Comparative study of different meteorological yield separation methods in waterlogging disaster assessment
-
摘要: 农业旱涝灾害评估中常采用不同方法计算作物气象产量, 但其中鲜见关于不同方法的比较研究。长江中下游棉区因涝减产现象严重, 因此本文以该地区6省为研究区, 利用标准化降水蒸散指数量化棉花生长期的涝渍强度, 选用7种常见气象产量分离方法(线性拟合、二次多项式拟合、三次多项式拟合、HP滤波法、3年滑动平均法、5年滑动平均法及五点二次平滑法)计算棉花气象产量, 并通过涝渍强度与气象产量的相关分析结果评估不同方法的表现。此外, 结合历史受涝面积对不同方法的涝渍灾害刻画能力进行对比。结果表明: 不同方法计算的棉花气象产量具有相似的长期趋势, 但在短期波动上存在较大差异。就气象产量和涝渍程度的相关性而言, HP滤波法、线性拟合和二次多项式效果更好; 而就气象产量结果与历史受涝面积的相符程度而言, HP滤波最优, 其次是二次多项式和三次多项式。此外, 基于7种方法的相关分析结果均判定湖北省和安徽省是棉花因涝减产最严重的省份, 但对于其余省份的判定结果会存在差异。湖北省和安徽省虽然因涝减产严重, 但两省棉花涝渍的发生倾向低于其余省份。因此, 在农业涝渍评估工作中推荐使用HP滤波法, 而长江中下游棉区的涝渍防治工作应重点关注湖北省和安徽省。Abstract: Various meteorological yield separation methods have been applied in research on agricultural drought and flood disaster assessment. However, a specific comparison of the performance of these methods is rarely performed. The middle-and-lower reach of Yangtze River is an important cotton-production belt in China, but the region is frequently flooded, resulting in severe cotton yield losses. Hence, the objective of the present work was to assess the impacts of waterlogging disasters on cotton yield fluctuation in this cotton-production belt and to compare the accuracy of different meteorological yield separation methods for characterizing the correlations between waterlogging intensity and cotton climatic yields. Six provinces located in this belt were selected as the study areas, and the cotton climatic yields were calculated using various meteorological yield separation methods: linear fitness (LF), quadratic polynomial fitness (QP), cubic polynomial fitness (CP), HP filtering (HP), 3-year moving average (TMA), 5-year moving average (FMA), and five-point quadratic smoothing (FPQS). The performances of the employed methods were compared and well-performing methods were recommended. Specifically, the waterlogging intensity over cotton growth periods was quantified using the widely used standardized precipitation evapotranspiration index (SPEI). Next, according to the correlation between waterlogging intensity and cotton climatic yield, the performances of the seven methods were compared; in addition, this comparison was further performed on historical waterlogging area data. The results indicated that the long-term trends of cotton climatic yield derived from different methods were similar, whereas the short-term trends could be different or even opposite. The absolute values of the cotton climatic yield in Zhejiang Province were obviously lower than those in other provinces, indicating that the cotton plants in Zhejiang suffered much lower yield losses from climatic disasters. Regarding the correlation between waterlogging intensity and cotton climatic yield, LF, QP, and HP were preferable at the provincial scale, and HP, TMA, FMA, and FPQS performed satisfactorily at the county scale. Considering the ability of the seven methods to make predictions in historical waterlogging areas, HP, QP, and CP were the most satisfactory. In general, HP performed the best in various aspects. In a few cases (e.g., counties in Anhui Province), the relationships identified by various methods between waterlogging intensity and cotton climatic yields were different, which implies that the selection of meteorological yield separation methods may alter conclusions in some areas. All methods concluded that Hubei and Anhui Provinces suffered the most severe yield-reducing effect of cotton waterlogging, whereas results in other provinces were inconsistent. Although Hubei and Anhui Provinces were identified as the most waterlogging-affected provinces, the waterlogging intensity over cotton growth periods in these two provinces was lower than that in other provinces. In conclusion, HP filtering was demonstrated to be a preferable method for agricultural waterlogging assessment, and to prevent and control cotton waterlogging disasters in the middle-and-lower reach of the Yangtze River Plain, special attention should be given to Hubei and Anhui Provinces.
-
图 1 研究区及其国家基准气象站分布
Figure 1. Study region and distribution of the national-level meteorological stations
图 2 不同气象产量分离方法拟合的1990—2019年长江中下游6省棉花趋势产量
Figure 2. Cotton trend yields fitted with different meteorological yield separation methods of the six provinces in the middle-and-lower reach of the Yangtze River from 1990 to 2019
AY is actual cotton yield. LF, QP, CP, HP, FMA, FPQS, and TMA indicate linear fitting method, quadratic polynomial method, cubic polynomial method, HP filter method, five-year moving average method, five point quadratic smoothing method, and three-year moving average method, respectively.
