Research of Brassica chinensis var. parachinensis under water stress based on machine learning
-
摘要:目的
以油青菜心Brassica chinensis var.parachinensis为试验对象，基于冠层温度研究其生长过程中的水分胁迫变化规律，并利用机器学习方法，以水分胁迫指数(Crop water stress index, CWSI)和光合有效辐射预测光合作用速率。
方法试验期间，在营养生长阶段(V期)和生殖生长阶段(R期)对油青菜心进行不同田间持水量处理，采集冠层温度、空气温湿度数据，建立无蒸腾作用基线(上限方程)、无水分胁迫基线(下限方程)，通过经验公式计算CWSI。利用基于密度的空间聚类方法和空气温度研究油青菜心的冠气温差上限分布情况，选取固定值作为上限；以CWSI经验公式为基础，使用不同温度定值的无蒸腾作用基线计算CWSI，验证聚类效果。为更简便获取光合作用速率，使用4种机器学习方法：最邻近节点算法(k-Nearest neighbor，KNN)、支持向量回归(Support vector regression，SVR)、极端梯度提升法(Extreme gradient boosting，XGBoost)、随机森林(Random forest，RF)进行预测，并对比预测效果。
结果在不同田间持水量处理下，CWSI能较好地监测油青菜心水分胁迫状况。通过聚类分析，将V期和R期冠气温差上限分类到2个簇中，得到簇心分别为3.4和4.2 ℃，与CWSI经验公式计算值显著相关，表明使用固定值作为油青菜心冠气温差上限值具有可行性。KNN、SVM、XGBoost和RF预测模型均取得较好效果，相关系数分别为0.873、0.877、0.887和0.863。
结论机器学习方法可用于油青菜心光合作用速率的预测，可以避免使用大型笨重仪器，降低对油青菜心叶片的损伤，减少测量时间。
Abstract:ObjectiveBrassica chinensis var. parachinensis was used as the experimental object to study the change rule of water stress during growth process based on canopy temperature, and the machine learning method was used for predicting the photosynthetic rate based on crop water stress index (CWSI) and photosynthetically active radiation.
MethodDuring the test, the experiment adopted different field capacities for B. chinensis var. parachinensis at the vegetative growth stage (V stage) and reproductive growth stage (R stage), collected the canopy temperature, air temperature and humidity data, established non-transpiration-baseline (upper limit equation), non-water-stress-baseline (lower limit equation), and calculated CWSI by empirical formulation. Cluster method of density-based spatial clustering of application with noise and air temperature were used to study the upper limit distribution of canopy temperature minus air temperature of B. chinensis var. parachinensis, and the fixed values were selected as the upper limit. Based on the CWSI empirical formulation, CWSI was calculated using the non-transpiration-baselines with different temperature fixed values to verify the clustering effect. In order to obtain the photosynthetic rate more easily, four machine learning methods of k-nearest neighbor (KNN), support vector regression (SVR), extreme gradient boosting (XGBoost) and random forest (RF) were used for prediction, and the prediction effects were compared.
ResultUnder different field capacities, CWSI could better monitor the water stress status of B. chinensis var. parachinensis. Through cluster analysis, the upper limit of canopy temperature minus air temperature at V stage and R stage was classified into two clusters, and the cluster centers were 3.4 and 4.2 ℃, respectively, which were significantly correlated with the values calculated by the empirical formula of CWSI, indicating that it was feasible to use a fixed value as the upper limit of canopy temperature minus air temperature in B. chinensis var. parachinensis. The prediction models of KNN, SVM, XGBoost and RF all achieved good results, and the correlation coefficients were 0.873, 0.877, 0.887 and 0.863, respectively.
ConclusionMachine learning can be used for predicting the photosynthetic rate of B. chinensis var. parachinensis, avoid the use of large and cumbersome instruments, reduce the damage to the leaves of B. chinensis var. parachinensis, and reduce the measurement time.
