基于不同分析模型的大豆叶片SPAD值和LAI光谱估算比较

为探讨在大豆鼓粒期采用光谱技术估算叶片SPAD值和LAI的有效分析模型和方法，本研究以大田鼓粒期大豆为试验材料，在3个不同时段9:45～10:15(10AM)、11:45～12:15(12PM)和13:45～14:15(2PM)测量冠层全波段光谱反射率，并分别使用极限学习机(ELM)、偏最小二乘回归(PLSR)、支持向量机(SVM)和随机森林(RF)构建大豆叶片SPAD值和LAI估算模型，并对比不同模型分析结果的估算精度。结果表明：在各模型中，12PM和2PM测定的光谱反射率与大豆叶片SPAD值和LAI的拟合精度均高于10AM。基于RF的大豆叶片SPAD值估算模型验证集的R²为0.910,RMSE为2.006,MRE为3.684;基于RF的大豆LAI估算模型验证集的R²为0.916,RMSE为0.209,MRE为4.383,与ELM、PLSR和SVM相比，有更高的估算精度。综上结果说明大豆鼓粒期在11:45～12:15和13:45～14:15采用RF模型，运用全波段的光谱反射率估算大豆叶片SPAD值和LAI可得到较准确的结果。     

基于Sentinel-2多光谱数据和机器学习算法的冬小麦LAI遥感估算

为了丰富大田尺度下冬小麦叶面积指数的遥感估算方法并提高估算精度,以关中地区冬小麦为对象,基于Sentinel-2多光谱卫星数据与地面同步观测的冬小麦叶面积指数样点数据,应用偏最小二乘回归(PLSR)、反向传播神经网络(BPNN)和随机森林(RF)法构建冬小麦叶面积指数估算模型,进行区域冬小麦叶面积指数遥感反演。结果表明,Sentinel-2多光谱卫星影像中心842nm近红外B8波段与冬小麦叶面积指数相关性最好,样本总体相关系数为0.778;植被指数中反向差值植被指数(IDVI)与冬小麦叶面积指数相关性最好,样本总体相关系数为0.776。各种估算模型中LAI-RF模型预测效果最佳,r~2为0.72,RMSE为0.53,RE为16.83%。基于LAI-RF估算模型,应用Sentinel-2多光谱卫星数据较好地反演了研究区冬小麦叶面积指数区域分布,其结果总体上与地面真实情况接近,说明以Sentinel-2卫星影像数据建立LAI-RF估算模型,可应用于区域冬小麦LAI反演制图。     

基于无人机可见光遥感的冬小麦株高估算

株高是作物生长过程中重要的生长指标。为探索快速准确获取作物株高的方法,利用无人机可见光图像采集系统,获取冬小麦拔节期至成熟期的高清数码图像,建立冬小麦拔节期、抽穗期、灌浆期及成熟期的作物数字表面模型(digital surface models,DSM)及作物高度模型(crop height model,CHM),并对模型进行验证。结果表明,冬小麦株高各生育时期CHM提取值与地面实测值极显著相关(P<0.01),误差为-0.10~0.09m,相对误差为17.64%~19.60%。株高预测值与实测值拟合性较高(R^2=0.82,RMSE=4.31cm)。这说明用无人机拍摄的高清数码影像可快速估算冬小麦的株高。     

基于新型植被指数的冬小麦覆盖度遥感估算

为提高冬小麦覆盖度估测精度,从增强近红外与红光差别的数学变换原理出发,构建了一种新型植被指数(NDVIn),再基于2013、2014年冬小麦冠层高光谱和模拟的资源三号卫星宽波段多光谱数据,分别构建基于常规植被指数(NDVI)与NDVIn的冬小麦覆盖度估算模型,然后利用留一交叉验证法对模型精度进行评价。结果表明,当n=6时,新生成的植被指数NDVI6对冬小麦农田覆盖度具有最好的估算性能,利用其基于小麦冠层高光谱及卫星多光谱数据建立的冬小麦覆盖度估算模型的决定系数r2分别为0.84、0.85,RMSE分别为0.092、0.091,模型精度均好于常规指数NDVI的估算结果。说明NDVI6用于估测冬小麦覆盖度具有可行性。     