图 3 1990—2019年不同气象产量分离方法分离的长江中下游6省棉花气象产量
Figure 3. Climatic cotton yields fitted by different methods of the six provinces in the middle-and-lower reach of the Yangtze River from 1990 to 2019
LF, QP, CP, HP, FMA, FPQS, and TMA indicate linear fitting method, quadratic polynomial method, cubic polynomial method, HP filter method, five-year moving average method, five point quadratic smoothing method, and three-year moving average method, respectively.
图 4 基于7种气象产量方法分离的长江中下游6省棉花气象产量与涝渍指标的相关系数
Figure 4. Correlation coefficients between cotton climatic yield fitted by different methods and cotton waterlogging indicator of the six provinces in the middle-and-lower reach of the Yangtze River
LF, QP, CP, HP, FMA, FPQS, and TMA indicate linear fitting method, quadratic polynomial method, cubic polynomial method, HP filter method, five-year moving average method, five point quadratic smoothing method, and three-year moving average method, respectively.
图 5 基于7种气象产量方法分离的长江中下游6省棉花气象产量与涝渍指标的相关系数的空间分布
Figure 5. Spatial distribution of correlation coefficients between cotton climatic yield fitted by different methods and cotton waterlogging indicator of the six provinces in the middle-and-lower reach of the Yangtze River
图中点标记表示该地区存在显著负相关结果。黄山市和湖州市棉花种植面积和产量均很小, 且个别年份出现了极端气象产量, 因此未纳入计算范畴。The points indicate negatively significant correlation. Cotton planting areas and cotton yields in Huangshan and Huzhou were very few, and their cotton climatic yields were abnormally high in a few years; thus, they were excluded in the calculation processes. LF, QP, CP, HP, FMA, FPQS, and TMA indicate linear fitting method, quadratic polynomial method, cubic polynomial method, HP filter method, five-year moving average method, five point quadratic smoothing method, and three-year moving average method, respectively.
图 6 长江中下游6省典型涝渍年与对应的气象产量
Figure 6. Cotton climatic yields during the typical waterlogging years in the six provinces in the middle-and-lower reach of Yangtze River Plain
LF, QP, CP, HP, FMA, FPQS, and TMA indicate linear fitting method, quadratic polynomial method, cubic polynomial method, HP filter method, five-year moving average method, five point quadratic smoothing method, and three-year moving average method, respectively.
图 7 长江中下游6省1990—2019年棉花涝渍指标均值比较(a)及空间分布特征(b)
Figure 7. Averaged cotton waterlogging intensity (a) and its spatial distribution (b) in the six provinces in the middle-and-lower reach of the Yangtze River Plain from 1990 to 2019
表 1 7种气象产量分离方法拟合的长江中下游6省棉花趋势产量序列与对应方法所得的研究区域平均趋势产量序列间的相关系数
Table 1. Correlation coefficients between the cotton trend yield series fitted by seven meteorological yield separation methods and the averaged trend yield series obtained by the corresponding method in the six provinces in the middle-and-lower reach of the Yangtze River
方法 Method 湖北 Hubei 湖南 Hunan 安徽 Anhui 江苏 Jiangsu 江西 Jiangxi 浙江 Zhejiang 线性拟合 LF 1.000** 1.000** 1.000** 1.000** 1.000** 1.000** 二次多项式 QP 0.398* 0.947** 0.999** 0.936** 0.998** 0.993** 三次多项式 CP 0.479** 0.932** 0.999** 0.739** 0.999** 0.976** HP滤波法 HP 0.411* 0.923** 0.998** 0.858** 0.996** 0.983** 5年滑动平均 FMA 0.462* 0.638** 0.938** 0.719** 0.967** 0.957** 五点二次平滑 FPQS 0.453* 0.643** 0.952** 0.733** 0.975** 0.965** 3年滑动平均 TMA 0.489** 0.623** 0.894** 0.700** 0.940** 0.918** *和**分别表示P<0.05和P<0.01显著性水平。计算样本量为6。* and ** indicate significance levels of P<0.05 and P<0.01, respectively. The calculation sample size is 6. LF, QP, CP, HP, FMA, FPQS, and TMA indicate linear fitting method, quadratic polynomial method, cubic polynomial method, HP filter method, five-year moving average method, five point quadratic smoothing method, and three-year moving average method, respectively.
下载:
