3.1 Validation of extracted spring maize information
3.1.1 Verification by confusion matrix
The accuracy of extraction of spring maize information was verified using a confusion matrix (also called an error matrix). The row of the matrix was the sampling data from field investigations, and the column was the extraction result of this study. From Table 2, it can be seen that the overall accuracies were higher than 85% from 2018 to 2020, and the kappa coefficients were greater than 0.83. From 2018 to 2020, spring maize was mainly in the west of Xiaoxing'an Mountains, Sanjiang, Songnen, and Liaohe Plains, north of Liaoxi Hills, and south of Liaodong Hills (Fig. 4a, b, and c). These areas are mainly plains and hills. Spring maize was also cultivated on the gentle terrain at the edge of the mountains.
Table 2
Accuracy evaluation of spring maize classification results. The overall accuracy refers to the percentage of checkpoints of all correctly classified land cover categories relative to the sum number of checkpoints. The producer’s accuracy represents the percentage of the actual reference data for each category that is correctly classified. The user’s accuracy represents the percentage of test data for each category that is correctly classified as belonging to that category. Kappa coefficient represents the proportion of error reduction generated by this classification versus that obtained by a completely random classification, where a higher Kappa value indicates higher accuracy in classification (Petropoulos et al. 2012; Chen et al. 2017; Foody et al. 2020).
Year | User’s accuracy (%) | Producer’s accuracy (%) | Kappa coefficient | Overall accuracy (%) |
2018 | 88.23 | 86.62 | 0.84 | 89.53 |
2019 | 91.34 | 84.34 | 0.85 | 87.49 |
2020 | 89.56 | 83.62 | 0.83 | 85.23 |
3.1.2 Verification of provincial results
The statistical data of crop planting areas were from the China Statistical Yearbook. Extracted data from 2018 to 2020 were calculated in this study. The extracted data were compared with the statistical data. A positive error indicates that the extracted area was bigger than the statistical, and a negative value corresponds to a larger statistical area. The results of this study were compared with the statistical area of spring maize. The annual errors were 11.39% in Liaoning, 7.00% in Jilin, and 54.34% in Heilongjiang. And compared with the statistical area of spring maize and soybean, the errors were lower for the three provinces: 7.67%, 5.49%, and 14.85%, respectively (Table 3). Therefore, the planting area of spring maize extracted in this study was a mix of soybean and spring maize.
Table 3
Extraction error of spring maize information in Northeast China from 2018 to 2020.
Year | Province | Extracted area (104 km2) | Spring maize | Spring maize & soybean |
Statistical area (104 km2) | Error (%) | Statistical area (104 km2) | Error (%) |
2018 | Liaoning | 3.09 | 2.71 | 14.02 | 2.79 | 10.75 |
Jilin | 4.63 | 4.23 | 9.46 | 4.51 | 2.66 |
Heilongjiang | 7.47 | 6.31 | 18.38 | 9.89 | -24.47 |
2019 | Liaoning | 2.98 | 2.67 | 11.61 | 2.76 | 7.97 |
Jilin | 3.99 | 4.21 | -5.23 | 4.56 | -12.50 |
Heilongjiang | 10.75 | 5.87 | 83.13 | 10.15 | 5.91 |
2020 | Liaoning | 2.92 | 2.69 | 8.55 | 2.80 | 4.29 |
Jilin | 4.55 | 4.28 | 6.31 | 4.61 | -1.30 |
Heilongjiang | 8.85 | 5.48 | 61.50 | 10.31 | -14.16 |
3.2 The spatio-temporal pattern of spring maize planting areas
As shown in Fig. 5a, 64.14% of the land in the study area was sown with spring maize from 2002 to 2020. The areas sown during multiple planting years were mainly in the hills and plains in the southwest and the plains in the northeast. The number showed a decreasing trend moving from these areas to the mountains in the north and northeast (Fig. 5a). The central Sanjiang Plain, Songnen, and Liaohe Plains, north and east of Liaoxi Hills, and south of Liaodong Hills were stable planting areas of spring maize, and these areas accounted for 11.86% of the study area (Fig. 5b). These stable planting areas were mainly plains and hills, rather than the mountains. There were fluctuating planting areas that were located near stable planting areas but not planted every year; these areas accounted for 17.41% (Fig. 5b). The low-frequency planting areas were mainly in the south of Xiao Xing'an Mountains, north and east of Sanjiang Plain, north of Songnen Plain, and central Liaoxi Hills, together accounting for 34.86% (Fig. 5b).