基于高光谱特征参数的冬小麦氮营养指数估算

为了实现快速高精度获取冬小麦氮营养指数的高光谱监测技术，利用美国SVC HR-1024I型野外光谱辐射仪对2017-2019年关中地区的冬小麦进行遥感监测，获取“三边”参数、任意两波段光谱指数和植被指数，通过相关性分析和逐步回归分析方法筛选冬小麦氮营养指数的敏感光谱参数，结合偏最小二乘回归(PLSR)、随机森林算法(RFR)、支持向量机回归(SVR)和梯度增强回归(GBDT)建立冬小麦氮营养指数模型，并对模型估算精度进行验证。结果表明，从拔节期到灌浆期，各时期的氮营养指数与任意两波段光谱指数均呈极显著相关，其中拔节期氮营养指数与任意两波段光谱指数相关性均高于其他时期，且基于一阶导数光谱的归一化光谱指数和比值光谱指数与氮营养指数的相关系数最大，为0.66。拔节期基于梯度增强回归的冬小麦氮营养指数预测模型的决定系数(r²)和均方根误差(RMSE)分别为0.96和0.05,模型验证的r²、RMSE和相对预测偏差(RPD)分别为0.95、0.12和2.12,模型预测精度最高。因此，拔节期基于梯度增强回归的冬小麦氮营养指数估算模型可用于冬小麦氮营养监测...     

基于无人机多光谱遥感的冬小麦冠层叶绿素含量估测研究

为探讨利用无人机多光谱影像监测冬小麦叶绿素含量的可行性,基于北京市大兴区中国水科院试验基地的2019年冬小麦无人机多光谱影像和田间实测冠层叶绿素含量数据,选取16种光谱植被指数,确定对冬小麦冠层叶绿素含量显著相关的植被指数,采用一元二次线性回归和逐步回归分析方法建立各生育时期及全生育期的SPAD值估测模型,通过精度检验确定对冬小麦冠层叶绿素含量监测的最优模型。结果表明,两种分析方法中逐步回归建模效果最佳。拔节期选取4个植被指数(MSR、CARI、NGBDI、TVI)建模效果最好,模型率定的决定系数(r~2)为0.73,模型验证的r~2、相对误差(RE)和均方根误差(RMSE)分别为0.63、2.83%、1.68;抽穗期选取3个植被指数(GNDVI、GOSAVI、CARI)建模效果最好,模型率定的r~2为0.81,模型验证的r~2、RE、RMSE分别为0.63、2.83%、1.68;灌浆期选取2个植被指数(MSR、NGBDI)建模效果最好,模型率定的r~2为0.67,模型验证的r~2、RE、RMSE分别为0.65、2.83%、1.88。因此,无人机多光谱影像结合逐步回归模型可以很好地监测冬小麦SPAD值动态变化。     

基于Sentinel-2影像和机器学习算法的冬小麦秸秆覆盖度遥感估算

为探究大范围小麦秸秆覆盖度(CRC)估测方法，以冬小麦秸秆为研究对象，基于Sentinel-2遥感卫星影像光谱指数、波段和纹理特征及其不同特征组合，利用灰色关联-随机森林(GRA-RF)敏感特征提取方法，结合高斯过程(GPR)、套索(LASSO)、岭回归(RR)和偏最小二乘(PLSR)等多种机器学习算法，开展小麦CRC估算的最优模型研究。结果表明，基于GRA-RF特征优选后的机器学习模型显著改善了小麦CRC的估算精度，LASSO算法总体对小麦CRC的估测效果最佳，并且针对不同的光谱特征组合表现出差异化的结果。其中，以光谱指数、波段和纹理信息构成的组合特征集构建的CRC遥感估算模型精度最优(r²=0.65,RMSE=9.25%),以波段与纹理两者组合特征估算的CRC精度次之(r²=0.63,RMSE=9.31%),仅利用单一的光谱指数、波段或者纹理特征估算冬小麦CRC的精度均劣于组合特征的结果。这说明应用GRA-RF组合筛选方法能够有效优选秸秆覆盖度的光谱特征；相比于单一特征，光谱指数、波段、纹理信息等构成的组合特征更能有效地监测小麦秸秆覆盖度...     

