基于基因表达式编程算法的参考作物腾发量模拟计算

选取都安气象站5年(2008—2012年)的逐日气象数据,包括日最高气温、最低气温、平均风速、日照时数以及相对湿度5个气象要素的不同组合作为输入,并以FAO-56 Penman-Monteith法(FAO P-M)的计算结果作为标准值,采用基因表达式编程算法(GEP)及径向基函数网络算法(RBFNN)对参考作物腾发量ETo进行模拟计算,并将模拟结果与Hargreaves模型的计算结果进行比较,用决定系数R2和均方根误差RMSE作为评价指标。结果表明,GEP模型能够捕捉到ETo的变化,具有较强的适用性,与FAO P-M公式的计算值有很高的一致性。引入关键气象因子(气温和相对湿度)后,模型的决定系数R2达到0.914,均方根误差RMSE为0.240 mm/d。在相同输入情况下GEP模型计算精度高于RBFNN模型和Hargreaves模型,并建立了可以替代Hargreaves模型的GEP模型及缺少相对湿度RH时的GEP模型。结果表明,在缺乏相关气象因子时,可以利用GEP模型模拟ETo。     

基于气温预报的参考作物腾发量预报方法比较

为探索精确预报未来短期ET0的方法,比较了4种基于气温预报ET0预报模型,即Hargreaves-Samani(HS)、Thornthwaite(TH)、简化的Penman-Monteith(PT)及McCloud(MC)模型。收集了西藏林芝站2001年1月1日至2013年12月31日的实测逐日气象数据和2012年6月6日至2013年12月31日逐日对未来7d的气象预报数据,在气温预报精度评价的基础上,采用4种基于温度的参考腾发量计算模型直接进行ET0预报,然后采用率定后的模型进行ET0预报,最后与实测气象数据和FAO-56PM公式计算的ET0值进行比较。结果表明,未率定的4种模型预报误差均较大,其中PT公式精度稍高。经率定后,4种公式的预报精度都有所提高,平均准确率为70%,MAE值HS模型最小,平均为0.57mm/d,其他3个模型为1.27~1.50mm/d;RMSE都在2.0mm/d左右;r值总体仍不高,TH模型平均仅有0.19,其他3种模型在0.6左右。综合来看,PT模型的预报效果稳定性优于其他3个模型。对于林芝地区附近的灌区,无论有无气象观测数据供模型率定,建议采用PT模型进行ET0预报。     

青海东部农业区参考作物蒸散量的估算方法

利用青海东部农业区5个气象站1960—2006年逐日气象资料,以Penman-Monteith公式估算结果为标准,分析了Hargreaves-Samani、McCloud与Priestley-Taylor法的适用性。结果表明,采用Hargreaves-Samani法和Priestley-Taylor法,年平均ET0估算值高于标准值,而McCloud估算结果显著偏小;气温较低月份(11月至次年3月),Hargreaves-Samani和Priestley-Taylor法月平均ET0估算值与标准值差异不显著,4—10月估算结果显著高于标准值。可见,Priestley-Taylor适用性最好,Hargreaves-Samani次之,McCloud最差,且Priestley-Taylor公式修正后估算精度更高。     

基于改进Makkink模型的四川盆地参考作物蒸散量估算

为了有效提高四川盆地参考作物蒸散量ET₀的预报精度,选取四川盆地16个代表性气象站点1961-2019年逐日气象数据,基于差分进化算法(DE)对辐射模型的经验参数校准改进Makkink模型并估算四川盆地ET₀,在日、月尺度上对改进的Makkink模型(M1-M6)和Jennsen-Haise(JH)及Irmak(IK)模型评价.结果表明:在日尺度上,改进的Makkink(M1-M6)模型(R²为0.77~0.87)模拟结果比JH和IK模型(R²为0.74~0.76)更精确,改进的Makkink模型中,M4模型估算精度最高,综合性指标GPI中位数为1.05;在月尺度上,改进的Makkink模型模拟结果(误差为3.59~15.71 mm/月)也优于JH和IK模型(误差为6.84~25.31 mm/月),其中M4模型估算精度最佳,综合性指标GPI为1.72.总体而言,推荐以温度和相对湿度作为输入数据的M4模型模拟四川盆地ET₀.     