3.3 Inter-annual variation of spring maize planting areas during three stages
There were three developmental stages with different characteristics in the evolution of spring maize distribution in Northeast China from 2002 to 2020. The pre-adjustment stage was a time of continuous increase, the in-adjustment stage experienced adjustment and reduction, and the post-adjustment stage included optimization and recovery.
In the pre-adjustment stage (Fig. 6a), the extent of planting did not change in the middle of Sanjiang Plain, central and south of Songnen Plain, north and west of Liaohe Plain, north and east of Liaoxi Hills, and south of Liaodong Hills. Compared with 2002, the reduced areas in 2014 were primarily in the middle of Sanjiang Plain and the north and east of Songnen Plain, showing minor shrinkage to the southward and westward. The increase was mainly in the north of Xiao Xing'an Mountains, north of Sanjiang Plain, and north of Songnen Plain, increasing to the east and north. The south of Liaohe Plain, central Liaoxi Hills, and west of Liaodong Hills increased to the south and east. The unchanged, reduced, and expanded planting areas accounted for 29.80%, 12.00%, and 58.20% of the sum planting area in 2014. The planting area mainly showed an increasing trend in this stage. The increased areas were mainly plains and hills, and reduced areas were at the junctions of plains and mountains, areas of high terrain.
During the in-adjustment stage (Fig. 6b), spring maize cultivation was unchanged in the middle Sanjiang Plain, Songnen, and Liaohe Plains, Liaoxi Hills, and south of Liaodong Hills. Compared with 2015, the reduced areas in 2017 were primarily distributed in the south of Xiao Xing'an Mountains, north of Sanjiang Plain, and north of Songnen Plain, with southward and westward contraction. There were smaller increases in the west and south of Songnen Plain, north and central of Liaohe Plain, and central Liaoxi Plain. The unchanged, reduced, and expanded planting areas accounted for 44.68%, 34.62%, and 20.70% of the total planted area in 2017. The planting area mainly showed a decreasing trend in this stage, with reduced areas primarily in plains and hills in the north, and increases s scattered in plains and hills.
In the post-adjustment stage (Fig. 6c), spring maize cultivation was unchanged in the middle Sanjiang Plain, Songnen, and Liaohe Plain, north of Liaoxi Hills, and south of Liaodong Hills. Compared with 2018, the distribution in 2020 shrank southward and westward in the west of Xiao Xing'an Mountains and east of Sanjiang Plain. The west of Songnen Plain shrank eastward, and the west and south of Liaohe Plain, south of Liaoxi Hills, and south of Liaodong Hills shrank northward. Cultivation mainly expanded northward and eastward in the north of Xiao Xing'an Mountains, north of Sanjiang Plain, east and north of Songnen Plain, and the central Liaoxi Hills expanded southward. The unchanged, reduced, and expanded planting areas accounted for 43.46%, 26.33%, and 30.21% of the total scope in 2020. There were roughly equivalent increases and decreases in the planting area of spring maize, with changes for all kinds of terrain.
