Research On Water Requirement Regular And Water Consumption Model Of Greenhouse Tomato Under Different Ventilation And Water Control

The study of water requirement regular and water consumption model of greenhouse crops under different ventilation and water control has important theoretical significance and practical value for realization of scientific environmental management of greenhouse crops,establishment of reasonable ventilation and irrigation regime,and development of efficient water saving in facility agriculture.This study took drip irrigation tomato in solar greenhouse in North China as the research object,and which was carried out at the Xinxiang Comprehensive Experimental Base of the Chinese Academy of Agricultural Sciences from March to July 2020-2021.Six treatments were set up in the experiment:two ventilation treatments were set up by opening the air vents at different positions of the greenhouse（T₁:opening the north window and top window;T₂:opening the north window,top window and south window）;Three irrigation treatments（W₁:0.9E_p,W₂:0.7E_p,W₃:0.5E_p）were set up with reference to the cumulative evaporation capacity（E_p）of the 20 cm standard evaporation dish.The meteorological,soil water and heat environment,tomato growth and water consumption index in the greenhouse were observed and analyzed.Crop coefficient（Kc’）and pan coefficient(K_pan)were used to investigate the water requirement characteristics of greenhouse tomato,and the resistance parameters（r_c and r_a）in PM model were improved.Finally,the effects of daily evapotranspiration（ET）and meteorological data(net solar radiation（Rn）,mean temperature（Ta）,minimum temperature(Ta_min),maximum temperature(Ta_max),relative humidity（RH）,minimum relative humidity(RH_min),maximum relative humidity(RH_max),and wind speed（V）)on ET in the greenhouse were studied.In addition,eight machine learning algorithms（Linear Regression（LR）,Support vector Regression（SVR）,K-Neighbor Regression（KNR）,Random Forest Regression（RFR）,Ada Boost Regression（ABR）,Bagging Regression（BR）,and Gradient Boosting Regression（GBR）were used to predict ET of greenhouse tomato under T₁W₁treatment,and universality evaluation was carried out for other treatments.The results are as follows:1）In the greenhouse,Ta showed T₁>T₂,and the range of indoor RH was the largest in cloudy days,while outdoor RH appeared in sunny days.Indoor wind speed（V）fluctuated little in seedling period,flower and fruit period,but fluctuated greatly in full fruit period and picking period.On the whole,water vapor pressure difference（VPD）at 2 m above the ground was larger than that at 30 cm above the canopy.Both indoor and outdoor solar radiation（Rs）reached the maximum in the full fruit period,and the correlation degree of meteorological factors inside and outside the greenhouse was from large to small:Rs>Ta>RH>V.The amplitude of temperature fluctuation decreased with the increase of soil depth,and the variation range of soil volume water content increased first and then decreases.The soil volume water content was W₃>W₁at 0-10cm depth.2)The water consumption of greenhouse tomato was influenced by both ventilation and irrigation.The proportion of total water consumption in each growth period was as follows（from large to small）:full fruit period,flower and fruit period,picking period,seedling period.Both ventilation and irrigation treatment affected the leaf area index（LAI）of greenhouse tomatoes,and ventilation had a significant effect on plant height（P<0.05）,while irrigation had more significant effect on stem flow rate than ventilation（P<0.05）.The influence of solar radiation and water vapor pressure difference on transpiration rate reached an extremely significant level（P<0.01）.The yield of T₂W₁ in two years was the highest（147.88 t/hm²）,and the yield of all treatments was significant（P<0.05）.With the decrease of irrigation amount,tomato yield,fruit quality and irrigation water compensation rate(I_rc)in greenhouse all showed a decreasing trend,while water use efficiency（WUE）and irrigation water use efficiency（IWUE）showed an increasing trend.3)The correlation between evapotranspiration of reference crops(ET₀’_-T1,ET₀’_-T2)and surface evaporation(E_pan-T1,E_pan-T2)was high,and the fitting degree of T₂（R²=0.97,RMSE=0.93 mm/d,MAE=0.83 mm/d）was higher than that of T₁.It could be seen that when ET₀’was calculated in the absence of meteorological data,it could be roughly estimated by using surface evaporation under the same treatment.The average crop coefficient(K_c’_-T1,K_c’_-T2)and pan evaporation coefficient(K_pan-T1,K_pan-T2)higher than the recommended value of FAO at flower and fruit period,full fruit period and picking period.4)In the seedling period,canopy resistance decreased rapidly with the increase of LAI,and in the flower and fruit period and full fruit period,canopy resistance decreased slowly.In two years.Within two years,the simulation accuracy of PM model was PM-r_a-h_s3（mixed convection）,PM-r_a-h_s1（free convection）and PM-r_a-h_s2（forced convection）in descending order,but the improvement of aerodynamic drag coefficient（r_a）would overestimate ET,and PM-r_a-h_s3,PM-r_a-h_s1 and PM-r_a-h_s2overestimated ET by 8.58%,10.44%and 10.73%respectively.5)Rn,Ta,Ta_min and Ta_max were positively correlated with ET,while RH,RH_min,RH_max and V were negatively correlated with ET.Rn is the most correlated with ET,and the correlation coefficient R=0.86,while V is the least correlated with ET,and the correlation coefficient R=0.46.The RMSE,MAE and R² of XGBR were 0.13 mm/d,0.20 mm/d and 0.88,respectively.XGBR was superior to the other 7 models in the accuracy predicting ET.Therefore,the XGBR model could better predict the daily evapotranspiration of tomato during the whole growth period.In two years,the RMSE of XGBR were less than 0.45 mm/d under different treatments;MAE were less than0.53 mm/d;R² were all greater than 0.81.XGBR achieved a high level of accuracy in predicting ET under different ventilation and water control.The characteristic importance of the input variables of XGBR was Rn>RH>RH_min>Ta_max>Ta_min>RH_max>Ta>V.When modeling greenhouse drip tomato ET based on XGBR,the selection of Rn,RH,RH_min,Ta_max,Ta_min and RH_max as model input variables,the model validity can be maximized（MSE=0.025 mm/d）.