integrating Impact Models and Sectors (PSIMS) framework. The model was used in three primary modes:
- Hindcasting: Simulating maize yields from 1979-2011 to evaluate the model's ability to reproduce historical impacts.
- Forecasting: Simulating 2012 US maize production before official statistics were released to assess predictive capabilities.
- Counterfactual Scenarios: Running simulations for the 1988 and 2012 droughts using swapped weather data and current (2012) technology and management practices. This allows for the isolation of the impact of technological and management changes.
Key methodological aspects include:
- Spatial Resolution: Simulations were conducted at a 10 km resolution for the conterminous US.
- Management and Technology Parameterization: Time-varying factors were incorporated, including:
- Planting Dates: Simulated using five distinct dates based on USDA crop progress data.
- Relative Maturity (RM) Group: Determined based on local historical growing degree unit accumulation.
- Planting Density: Based on state-level crop progress data.
- Genetic Yield Improvement: Linear increases in kernel number (9%) and radiation use efficiency (10%) were applied to mimic advancements in hybrid genetics.
- Land Use Change: Post-processing adjustments for the amount of cultivated corn area and the fraction of irrigated vs. rainfed land were included.
- Statistical Correction and Validation: Raw simulation outputs were aggregated to the county level and statistically corrected for biases and trends against USDA NASS survey data. A leave-one-out cross-validation approach was used to estimate forecast errors.
Datasets
The study utilized a combination of meteorological, soil, and agricultural statistics datasets:
- Meteorological Data: Daily time-series of key weather variables from January 1, 1979, to November 30, 2012, from the North American Regional Reanalysis (NARR).
- Soil Data: Soil profile parameters (textures, bulk density, organic carbon, water holding capacity) were derived from the Harmonized World Soils Database.
- Agricultural Data:
- Observed planting and maturity dates, and planting densities from USDA crop progress reports (1979-2013).
- County-level data on irrigated and rainfed harvested areas from USDA NASS (1979-2011).
- Estimates of sub-county land and management practices from the Spatial Production and Allocation Model (SPAM) dataset.
- Observed county, state, and national maize yields from USDA NASS for validation.
- USDA Risk Management Agency (RMA) Summary of Business (SOB) and Cause of Loss (COL) reports for independent estimation of crop losses.
Key Insights and Ideas
- Model Reproducibility: Process-based models like CERES-Maize, when properly parameterized and statistically corrected, can accurately reproduce historical crop yield anomalies, achieving a median absolute deviation of 6.7% nationally and 11% at the state level.
- Technological Adaptation: Significant advancements in agricultural technology and management practices (e.g., improved genetics, irrigation, earlier planting) have demonstrably reduced the US maize industry's sensitivity to large-scale drought events.
- Counterfactuals as a Tool: Using historical counterfactual scenarios (swapping weather between drought years) is a powerful method to isolate and quantify the impact of technological and management changes on drought sensitivity.
- Economic Impact Quantification: The modeling approach provides a robust framework for estimating and comparing the economic costs of drought events, even when historical data for insurance claims or direct losses are inconsistent or unavailable.
- Drought Severity vs. Impact: The 2012 drought, while climatically severe, had a less pronounced negative impact on crop productivity compared to the 1988 drought, largely due to these technological and management adaptations.
Key Results
- Yield Simulation Accuracy: The model achieved a median absolute deviation of 6.7% at the national level and 11% at the state level when simulating maize yields from 1979-2011. The 2012 simulated national average yield (7.72 t/ha) was within 0.3% of the official USDA statistic (7.75 t/ha).
- Drought Sensitivity Reduction: Changes in technology and management between 1988 and 2012 reduced system-level drought sensitivity in US maize production by approximately 25%.
- Economic Costs:
- The 1988 drought's agricultural production losses were estimated at $9 billion (adjusted for inflation to 2012 dollars).
- The 2012 drought's agricultural production losses were estimated at $21.6 billion.
- Counterfactual Cost Comparison: If farmers in 1988 had experienced the 2012 drought conditions, the losses would have been approximately $1 billion higher. Conversely, if farmers in 2012 had experienced 1988 drought conditions, losses would have been reduced.
- Spatial Impact: Drought impacts in 2012 were more widespread and generally larger at the county level than in 1988, though irrigation played a crucial role in mitigating losses in states like Nebraska.
Limitations and Future Work
- Model Limitations: The CERES-Maize model does not explicitly include dynamic functions for pests, diseases, or ozone damage. Nutrient stresses are limited to nitrogen.
- Data Limitations: While the study uses high-resolution data, limitations in climate data (1/8th degree NARR) and the need for statistical corrections highlight the challenges of capturing sub-county variability.
- Future Work:
- The authors suggest that similar approaches could be applied to developing countries, though challenges with data availability (yield, soil, climate) will need to be addressed.
- Improvements in the accuracy and fidelity of seasonal temperature and precipitation forecasts are needed to further enhance prediction, monitoring, and evaluation capabilities.
- Further research into the benefits of specific adaptation strategies (e.g., drought-tolerant varieties, efficient irrigation, cover crops) is recommended.
- The study's findings have implications for agricultural insurance and reinsurance, particularly in regions with limited historical data.
Summary updated