3.4 Drought in spring maize planting areas
3.4.1 Drought in specific spatial patterns
Occurrences of mild, moderate, and severe drought in the stable planting areas of spring maize mainly occurred in the middle of Sanjiang Plain, Songnen, and Liaohe Plains, north and east of Liaoxi Hills, and south of Liaodong Plain. Mild drought mainly occurred at high frequency in the central Sanjiang Plain, Songnen Plain, and north of Liaoxi Hills. Moderate drought primarily occurred in the Songnen, and Liaohe Plains, north of Liaoxi Hills, and south of Liaodong Hills at high frequency. Severe drought mainly occurred frequently in the south of Liaohe Plain, north and east of Liaoxi Hills, and south of Liaodong Hills (Fig. 7a). Drought in the fluctuating planting areas mainly occurred in the south of Xiaoxing'an Mountains, middle of Sanjiang Plain, Songnen, and Liaohe Plains, north of Liaoxi Hills, and south of Liaodong Hills. Mild drought occurred at high frequency in the south of Xiaoxing'an Mountains, middle of Sanjiang Plain, Songnen, and Liaohe Plains. Moderate drought mainly occurred frequently in the south of Xiaoxing'an Mountains, middle of Sanjiang Plain, Songnen, and Liaohe Plains, and north of Liaoxi Hills. Severe drought was primarily distributed in the south of Xiaoxing’an Mountains, south of Liaohe Plain, and north and east of Liaoxi Hills at high frequency (Fig. 7b). The drought in the low-frequency growing areas was distributed in the west of Xiaoxing’an Mountains, Sanjiang, and Songnen Plains, central Changbai Mountains, Liaohe Plain, Liaoxi Hills, and south of Liaodong Hills. Mild drought frequently occurred in the Sanjiang, Songnen, and Liaohe Plains. Moderate drought frequently occurred in the Songnen and Liaohe Plains. Severe drought occurred in the west of Songnen Plain, Liaoxi Hills, and south of Liaodong Hills at high frequency (Fig. 7c). Drought mainly occurred frequently in plains and hills in the west of the study area. Compared with the other two areas, moderate and severe droughts were most frequent in stable planting areas, 33.62%, and 19.83%. Heavier droughts were also more likely to occur in this area, and severe drought mainly occurred frequently in the hills in the southwest Liaoning Province.
Table 4
The maximum numbers and percentages of drought occurrences for different planting areas, 2002–2020.
Planting areas | Mild | Percentage (%) | Moderate | Percentage (%) | Severe | Percentage (%) |
Stable | 81 | 46.55 | 61 | 33.62 | 67 | 19.83 |
Fluctuating | 80 | 49.81 | 59 | 31.46 | 69 | 18.73 |
Low frequency | 78 | 52.02 | 61 | 29.49 | 71 | 18.49 |
3.4.2 Drought in expanded planting areas of spring maize
The three levels of drought in the expanded planting areas in the pre-adjustment stage mainly occurred in the middle of Sanjiang Plain, Songnen, and Liaohe Plains, Liaoxi Hills, and south of Liaodong Plain. Mild and moderate droughts mainly occurred at high frequency in the middle of Sanjiang Plain, Songnen, and Liaohe Plains. Severe drought mainly occurred frequently in Liaoxi Hills (Fig. 8a). Drought during the in-adjustment stage occurred primarily in the Songnen Plain, Liaohe Plain, and Liaoxi Hills. Mild and moderate droughts occurred at high frequency in the Songnen and Liaohe Plain. Severe drought was mainly distributed in the Liaoxi Hills and south of Liaohe Plain at high frequency (Fig. 8b). Drought in the post-adjustment stage was mainly distributed in the Sanjiang, Songnen, and Liaohe Plains, and Liaoxi Hills. Mild and moderate droughts were frequently distributed in the Sanjiang, Songnen, and Liaohe Plains. Severe drought was distributed in the Liaoxi Hills at a high frequency (Fig. 8c). Drought in expanded planting areas occurred most frequently in plains and hills in the west of Northeast China. The largest percentages of moderate and severe droughts occurred in the pre-adjustment and in-adjustment stages, respectively. There were more droughts in the pre-adjustment stage, with gradual decreases in the latter two stages. Drought in the post-adjustment stage was mainly mild, and severe drought frequently occurred in the Liaoxi Hills.
Table 5
The maximum numbers and percentages of drought occurrences in the different stages, 2002–2020.
Stages | Mild | Percentage (%) | Moderate | Percentage (%) | Severe | Percentage (%) |
Pre-adjustment | 58 | 48.75 | 42 | 32.68 | 44 | 18.57 |
In-adjustment | 20 | 44.78 | 14 | 32.54 | 16 | 22.68 |
Post-adjustment | 21 | 58.87 | 15 | 26.38 | 16 | 14.75 |