摘要:近几十年来, 我国大豆产需缺口不断扩大, 提升大豆单产水平已成为当前提高大豆总产量的首要可行举措。然而, 影响我国大豆单产的驱动因子及其地域空间差异特征并不明晰。本文通过搜集1952年、1965年、1978年、1990年、2000年、2010年和2017年的全国各省市农业统计年鉴等数据, 从大豆种植的管理措施、自然因素、科技水平、社会因素、经济因素等方面选取与大豆生产密切相关的13个因子, 以大豆单产作为目标变量构建增强回归树模型, 量化各因子的相对重要性及其与大豆单产之间的关系, 分析大豆单产的变异特征, 揭示全国尺度及4个大豆主产区之间的大豆单产驱动力时空分异特征。研究结果表明: 1)各年份的大豆单产变异系数为34.1%~73.2%, 表明全国各地市大豆单产之间存在较大的差异。本研究构建的增强回归树模型可有效解释43.3%的大豆单产变异性, 并可量化揭示各因子与大豆单产之间的非线性关系。2) 1952年以来影响我国大豆单产水平的最重要因素依次为大豆播种面积占农作物总种植面积的百分比(相对重要性为20.9%)、文盲率(18.9%)、每公顷化肥(折纯)施用量(10.7%)。3)不同主产区的大豆单产核心驱动力存在空间差异, 北方春大豆区的最重要因素为每公顷农业机械总动力(13.1%)、文盲率(11.8%), 黄淮海流域夏大豆区的最重要因素为每公顷化肥(折纯)施用量(25.6%)、每公顷农药(折纯)施用量(18.4%), 长江流域春夏大豆区的最重要因素为研发支出占地区生产总值的百分比(21.5%)、有效灌溉面积占农作物播种面积的百分比(14.3%), 南方多熟大豆区的最重要因素为每公顷化肥(折纯)施用量(22.7%)、第一产业占地区生产总值的百分比(13.3%)。4)大豆播种面积占农作物总播种面积的百分比对于全时期、改革开放前、改革开放后3个时期均是影响大豆单产最重要的因子, 改革开放前其他重要的因子包括文盲率和每公顷化肥(折纯)施用量, 改革开放后则包括每公顷农业机械总动力和年均温。总之, 我国各大豆主产区需合理施用化肥和农药, 努力提高机械化水平和农业生产者的知识水平, 本研究结果可为各省市采取有效措施提升大豆单产水平提供科学依据。
关键词:大豆/
产量/
驱动因素/
增强回归树模型/
空间异质性
Abstract:Over the past several decades, the consumers’ demand for soybeans has grown rapidly in China, resulting in a significant increase in the gap between production and demand. Therefore, increasing the total soybean output is of critical importance to ensure food security. Given that it is difficult to increase the total area of cultivated land in China, improving soybean yield per unit area land has become the primary measure for increasing the total soybean output. However, the determinants that directly affect soybean yield, the regional spatial heterogeneity of yield remain unclear. In this study, data from agricultural statistical yearbooks at both the provincial and prefecture levels in China as well as meteorological data (e.g., temperature, precipitation, and sunshine duration) from 1952 to 2017 (comprising 1952, 1965, 1978, 1990, 2000, 2010, and 2017) were collected, whereupon 13 factors closely related to soybean production were selected from the perspective of planting management measures, natural factors, scientific and technogical levels, social factors, and economic factors. Several boosted regression tree models were built to quantify the relative importance of each factor and to determine the mechanism through which it influenced soybean yield; to analyze the variation characteristics of soybean yield; and to reveal the spatiotemporal characteristics of key driving forces across the national scale and among the four major soybean-producing areas (i.e., the northern spring soybean area, the summer soybean area in the Huang-Huai-Hai Basin, the spring and summer soybean area in the Yangtze River Basin, and the southern soybean area) over a long period since 1952. The following results were obtained. 1) The coefficient of variation of soybean yields in different years ranged from 34.1% to 73.2%, indicating that there were substantial differences in yield across the regions in China. The boosted regression tree model could effectively explain 43.3% of the soybean yield variability and quantitatively revealed the nonlinear relationship between each factor and soybean yield in the national scale. 2) The most important factor affecting soybean yield in China since 1952 was the soybean sown area as a percentage of the total crop sown area (relative importance of 20.9%), followed by the illiteracy rate (18.9%) and fertilizer consumption (pure amount) per hectare (10.7%). 3) Spatial differences existed in the dominant driving factors of soybean yield among different main production areas. The main driving factors of the northern spring soybean area were the total power of agricultural machinery per hectare (13.1%) and the illiteracy rate (11.8%); those for the summer soybean area in the Huang-Huai-Hai Basin were the fertilizer consumption (pure amount) per hectare (25.6%) and pesticide consumption (pure amount) per hectare (18.4%); those for the spring and summer soybean area in the Yangtze River Basin were the R&D expenditure as a percentage of regional GDP (21.5%) and the effective irrigation area as a percentage of the crop sown area (14.3%); and those for the southern soybean area were the fertilizer consumption (pure amount) per hectare and the primary industry as a percentage of regional GDP (13.3%). 4) The soybean sown area as a percentage of the total crop sown area was the most important factor that affected soybean yield during 1952–2017, both before and after the reformation and opening up of China. Additionally, the illiteracy rate and fertilizer consumption (pure amount) per hectare were two other important factors for the period before the reformation and opening up of the country, whereas the total power of agricultural machinery per hectare and annual average temperature were important factors afterwards. This study revealed the determinants of soybean yield and its spatiotemporal heterogeneity in China since 1952 and determined the effective measures for improving the yield of this important crop. These findings should be useful for soybean production-related departments at both the provincial and prefecture levels in China for improving the rational usage of fertilizers and pesticides, increasing the level of mechanization, and enhancing the knowledge level of agricultural producers.
Key words:Soybean/
Yield/
Driving factors/
Boosted regression tree model/
Spatial heterogeneity
PDF全文下载地址:
http://www.ecoagri.ac.cn/article/exportPdf?id=3dd2c61b-68a9-4379-88d7-4b775db686d9