基于MEA-BPNN的西北旱区参考作物蒸散量预报模型

为有效提高西北旱区参考作物蒸散量(Reference crop evapotranspiration,ET0)预报精度,在西北旱区选择5个代表性气象站点,构建10种基于思维进化算法(Mind evolutionary algorithm,MEA)优化的误差反向传波神经网络(Back propagation neural network,BPNN)ET0预报模型,并将其与Hargreaves-Samani模型、Irmak模型和48-PM模型等3种在西北旱区ET0计算精度较高的模型进行比较。结果表明:在不同输入的情况下MEA-BPNN模型模拟精度具有相对较高水平,其中MEA-BPNN1(输入最高气温Tmax、最低气温Tmin、相对湿度RH、日照时数n和距地面2 m高处的风速u2)、MEA-BPNN2(输入Tmax、Tmin、n和u2)及MEA-BPNN3(输入Tmax、Tmin、RH和u2)模型的R2、NSE均大于0.96,RMSE、MAE也分别小于0.34、0.25 mm/d,以上3种MEA-BPNN模型的整体评价指标(Global performance indicator,GPI)排名分别为1、2、3;MEA-BPNN7(输入Tmax、Tmin和u2)的R2、NSE分别为0.966 2、0.962 2,RMSE、MAE分别为0.361 0、0.276 1 mm/d,模拟精度较高;MEA-BPNN模型可移植性的分析表明:MEA-BPNN模型在西北旱区具有较强的泛化能力,基于不同站点数据构建的预报模型也有较高精度;在相同输入情况下MEA-BPNN模型模拟精度均高于Hargreaves-Samani模型、Irmak模型和48-PM模型。因此,在气象资料缺乏情景下MEA-BPNN模型可作为西北旱区ET0计算的推荐模型,可为实时精准灌溉预报的实现提供科学依据。     

基于气温预报和HS公式的不同生育期参考作物腾发量预报

根据南京站2001-2011年实测气象数据,以Penman-Monteith(PM)公式计算得到的参考作物腾发量ET0值作为基准值,对仅需要气温数据计算参考作物腾发量的Hargreaves-Samani(HS)公式进行参数率定,采用率定后的HS公式依据2012年6月-2015年6月气温预报数据对南京水稻、冬小麦不同生育期未来1~7d的ET0进行预报,并与基于实测气象数据的PM法计算的ET0值进行比较,评价HS法的ET0预报精度。结果表明:最低、最高气温实测值与预报值相关系数分别为0.97和0.93,最低气温预报精度略高于最高气温;预见期1~7d内,水稻、冬小麦不同生育期ET0预报值与PM法计算值变化趋势基本一致,整个生育期内冬小麦ET0预报值与PM法计算值吻合程度更好,水稻、冬小麦相关系数分别达0.60、0.80左右;水稻各生育期平均准确率为66.0%~97.5%,平均绝对误差为0.65~1.22mm/d,均方根误差为0.76~1.42mm/d,冬小麦各生育期平均准确率为75.4%~99.5%,平均绝对误差为0.33~1.06mm/d,均方根误差为0.43~1.23mm/d;作物生育期各阶段对气温预报误差越敏感,ET0预报精度越低,随着生育期的推进,水稻对气温预报误差的敏感程度逐渐减小,相应的ET0预报精度逐渐增加,而冬小麦反之;但整体上预见期1~7d的气温预报及ET0预报精度达到可利用程度,可为快速灌溉预报及灌溉决策提供数据支撑。     

参考作物蒸散量计算方法在西宁的适用性分析

利用青海西宁气象站点1970-2016年的逐日气象资料,以Penman-Monteith估算结果作为计算参考作物蒸散量(ET0)的标准,讨论Hargreaves-Samani法、Mc Cloud法与1948 Penman法3种方法在青海省西宁地区的适用性。结果表明:1948 Penman法和Hargreaves-Samani法优于Mc Cloud法;Mc Cloud估算结果显著偏小;Hargreaves-Samani法,在ET0较小时估算值与Penman-Monteith方法计算值相当,在ET0较大时,估算结果显著高于标准值;1948 Penman法与Penman-Monteith法相关系数较高,一致性最好,可以对1948 Penman法的估算值乘以一个常数就能够近似代替PenmanMonteith公式的估算值,提高估算精度。     

基于MEA-BPNN的中国西北旱区参考作物蒸散量预报模型

为有效提高西北旱区参考作物蒸散量(Reference crop evapotranspiration,ET_0)预报精度,在西北旱区选择5个代表性气象站点,构建10种基于思维进化算法(Mind evolutionary algorithm,MEA)优化的误差反向传波神经网络(Back propagation neural network,BPNN)ET_0预报模型,并将其与Hargreaves-Samani模型、Irmak模型和48-PM模型等3种在西北旱区ET_0计算精度较高的模型进行比较。结果表明:在不同输入的情况下MEA-BPNN模型模拟精度具有相对较高水平,其中MEA-BPNN1(输入最高气温T_(max)、最低气温T_(min)、相对湿度R_H、日照时数n和距地面两米高处的风速u_2)、MEABPNN2(输入T_(max)、T_(min)、n和u_2)及MEA-BPNN3(输入T_(max)、T_(min)、R_H和u_2)模型的R~2、NSE均大于0.96,RMSE、MAE也分别小于0.34、0.25 mm/d,以上3种MEA-BPNN模型的整体评价指标(Global performance indicator,GPI)排名分别为1、2、3;MEA-BPNN7(输入T_(max)、T_(min)和u_2)的R~2、NSE分别为0.966 2、0.962 2,RMSE、MAE分别为0.3610、0.276 1 mm/d,模拟精度较高;MEA-BPNN模型可移植性的分析表明:MEA-BPNN模型在西北旱区具有较强的泛化能力,基于不同站点数据构建的预报模型也有较高精度;在相同输入的情况下MEA-BPNN模型模拟精度均高于Hargreaves-Samani模型、Irmak模型和48-PM模型。因此,在气象资料缺乏情景下MEA-BPNN模型可作为西北旱区ET_0计算的推荐模型,可为实时精准灌溉预报的实现提供科学依据。     

MARS模型在渭河流域参考作物蒸散量计算中的适应性研究

为有效提高气象资料缺失时渭河流域参考作物蒸散量(ET₀)计算精度,选取流域及附近20个气象站58 a (1960-2017年)逐日气象资料,基于不同气象要素组合,构建16种基于多元自适应回归样条(MARS)的ET₀计算模型,并将计算结果与Hargreaves-Samani、Makkink和Irmark-Allen模型进行对比,评价MARS模型在渭河流域的适应性及可移植性。结果表明：MARS模型能很好地甄别ET₀与各输入因子间的非线性关系,MARS2 (T_max、T_min、R_a)计算精度(平均MAE为0.225 mm/d,平均RMSE为0.327 mm/d,平均R²为0.897)能满足应用要求,模型精度随输入气象要素数量的增加而升高;输入因子中引入地球外辐射R_a,可明显提高MARS模型精度;在输入因子相同时,MARS模型精度高于Hargreaves-Samani、Irmark-Allen和Makkink模型;MAR...     

西藏高原灌区参考作物蒸散量模型的适用性研究

为明确参考作物蒸散量(ET_0)计算模型在西藏高原灌区的适用性,推荐适宜于气象资料短缺条件下的ET_0计算模型,本研究基于满拉、墨达、江北3个灌区的气象站的长系列数据,以FAO推荐的Penman-Monteith(FAO 56 PM)模型计算的ET_0为标准,对ET_0的5种常用计算模型的适用性进行评价。结果表明:Makkink、Irmark-Allen、FAO 17Penman、Hargreaves-Samani和Priestley-Taylor 5种模型模拟的日尺度ET_0变化趋势与FAO 56 PM模型一致,在年际间均呈先增后减的变化规律,且峰值出现在6～7月份,但各模型适用性存在显著差异。Makkink模型的日尺度MAE、RMSE、NSE值分别为0.37 mm/d、0.45 mm/d和0.84,模拟精度及可信度最高;Irmark-Allen模型次之,MAE、RMSE、NSE分别为0.65 mm/d、0.71 mm/d、0.62;Priestley-Taylor模型最差,MAE值最大达4.91 mm/d且NSE值小于0。年尺度下,各模型较FAO 56 PM均存在高估现象,其中FAO 17Penman、Hargreaves-Samani、Priestley-Taylor模型的NSE值介于-3 571.76～-118.00之间,模拟结果不可信;Makkink模型的NSE值最接近于0,模拟结果可信,但模拟过程的误差较大。综合评定,推荐Makkink为西藏高原灌区气象资料短缺条件下的ET_0简化模型。     

Comparison of artificial neural network models and empirical and semi-empirical equations for daily reference evapotranspiration estimation in the Basque Country (Northern Spain)

Reference evapotranspiration (ETo) determination is a key factor for water balance and irrigation scheduling. Evapotranspiration can be measured directly by high-cost micrometeorological techniques, or estimated by mathematical models. The combination equation of Penman–Monteith, modified by Allen et al. [Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop evapotranspiration. Guidelines for computing crop water requirements. FAO Irrigation and Drainage, Paper no. 56. FAO, Rome] (PM56), is the reference equation for ETo estimation. This method is also appropriate for the calibration of other ETo estimation equations. The utilization of these calibrated ETo equations is recommended in the absence of data of any of the meteorological parameters necessary for the application of PM56. In addition to the use of classic ETo equations, the adoption of artificial neural network (ANN) models for the estimation of daily ETo has been evaluated in this study. ANNs are mathematical models, whose architecture has been inspired by biological neural networks. They are highly appropriate for the modelling of non-linear processes, which is the case of the evapotranspiration process. Seven ANNs (with different input combinations) have been implemented and compared with ten locally calibrated empirical and semi-empirical ETo equations and variants of these equations (with estimated meteorological parameters as inputs). The comparisons have been based on statistical error techniques, using PM56 daily ETo values as a reference. ANNs have obtained better results than the locally calibrated ETo equations in the three groups of evaluated methods: temperature and/or relative humidity-based methods (0.385 mm d⁻¹ of root mean square error (RMSE)), solar radiation-based methods (0.238 mm d⁻¹ of RMSE), and methods based on similar requirements to those of PM56 except for the estimation of solar radiation and/or relative humidity (0.285 mm d⁻¹ of RMSE).     

Comparative study of conventional and artificial neural network-based ETo estimation models

Accurate estimation of reference crop evapotranspiration (ETo) is required for several hydrological studies and thus, in the past, a number of ETo estimation methods have been developed with different degree of complexity and data requirement. The present study was carried out to develop artificial neural network (ANN) based reference crop evapotranspiration models corresponding to the ASCE’s best ranking conventional ETo estimation methods (Jensen et al. ASCE Manual and Rep. on Engrg. Pract. no. 70, 1990). Among the radiation methods, FAO-24 radiation (or Rad) method for arid and Turc method for humid region, and among the temperature methods, FAO-24 Blaney–Criddle (or BC) method were studied. The ANN architectures corresponding to the above three less data-intensive methods were developed for four CIMIS (California Irrigation Management Information System) stations, namely, Davis, Castroville, Mulberry, and West Side Field station. The comprehensive ANN architecture developed by Kumar et al. (J Irrig Drain Eng 128(4):224–233, 2002) corresponding to Penman–Monteith (PM) ETo for Davis was also tried for the other three stations. Daily meteorological data for a period of more than 10 years (01 January 1990 to 30 June 2000) were collected from these stations and were used to train, test, and validate the ANN models. Two learning schemes, namely, standard back-propagation with learning rate of 0.2 and standard back-propagation with momentum having learning rate of 0.2 and momentum term of 0.95 were considered. ETo estimation performance of the ANN models was compared with the FAO-56 PM method. It was found that the ANN models gave better closeness to FAO-56 PM ETo than the best ranking method in each category (radiation and temperature). Thus these models can be used for ETo estimation in agreement with climatic data availability, when not all required climatic variables are observed.     

玉米农田生态系统蒸散发模型参数优化

[目的]在无法根据实测值得到具体模型参数的地域,对经验参数进行优化以提高区域蒸散发模型的精度.[方法]通过黑河流域生态水文过程综合遥感试验水文气象观测数据集中的大满超级站气象要素梯度观测系统的数据,研究玉米农田生态系统的蒸散发模型优化问题.采用差分进化自适应算法,以潜热通量和感热通量为优化目标,引入能量闭合因子对模型参...     

参考作物蒸发蒸腾量计算方法的应用比较

应用国家"863"节水农业重大专项子课题示范现场的气象资料,对4种分属于不同类型的参考作物蒸发蒸腾量ET0计算公式进行了日ET0值的验证计算。结果显示,在时间序列上,随气象因素变化各方法计算的日ET0值呈相同的变化趋势,但计算值有较大的差异;选取FAO56Penman-Monteith公式计算结果为标准,Priestley-Taylor(1972)方法结果与之最为接近,其余依次是Irmark-Allen拟合法和Hargreaves-Samani(1985)法;不同天气类型条件下,Priestley-Taylor(1972)结果与FAO56Penman-Monteith有较高的一致性,而其他2种方法随n/N的减小,误差急剧增加,尤其是Hargreaves-Samani(1985)方法。     

Irrigation requirements and transpiration coupling to the atmosphere of a citrus orchard in Southern Brazil

Crop evapotranspiration (ETc) was measured as evaporative heat flux from an irrigated acid lime orchard (Citrus latifolia Tanaka) using the aerodynamic method. Crop transpiration (T) was determined by a stem heat balance method. The irrigation requirements were determined by comparing the orchard evapotranspiration (ETc) and T with the reference evapotranspiration (ETo) derived from the Penman-Monteith equation, and the irrigation requirements were expressed as ETc/ETo (Kc) and T/ETo (Kcb) ratios. The influence of inter-row vegetation on the ETc was analyzed because the measurements were taken during the summer and winter, which are periods with different regional soil water content. In this study, the average Kc values obtained were 0.65 and 0.24 for the summer and winter, respectively. The strong coupling of citrus trees to the atmosphere and the sensitivity of citrus plants to large vapor pressure deficits and air/leaf temperatures caused variations in the Kcb in relation to the ETo ranges. During the summer, the Kcb value ranged from 0.34 when the ETo exceeded 5 mm d⁻¹ to 0.46 when the ETo was less than 3 mm d⁻¹.     

江苏地区不同参考作物蒸发蒸腾量估算模型

为了研究不同参考作物蒸发蒸腾量ET₀估算方法在江苏地区的适用性,收集了江苏省徐州市、高邮市和昆山市1957年1月至2019年12月的气象数据,采用12种不同模型估算了各站点的ET₀,其中模型Priestly-Taylor,Hansen,Jensen-Haise,Makkink是基于辐射数据的模型;MC-Cloud,1985 Hargreaves,Thornthwaite是基于温度数据的;Copais,Valiantzas 1和Valiantzas 2是综合法模型;XGBoost和SVM是机器学习模型.12种ET₀的估算模型计算值分别与Penman-Monteith模型(PM)计算值进行比较,结果表明:各站点的综合评价指数GPI最高的为机器学习模型中的SVM模型;在输入参数相同的情况下,机器学习模型模拟精度优于综合法和温度法以及辐射法中的Pristley-Taylor和Makkink模型;机器学习模型随着输入参数减少,模拟精度依次降低.研究结果可以为江苏地区气象数据不完善时估算ET₀提供科学依据.     

A comparison of the Priestley-Taylor and Penman methods for estimating reference crop evapotranspiration in tropical countries

An equation for Potential Evaporation (PE) proposed by Priestley and Taylor in 1972 has fewer data requirements than the well established Penman Potential Transpiration (Et) equation. From their definitions, PE and Et values should both provide acceptable estimates of Reference Crop Evapotranspiration (ETo), as defined by Doorenbos and Pruitt. Analysis of mean monthly climatic data from 30 tropical stations, widely spread within the latitude zone 25°N to 25°S, showed that PE and Et estimates agreed closely when monthly rainfall exceeded monthly Et. The minimum data requirements for the Priestley-Taylor equation are daily net radiation and mean air temperature. The Penman equation additionally requires daily data for humidity and run of wind. As reliable field net radiometers become more widely available, the Priestley-Taylor PE equation offers a satisfactory alternative to the Penman Et equation for estimating ETo in humid tropical climates.     

基于神经网络算法的果树需水预测研究

【目的】准确预测果树需水量。【方法】对采集地果园环境数据进行主成分分析,筛选出影响果树蒸腾量的关键因子。建立以长短时记忆(LSTM)神经网络为基础的预测模型来预测果树蒸腾量。为提高预测的精度,在LSTM神经网络的基础上加入了注意力(Attention)机制,形成Attention-LSTM预测模型。【结果】将改进的模型与其他模型的预测精度进行对比,仿真试验表明,该模型的预测精度最高,RMSE和MSE分别为0.487和0.062。【结论】该预测模型可以准确预测果树蒸腾量,从而实现果园精准灌溉并提高水果产量,具有一定的实际意义。     

基于分形理论的参考作物蒸发蒸腾量估算

根据成都、雅安和乐山等站点实测的1954~2000年的气象资料和地理参数,建立了温度推算辐射的模型,从分形理论的角度,进一步揭示了温度和辐射的关系特征,并结合Penman-Monteith模型对参考作物蒸发蒸腾量(ET0)进行了估算。研究表明,温度(T)、净辐射(Rn)和净短波辐射(Qns)无标度区拐点的界限时间和分形维数非常接近,具有相似的分形特征,可利用T估算Rn和Qns;通过T估算的Rn计算的ET0的平均绝对偏差和平均相对偏差,均小于通过T估算的Qns计算的ET0的平均绝对偏差和平均相对偏差,而且偏差较小。因此,在实际应用中可通过T估算Rn来计算ET0,简化ET0的计算过程。     

四川省不同SPEI指数计算方法适用性评价

为探究不同参考作物腾发量(ET₀)算法及相应标准化降水蒸散指数(SPEI)在四川省的适用性,针对四川省3个区域(川西高原、川西南山地和川中盆地),利用34个气象站点1967—2016年的气象资料,以Penman-Monteith(PM)法计算的ET₀为标准,对FAO-24Radiation(FAO-Ra)、Priestley-Taylor(PT)、Makkink(MK)、Hargreaves-Samani(HS)、Blaney-Criddle(BC)、World Meteorological Organization(WMO)、Rohwer(Ro)7种方法的ET₀计算结果进行比较,并选取其中综合表现较好的3种方法进行相应的SPEI计算。通过时间序列分析、误差分析、K-S检验及小波分析等方法,探讨各区域不同ET₀算法下的SPEI适用性。结果表明：7种方法在不同区域计算精度差异显著,在川西高原及川西南山地,PT法均方根误差(RMSE)均在99.11 mm以下,大部分气象站点的相对误差(RE)介于-3.8...