-
-
![]()
图 2 冠气温差与饱和水汽压差(VPD)的关系
Figure 2. Relationship between canopy and air temperature difference and vapor pressure deficit (VPD)
![]()
图 3 5种田间持水量处理的水分胁迫指数(CWSI)变化曲线
T1：田间持水量为32.2%；T2~T5：田间持水量分别为T1的85%、70%、55%和40%
Figure 3. Crop water stress index(CWSI) change curves under five field water holding capacities
T1： Field water holding capacity was 32.2%； T2~T5: Field water holding capacity was 85%, 70%, 55% and 40% of T1 respectively
![]()
图 4 基于DBSCAN的冠气温差上限聚类
−1表示离群点，0~5表示聚类形成的簇
Figure 4. Clusters of the upper limit of canopy temperature minus air temperature by DBSCAN
−1 means outliers, 0-5 means clusters formed by clusters
![]()
图 5 不同水分胁迫处理的油青菜心光合作用速率日变化
T1：田间持水量为32.2%；T2~T5：田间持水量分别为T1的85%、70%、55%和40%
Figure 5. Diurnal variations of photosynthetic rate of Brassica chinensis in different water stress treatments
T1： Field water holding capacity was 32.2%； T2~T5: Field water holding capacity was 85%, 70%, 55% and 40% of T1 respectively
![]()
图 6 4种模型的光合作用速率测量值与预测值的散点图
Figure 6. The scatter plots of predicted and measured photosynthetic rates based on four models
表 1 不同冠气温差上限的水分胁迫指数(CWSI)误差分析
Table 1 Error statistics of crop water stress index (CWSI) under different upper limits of canopy and air temperature difference
生长期
Stage处理1)
Treatment固定温度/℃
Fixed temperatureCWSI 平均值
Mean与无蒸腾基线的平均值差值2)
Difference with average of non-transpiration-baseline均方根误差
RMSEV T1 2.0 0.129 −0.019 0.042 3.4 0.103 −0.002 0.008 5.0 0.084 0.011 0.027 T5 2.0 0.801 −0.171* 0.179 3.4 0.635 −0.005* 0.029 5.0 0.515 0.115* 0.125 R T1 2.0 0.066 −0.025 0.066 4.2 0.043 −0.002 0.007 5.0 0.038 0.003 0.010 T5 2.0 0.908 −0.375* 0.402 4.2 0.555 −0.021* 0.033 5.0 0.487 0.047* 0.058 1) T1：田间持水量为32.2%， T5：田间持水量为T1的40%；2) “*”表示T5处理下各固定温度的CWSI平均值与无蒸腾基线的CWSI平均值差值达0.05的显著水平(LSD法)
1) T1： Field water holding capacity was 32.2%, T5: Field water holding capacity was 40% of T1; 2) “*” indicates the significant difference at 0.05 level between CWSI average values of each fixed temperature and non-transpiration-baseline in T5 treatment (Least significant difference method)
下载: 导出CSV
表 2 4种预测模型的误差分析
Table 2 Error statistics of four predicted models
模型
Model决定系数
R2平均绝对误差
MAE均方根误差
RMSEKNN 0.873 0.226 0.294 SVM 0.877 0.234 0.289 XGBoost 0.887 0.227 0.279 RF 0.863 0.239 0.310
下载: 导出CSV
-
[1] 卢宇鹏, 夏岩石, 温少波, 等. 不同熟性菜心品质性状的多样性分析[J]. 广东农业科学, 2020, 47(5): 29-36. [2] 叶红霞. 菜心的栽培季节和栽培方式[J]. 新农村, 2020(10): 26-27. doi: 10.3969/j.issn.1674-8409.2020.10.012 [3] 杨树涛, 黄永文, 刘泳涛, 等. 普宁市菜心生产特色农业气象指标研究[J]. 河南农业, 2018(11): 17-20. [4] 徐燕. 土壤水分胁迫对菜心生理生化指标及气孔发育的影响[D]. 广州: 暨南大学, 2010. [5] IDSO S B, JACKSON R D, PINTER P J, et al. Normalizing the stress-degree-day parameter for environmental variability[J]. Agricultural Meteorology, 1981, 24(1): 45-55.
[6] JACKSON R D, IDSO S B, REGINATO R J, et al. Canopy temperature as a crop water stress indicator[J]. Water Resources Research, 1981, 17(4): 1133-1138. doi: 10.1029/WR017i004p01133
[7] JONE H G. Use of infrared thermometry for estimation of stomatal conductance as a possible aid to irrigation scheduling[J]. Agricultural and Forest Meteorology, 1999, 95(3): 139-149. doi: 10.1016/S0168-1923(99)00030-1
[8] RUD R, COHEN Y, ALCHANATIS V, et al. Crop water stress index derived from multi-year ground and aerial thermal images as an indicator of potato water status[J]. Precision Agriculture, 2014, 15(3): 273-289. doi: 10.1007/s11119-014-9351-z
[9] 崔晓, 许利霞, 袁国富, 等. 基于冠层温度的夏玉米水分胁迫指数模型的试验研究[J]. 农业工程学报, 2005, 25(8): 22-24. doi: 10.3321/j.issn:1002-6819.2005.08.005 [10] 赵福年, 王瑞君, 张虹, 等. 基于冠气温差的作物水分胁迫指数经验模型研究进展[J]. 干旱气象, 2012, 30(4): 522-528. [11] JONES H G, STOLL M, SANTOS T, et al. Use of infrared thermography for monitoring stomatal closure in the field: Application to grapevine[J]. Journal of Experimental Botany, 2002, 53(378): 2249-2260. doi: 10.1093/jxb/erf083
[12] BIAN J, ZHANG Z T, CHEN J Y, et al. Simplified evaluation of cotton water stress using high resolution unmanned aerial vehicle thermal imagery[J]. Remote Sensing, 2019, 11(3): 267. doi: 10.3390/rs11030267.
[13] 王卫星, 罗锡文, 区颖刚, 等. 基于冠层温度的菜心缺水指数模型初步试验研究(英文)[J]. 农业工程学报, 2003, 19(5): 47-50. doi: 10.3321/j.issn:1002-6819.2003.05.010 [14] 孙道宗, 王卫星, 唐劲驰, 等. 茶树水分胁迫建模及试验[J]. 排灌机械工程学报, 2017, 35(1): 65-70. doi: 10.3969/j.issn.1674-8530.15.0249 [15] 张立元, 牛亚晓, 韩文霆, 等. 大田玉米水分胁迫指数经验模型建立方法[J]. 农业机械学报, 2018, 49(5): 233-239. doi: 10.6041/j.issn.1000-1298.2018.05.027 [16] AGAM N, COHEN Y, ALCHANATIS V, et al. How sensitive is the CWSI to changes in solar radiation?[J]. International Journal of Remote Sensing, 2013, 34(17): 6109-6120. doi: 10.1080/01431161.2013.793873
[17] KUMAR N, ADELOYE A J, SHANKAR V, et al. Neural computing modelling of the crop water stress index[J/OL]. Agricultural Water Management, 2020, 239: 1-10. [2021-01-05]. https://doi.org/10.1016/j.agwat.2020.106259.
[18] KHORSANDI A, HEMMAT A, MIREEI S A, et al. Plant temperature-based indices using infrared thermography for detecting water status in sesame under greenhouse conditions[J]. Agricultural Water Management, 2018, 204: 222-233. doi: 10.1016/j.agwat.2018.04.012
[19] 屈振江, 郑小华, 王景红, 等. 渭北旱塬苹果园内外温度变化特征研究[J]. 干旱区地理, 2016, 39(2): 301-308. [20] 吐尔逊·买买提, 谢建华. 基于DBSCAN的农机作业轨迹聚类研究[J]. 农机化研究, 2017, 39(4): 7-11. doi: 10.3969/j.issn.1003-188X.2017.04.002 [21] 谢慧婷. 基于红外热成像技术的生菜缺水指标的研究[D]. 福州: 福建农林大学, 2016. [22] MATESE A, BARALDI R, BERTON A, et al. Estimation of water stress in grapevines using proximal and remote sensing methods[J]. Remote Sensing, 2018, 10(1): 114. doi: 10.3390/rs10010114.
[23] ZHANG L Y, NIUY X, ZHANG H H, et al. Maize canopy temperature extracted from UAV thermal and RGB imagery and its application in water stress monitoring[J]. Frontiers in Plant Science, 2019, 10: 1270. doi: 10.3389/fpls.2019.01270.
[24] SONG X Y, ZHOU G S, HE Q J, et al. Stomatal limitations to photosynthesis and their critical water conditions in different growth stages of maize under water stress[J/OL]. Agricultural Water Management, 2020, 241: 1-12. [2021-01-05]. https://doi.org/10.1016/j.agwat.2020.106330.
[25] 刘煦. 林下参种植光环境的动态预测与评价研究[D]. 长春: 吉林大学, 2014. [26] 陈硕博. 无人机多光谱遥感反演棉花光合参数与水分的模型研究[D]. 杨凌: 西北农林科技大学, 2019. [27] 陈俊英, 陈硕博, 张智韬, 等. 无人机多光谱遥感反演花蕾期棉花光合参数研究[J]. 农业机械学报, 2018, 49(10): 230-239. doi: 10.6041/j.issn.1000-1298.2018.10.026 [28] 宋飞扬, 铁治欣, 黄泽华, 等. 基于KNN-LSTM的PM_(2.5)浓度预测模型[J]. 计算机系统应用, 2020, 29(7): 193-198. [29] BOTULA Y D, NEMES A, MAFUKA P, et al. Prediction of water retention of soils from the humid tropics by the nonparametric k-nearest neighbor approach[J]. Vadose Zone Journal, 2013, 12(2). doi: 10.2136/vzj2012.0123.
[30] MOHAMMADI B, MEHDIZADEH S. Modeling daily reference evapotranspiration via a novel approach based on support vector regression coupled with whale optimization algorithm[J/OL]. Agricultural Water Management, 2020, 237: 1-13. [2020-01-03]. https://doi.org/10.1016/j.agwat.2020.106145.
[31] WANG T T, YANG C H, LIANG J. Soil pH value prediction using UWB radar echoes based on XGBoost[C]//CSPS. International Conference in Communications, Signal Processing, and Systems. Urumqi: CSPS, 2019: 1941-1947.
[32] 白婷, 丁建丽, 王敬哲. 基于机器学习算法的土壤有机质质量比估算[J]. 排灌机械工程学报, 2020, 38(8): 829-834. [33] 罗清元, 杨丹, 刘丽娜, 等. 基于不同环境下的河南省典型区域土壤田间持水量研究[J]. 节水灌溉, 2019, 4929(6): 35-38. doi: 10.3969/j.issn.1007-4929.2019.06.009 [34] MARTIN E, KRIEGEL H P, SANDER J, et al. A density-based algorithm for discovering clusters in large spatial databases with noise[C]//AAAI. Proceedings of the Second International Conference on Knowledge Discovery and Data Mining (KDD-96). California: AAAI, 1996: 226-231.
[35] ALTMAN N S. An introduction to kernel and nearest-neighbor nonparametric regression[J]. The American Statistician, 1992, 46(3): 175-185.
[36] CORTES C, VAPNIK V. Support-vector networks[J]. Machine Learning, 1995, 20(3): 273-297.
[37] CHEN T Q, GUESTRIN C. XGBoost: A scalable tree boosting system[C]//ACM. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. California: ACM, 2016: 785-794.
[38] BREIMAN L. Random forests[J]. Machine Learning, 2001, 45(1): 5-32. doi: 10.1023/A:1010933404324