基于机器学习算法的冬小麦不同生育时期生物量高光谱估算

为探索适用于冬小麦不同生育时期的高光谱估算方法,基于4年大田试验,以江苏省主要冬小麦品种为材料,以8种对常用生物量敏感的高光谱指数为基础,分别采用偏最小二乘算法、支持向量回归算法、随机森林算法在冬小麦4个主要生育时期(抽穗期前、抽穗期、开花期和灌浆期)进行了高光谱生物量估算和预测能力比较。结果表明,在冬小麦不同生育时期,高光谱估算生物量精度差异显著;利用随机森林构建的生物量估算模型在4个生育时期均表现出很好的效果,决定系数(r^2)和均方根误差(RMSE)在抽穗期前分别为0.79和44.82 g·m^-2,在抽穗期分别为0.71和62.07 g·m^-2,在开花期分别为0.70和97.63 g·m^-2,在灌浆期分别为0.71和106.98 g·m^-2;随机森林模型在4个生育时期的预测能力都高于或接近于支持向量回归模型,高于偏最小二乘回归模型,r^2和RMSE在抽穗期前分别为0.60和72.54 g·m^-2,在抽穗期分别为0.60和75.07 g·m^-2,在开花期分别为0.68和109.9 g·m^-2,在灌浆期分别为0.61和127.93 g·m^-2。这说明随机森林算法在冬小麦不同生育时期生物量高光谱遥感估算方面具有较高的精度和稳定性。     

基于数字图像特征的冬小麦渍害监测研究

为探索渍害胁迫下冬小麦灾损程度的可视化监测方法,通过田间试验,分析了麦田16个常用图像特征指数在不同受渍时间下的变化特征及其与冬小麦SPAD值、产量和千粒重的相关关系,并建立了基于图像特征指数衰减量的冬小麦渍害估算模型。结果表明,随渍水时间的增加,红光(R)、红光标准化值(NRI)、超红指数(EXR)、植被颜色指数(CIVE)极显著上升,而绿光标准化值(NGI)、归一化绿红差值指数(NGRDI)、绿-红差值指数(GMR)、超绿指数(EXG)、绿红比值指数(GRVI)则极显著下降;且这9个图像特征指数均与冬小麦SPAD值、产量和千粒重呈极显著相关,相关系数的最大绝对值分别为0.92、0.85和0.91;基于图像指数衰减量所建的SPAD值、产量和千粒重减少量的估算模型均以二次多项式最优,且以CIVE指数衰减量构建的SPAD值、产量和千粒重减少量估算模型的预测精度最高,验证集决定系数分别达到0.98、0.95、0.96。因此,数字图像技术可用于冬小麦渍害监测,且以基于CIVE指数的监测效果最佳。     

冬小麦白粉病冠层光谱特征解析与病情指数反演

为探讨利用高光谱技术快速无损地监测小麦白粉病灾情的方法,通过人工田间诱发白粉病,在灌浆期对不同发病等级(病情指数)的冬小麦进行冠层高光谱测定,对原始光谱数据进行一阶微分处理,筛选最佳光谱特征参量和植被指数,构建冬小麦白粉病病情指数反演模型。结果表明,在冠层尺度,小麦白粉病"红边"位置均在730nm左右(±1nm);经验证,5种模型中三角植被指数(TVI)模型估算精度最好,r2和RMSE分别达到了0.700和0.112,与精度最低的优化土壤调节植被指数(OSAVI)模型相比,r2提高了0.071,RMSE降低了0.013。小麦白粉病"红边"蓝移现象并不明显;五种模型r2都达到了0.6以上,说明高光谱技术都能够有效地对冬小麦白粉病病情指数进行无损、快速、精确的反演,其中TVI的反演精度最佳。     

