Comprehensive Analysis of Grain Production Based on Three-Stage Super-SBM DEA and Machine Learning in Hexi Corridor, China

Zhengxiao Yan, Wei Zhou, Yuyi Wang, Xi ChenSustainability 2022 · DOI

TL;DR

This study uses a novel three-stage DEA model combined with machine learning to analyze grain production efficiency in China's Hexi Corridor. It reveals that human-driven factors, particularly increased inputs like land and labor, are overwhelmingly responsible for production gains, while efficiency improvements and natural factors play a minor role.

What Problem Does This Paper Address?

Global food security is a persistent challenge, exacerbated by declining agricultural investment and the impacts of climate change. While grain production has increased in many regions, understanding the true drivers of this growth – whether it's genuine efficiency gains or simply increased resource input – is crucial for sustainable agricultural policy. This paper specifically investigates the Hexi Corridor, a key grain-producing region in China, to disentangle the contributions of human inputs, natural factors, and production efficiency to its grain output.

Methods

The researchers employed a sophisticated three-stage approach:

  1. Stage 1: Three-Stage Super-SBM DEA: This stage uses a modified Data Envelopment Analysis (DEA) model to measure agricultural production efficiency. The "Super-SBM" aspect allows for ranking efficient units, and the "three-stage" design is key to isolating the true drivers of efficiency.
  2. Stage 2: Stochastic Frontier Analysis (SFA): Here, the study uses SFA to remove the influence of external environmental factors and random errors from the efficiency scores calculated in Stage 1. This is achieved by modeling the "slacks" (inefficiencies) from Stage 1 as dependent variables and environmental factors as independent variables.
  3. Stage 3: Adjusted DEA: The adjusted input data, now free from environmental and stochastic influences, is fed back into the Super-SBM DEA model to obtain a more accurate measure of efficiency, solely attributable to managerial decisions.
  4. Extra-Trees Algorithm: To determine the relative importance of various factors (inputs, efficiency, and natural variables) on grain production, the researchers utilized the Extra-Trees machine learning algorithm. This algorithm is known for its ability to handle non-linear relationships and provide feature importance rankings.

Datasets

The study utilized data from the Hexi Corridor region of Gansu Province, China, spanning the years 2002–2018. The data sources included:

  • Gansu Development Yearbook (for agricultural machinery, labor, fertilizer, and planting area).
  • Gansu Rural Yearbook (for pesticide and plastic film consumption).
  • Gansu Water Resources Bulletin (for irrigation water consumption).
  • China Meteorological Science Data Sharing Service Network (for precipitation, temperature, and sunshine duration).
  • Published literature for CO2 emissions data.

The study did not introduce new benchmark datasets but rather synthesized existing public data for its analysis.

Key Insights and Ideas

  • Human Input Dominance: The most striking finding is that human-driven factors are overwhelmingly responsible for the increase in grain production, accounting for 93.7% of the influence. This suggests that simply increasing inputs like land, labor, and fertilizer is the primary strategy for boosting output, rather than improving efficiency.
  • Efficiency's Diminishing Role: Despite significant increases in grain production, the actual production efficiency (GPE) has either stagnated or declined over the study period, especially after accounting for environmental factors. This highlights a potential disconnect between output growth and sustainable, efficient practices.
  • Environmental Factors' Nuanced Impact: While environmental factors were removed to isolate managerial efficiency, their initial influence on efficiency scores was significant, particularly for economically developed regions. This suggests that policy interventions need to consider regional environmental contexts.
  • The Power of Machine Learning: The Extra-Trees algorithm proved highly effective (R² = 0.989) in predicting grain production and quantifying the importance of various factors, demonstrating its utility in complex agricultural analysis.

Key Results

  • Human-Driven Factors: Grain Crops Planting Area (37.42%), Agricultural Labor Force (27.69%), and Chemical Fertilizer Consumption (11.63%) were the most important factors driving grain production.
  • Nature-Driven Factors: Annual Precipitation (4.06%) and Average Sunshine Duration (1.76%) were the most influential natural factors, but their overall contribution was minimal (6.3% combined).
  • Grain Production Efficiency (GPE): GPE accounted for a mere 0.14% of the factors influencing grain production, underscoring its limited role in recent production increases.
  • Efficiency Trends: After accounting for environmental factors, the average grain production efficiency in the Hexi Corridor showed a slight downward trend from 2002 to 2018.
  • Regional Differences: Significant spatial variations in GPE were observed, with Jiuquan and Jiayuguan showing particularly low efficiency by 2018 compared to other cities in the Hexi Corridor.

Limitations and Future Work

  • Data Granularity: The study acknowledges that the yearbooks used for data collection are typically published a year after the actual data, introducing a slight temporal lag.
  • Model Simplification: While the Extra-Trees model is powerful, the paper notes that the GPE was almost ignored by the prediction model, suggesting that its non-linear relationship with production might not have been fully captured or that its influence is indeed very small.
  • Future Directions: The authors suggest a long-term science and technology strategy for agricultural informatization, including building agricultural informatization strategies, strengthening agricultural science and technology networks, and promoting smart agriculture through big data projects. They also emphasize the need to protect arable land and promote efficient water management technologies.

Summary updated

Abstract

Food security is always a pressing agenda worldwide. The grain production in many areas has decreased due to the reduction in agricultural research funding and infrastructure investment. In this paper, we employed the Extreme-Tree algorithm to determine the main effectors in grain production in Hexi Corridor, Gansu, China, during 2002–2018. First, we applied the three-stage super-SBM DEA to precisely assess agricultural production. Then, we used the Extremely randomized trees algorithm to quantify the importance of each factor. Our results show that the variant of average efficiency score at the first stage was minimal. After removing the influence of environmental factors on production efficiency, the more accurate efficiency score was decreasing from 2002 to 2018. The R2 value of the Extra-Tree model was 0.989 in the grain production analysis. Our research shows that grain production in the Hexi Corridor was controlled by human-driven but not nature-driven during our research period. Based on the importance attribution analysis of each model, it showed that the importance of human-driven investment occupied 93.7% of grain production. The importance of nature-driving was about 6.3%. Accordingly, we proposed corresponding opinions and suggestions to government and growers.
BibTeX
@inproceedings{yan2022comprehensive,
  author = {Zhengxiao Yan and Wei Zhou and Yuyi Wang and Xi Chen},
  title = {{Comprehensive Analysis of Grain Production Based on Three-Stage Super-SBM DEA and Machine Learning in Hexi Corridor, China}},
  booktitle = {Sustainability},
  year = {2022},
  doi = {10.3390/su14148881},
}
Figure