灌溉策略对冬小麦水分利用和生长的影响

为探究不同灌溉策略下冬小麦水分利用和生长的情况,在总灌溉量相同的前提下设置拔节水+开花水单次参比蒸散30%灌溉(W1)、拔节水+开花水单次参比蒸散60%灌溉(W2)和拔节水+开花水大水漫灌(W3)3种灌溉策略,利用称重式蒸渗仪和diviner 2000研究了不同灌溉策略下冬小麦的耗水动态、蒸散特征和水分利用效率。结果表明,大水漫灌处理(W3)下冬小麦主要利用上层(0～50 cm)土壤水分,而低速率灌溉(W2和W1)处理增强了植株根系对深层(70～100 cm)土壤水分的吸收;同时,低速率灌溉可以降低蒸散速率,W3、W2和W1的日蒸散速率最大值在拔节水灌溉期间分别为13.20、10.82和10.58 mm·d^-1,在开花水灌溉期间分别为15.10、10.57和9.10 mm·d^-1,其中低速率灌溉主要降低了单日蒸散的午间高峰值,减少了无效耗水。大水漫灌处理不利于生长后期株高的增加,而低速率灌溉不仅有利于株高的形成,也有利于叶片维持较高水平且稳定的SPAD值,保证了籽粒灌浆,使得W2处理的穗粒数和千粒重较W3处理分别提高7.25%和3.93%。综合来看,低速率灌溉策略通过低量持续的供水改变了冬小麦植株根系对土壤水利用的层次,减少无效水蒸散,维持叶片稳定的光合能力,提高了产量和水分利用效率。     

Short-term daily forecasting of crop evapotranspiration of rice using public weather forecasts

Accurate forecasts of daily crop evapotranspiration (ET_c) are essential for real-time irrigation management and water resource allocation. This paper presents a method for the short-term forecasting of ET_c using a single-crop coefficient approach and public weather forecasts. Temperature forecasts with a 7-day lead time in 2013–2015 were retrieved and entered into a calibrated Hargreaves–Samani model to compute daily reference evapotranspiration (ET₀) forecasts, while crop coefficient (K_c) empirical values were estimated from both observed ET_c value and calculated ET₀ values using the Penman–Monteith equation for the period of 2010–2012. Daily ET_c forecasts of irrigated double-cropping rice were determined for three growing seasons during the period of 2013–2015 and were compared with ET_c values measured by the weighing lysimeters at the Jiangxi experimental irrigation station in southeastern China. During the early rice season, the average mean absolute error (MAE) and root-mean-square-error (RMSE) values of ET_c forecasts ranged from 0.95 to 1.06 mm day^?1 and from 1.18 to 1.31 mm day^?1, respectively, and the average correlation coefficient (R) ranged from 0.39 to 0.54; for late rice, the average MAE and RMSE values ranged from 1.01 to 1.09 mm day^?1 and from 1.32 to 1.40 mm day^?1, respectively, and the average R value ranged from 0.54 to 0.58. There could be three factors responsible for errors in ET_c forecasts, including temperature forecast errors, K_c value errors and neglected meteorological variables in the HS model, including wind speed and relative humidity. In addition, ET_c was more sensitive to changes in temperature than K_c. The overall results indicated that it is appropriate to forecast ET_c with the proposed model for real-time irrigation management and water resource allocation.     

Biomass production and nutrient removal by tropical grasses subsurface drip‐irrigated with dairy effluent

Effluent lagoons on dairy farms can overflow and potentially pollute adjacent land and associated water bodies. An alternative solution to effluent disposal is needed by dairy operators in island environments. An attractive win‐win alternative is to recycle nutrients from this resource through effluent irrigation for forage grass production that minimizes environmental pollution. This study assessed biomass production and nutrient removal by, and high application rates to, tropical grasses that were subsurface drip‐irrigated with dairy effluent. Four grass species – Banagrass (Pennisetum purpureum K. Schumach.), California grass (Brachiaria mutica (Forssk.) Stapf.), Stargrass (Cynodon nlemfuensis Vanderyst) and Suerte grass (Paspalum atratum Swallen) – were subsurface (20–25 cm) drip‐irrigated with effluent at two rates based on potential evapotranspiration (ET_p) at the site (Waianae, Hawaii) ?2·0 ET_p (16 mm d^?1 in winter; 23 mm d^?1 in summer) and 0·5 ET_p (5 mm d^?1 in winter; 6 mm d^?1 in summer). Treatments were arranged in an augmented completely randomized design. Brachiaria mutica and P. purpureum had the highest dry‐matter yield (43–57 t ha^?1 year^?1) and nutrient uptake especially with the 2·0 ET_p irrigation rate (1083–1405 kg ha^?1 year^?1 N, 154–164 kg ha^?1 year^?1 P, 1992–2141 kg ha^?1 year^?1 K). Average removal of nutrients by the grasses was 25–94% of the applied nitrogen, 11–82% of phosphorus and 2–13% of the potassium. Average values of crude protein (90–160 g kg^?1), neutral detergent fibre (570–620 g kg^?1) and acid detergent fibre (320–360 g kg^?1) were at levels acceptable for feeding to lactating cattle. Results suggest that P. purpureum and B. mutica irrigated with effluent effectively recycled nutrients in the milk production system.     

Modelling the nitrogen-driven trade-off between nitrogen utilisation efficiency and water use efficiency of wheat in eastern Australia

The nitrogen-driven trade-off between nitrogen utilisation efficiency (yield per unit nitrogen uptake) and water use efficiency (yield per unit evapotranspiration) is widespread and results from well established, multiple effects of nitrogen availability on the water, carbon and nitrogen economy of crops. Here we used a crop model (APSIM) to simulate the yield, evapotranspiration, soil evaporation and nitrogen uptake of wheat, and analysed yield responses to water, nitrogen and climate using a framework analogous to the rate-duration model of determinate growth. The relationship between modelled grain yield (Y) and evapotranspiration (ET) was fitted to a linear-plateau function to derive three parameters: maximum yield (Y_max), the ET break-point when yield reaches its maximum (ET^#), and the rate of yield response in the linear phase (ΔY/ΔET). Against this framework, we tested the hypothesis that nitrogen deficit reduces maximum yield by reducing both the rate (ΔY/ΔET) and the range of yield response to evapotranspiration, i.e. ET^# − E_s, where E_s is modelled median soil evaporation.     

县域冬小麦生物量动态变化遥感估测研究

为给生产管理中及时掌握县域冬小麦长势的动态变化提供有效手段,以江苏省沭阳县为研究区,基于冬小麦生物量形成的生理生态过程,重构冬小麦生物量遥感估测模型。选用两景不同时相的HJ星影像数据,利用植被指数反演的LAI数据,对冬小麦生物量模型进行参数修订,并对县域冬小麦拔节期生物量的空间分布进行估测。在此基础上,进一步估测冬小麦抽穗期生物量分布特征及其动态变化特点。结果表明：（1）冬小麦拔节期生物量估测值和观测值范围分别为2 054.3~4 828.3 和1 962.5~4 568.4 kg·hm^-2 ,平均值分别为3 148和3 045.5 kg·hm^-2 ,RMSE为214.8 kg·hm^-2 ,决定系数为0.919 1,表明冬小麦生物量模型模拟精度较好;（2）冬小麦抽穗期生物量较拔节期发生明显变化,其中长势变化快的田块面积为20 108.7hm,占总种植面积的23.4%。春季气候因素的转好以及肥水措施的实施对冬小麦营养与生殖共生阶段的生长起到明显促进作用。说明本研究提出的基于遥感反演信息与生长模型协同的冬小麦生物量估测方法能有效估测县域冬小麦不同生长时期生物量的空间分布及其动态变化。     

Estimated evapotranspiration of rice based on pan evaporation as a surrogate to lysimeter measurement

Reference crop evapotranspiration (ET _o), used to determine actual crop evapotranspiration, is often estimated from pan evaporation (EP) data. However, uncertainties in the relationship between ET _o and EP often result in unreliable estimate of crop evapotranspiration. This study investigated the relationship between measured and estimated crop evapotranspirations, ET _m and ET _e, respectively, at tillering (9–30 days after transplanting, DAT) and mid-growth (51–72 DAT) stages of a rice variety. ET _m was measured with a Marriott Tube-type Micro-lysimeter (hereafter referred to Micro-lysimeter) in a ponded rice field and ET _e was estimated from EP, which was measured by employing the US Weather Bureau Class ‘A’ Evaporation Pan (hereafter referred to Class A Evaporation Pan). A strong linear relation (r ² = 0.89) at the tillering stage and a weak relation (r ² = 0.48) at the mid-growth stage were obtained between ET _m and EP. The slope of this plot provided a pan-crop factor (K _p K _c), which was 0.81 at the tillering stage and 0.79 at the mid-growth stage. The ET _e versus ET _m relationship was also strongly linear (r ² = 0.90) at the tillering stage but weakly linear (r ² = 0.50) at the mid-growth stage. The pan-based method thus provided reliable estimates of evapotranspiration during the tillering stage of rice.     

基于多源遥感数据和机器学习算法的冬小麦产量预测研究

为探讨基于多源遥感数据和机器学习算法预测冬小麦产量的可行性,利用中麦175/轮选987重组自交系F₇代群体中70个家系开展田间试验,通过无人机遥感平台和地面表型车平台及手持式冠层鉴定平台,获取冬小麦灌浆期光谱数据,分别用4种机器学习方法和集成方法建立产量预测模型。结果表明,在61个光谱指数中,除MCARI、DSI、PVI外,其余指数均与产量显著相关或极显著相关,700 nm和800 nm组合的高光谱指数能够比较准确地预测产量。相对于高光谱和多光谱,RGB传感器预测产量精度最高,平均决定系数（r²）为0.74,平均均方根误差（RMSE）为517.78 kg·hm_-2。相对于决策树（DT）、随机森林（RF）、支持向量机（SVM）三种传统机器学习算法,岭回归（RR）算法预测产量的精度最高,平均r²为0.73,平均RMSE为516.1 kg·hm_-2。与单一的传统机器学习算法相比,DT、RF、SVM、RR结合集成算法的预测精度高且稳定,r²高达0.77,RMSE也较低。SVM 、RF、DT、RR四种机器学习算法和RGB、ASD、UAV、UGV四个传感器构成的算法-传感器集成方法的预测精度提升,r²为0.79,RMSE降至469.98 kg·hm_-2。因此,利用Stacking集成方法将不同算法、传感器进行结合,能够有效地提高冬小麦产量预测精度。     

冬小麦不同冠层叶片光合蒸腾和水分利用效率变化特征及对灌溉的响应

为了解冬小麦不同冠层叶片光合和蒸腾作用特征以及随水分条件的变化规律,通过田间试验,以冬小麦京冬22为试验材料,设置0 mm（T₀）、220 mm（T₁）、280 mm（T₂）3种水分处理,比较分析了冬小麦不同冠层叶片净光合速率、蒸腾速率及水分利用效率对光合有效辐射和灌溉响应的差异。结果表明,三种水分处理下,冬小麦不同冠层叶片的蒸腾速率和光合速率随光合有效辐射的增加而增加,随后趋于平缓。不同冠层叶片蒸腾速率、光合速率对光合有效辐射的响应表现为上层>中层>下层;不同冠层叶片WUE对光合有效辐射的响应表现为上中层>下层。光合有效辐射相同时,灌水处理（T₁和T₂）的叶片光合蒸腾速率均高于不灌水处理（T₀）。T₁、T₂处理下,叶片光合速率对光合有效辐射响应在整个生育期内表现为灌浆期>抽穗期>成熟期>拔节期,蒸腾速率对光合有效辐射响应在整个生育期内表现为抽穗期>灌浆期>拔节期>成熟期;T₀处理下,叶片光合速率对光合有效辐射响应在整个生育期内表现为灌浆期>拔节期>抽穗期>成熟期,蒸腾速率对光合有效辐射响应在冬小麦整个生育期内表现不显著。因此,在进行小麦叶片到单株光合蒸腾尺度拓展估算时,应考虑冠层位置和水分条件对拓展结果的影响。     

The effect of water stress in regulated deficit irrigation on soybean yield (Glycine max [L.] Merr.)

The objective of this research was to investigate the effect of water stress in regulated deficit irrigation (RDI) on the yield of soybean growing on Ultisol soil. This research was conducted under plastic house on the experimental farm of Lampung Polytechnique from August to November 2004. The water stress treatments in regulated deficit irrigation were ET₁ (1.0 × ET_c), ET₂ (0.8 × ET_c), ET₃ (0.6 × ET_c), ET₄ (0.4 × ET_c) and ET₅ (0.2 × ET_c), arranged in a randomized block design with four replications. ET_c means crop evapotranspiration under standard condition, which was well watered. For example, the ET₂ (0.8 × ET_c) treatment means that the amount of supplied water per a day is the same as the crop adjustment evapotranspiration (ET_cadj) with the value 0.8 of water stress coefficient (K _s). The RDI treatments were carried out just at vegetative phase and its treatments were stopped at the beginning of flowering phase, and afterwards the treatments were watered at 1.0 × ET_c. The results showed that since week II, the soybean experienced stress throughout the growth period except ET₂ treatment. ET₂ treatment started to be stressed at week V and continued to be stressed until the harvest time. At the ET₃ treatment, the critical water content (θ_c) of soybean was reached at week II, and the θ_c was 0.24 m³/m³ on the average. The RDI at vegetative period significantly affected the yield. The highest yield was ET₁ (35.2 g/plant), followed by ET₂ (31.0 g/plant), ET₃ (18.1 g/plant), ET₄ (7.6 g/plant), and ET₅ (3.3 g/plant). The optimal water management of soybean with the highest yield efficiency was regulated deficit irrigation with water stress coefficient (K _s) of 0.80 for vegetative phase.     

基于无人机可见光影像与生理指标的小麦估产模型研究

为及时、准确地掌握小麦产量动态信息,基于无人机遥感平台,分别分析了小麦4项生理指标［地面实测叶面积指数、叶片含氮量、叶片含水量及叶片叶绿素相对含量（SPAD值）］及10项植被指数与产量的相关性,以筛选出与产量最为敏感的生理指标与植被指数,并比较了3种建模方法（一元回归UR、多元逐步回归SMLR和主成分回归PCAR）在小麦各生育时期估产的适用性,进而得到小麦最优估产模型。结果表明：（1）不同生育时期两类变量与产量的相关性变化特征一致,均表现为抽穗期>灌浆期>成熟期;不同生理指标、植被指数与产量的相关性在各生育时期均存在差异,生理指标表现为叶片含氮量>LAI>SPAD>叶片含水量;而植被指数在各时期表现不同;（2）以生理指标与植被指数为自变量,采用SMLR模型构建的抽穗期估产模型拟合精度最高,R、RMSE和nRMSE分别为0.828、362.53 kg·hm^-2和12.35%;（3）小麦估产模型在各生育时期的预测精度表现为抽穗期>灌浆期>成熟期。