Results and discussion Directory UMM :Data Elmu:jurnal:A:Agricultural & Forest Meterology:Vol104Issue4Sept2000:

306 W. Luo, J. Goudriaan Agricultural and Forest Meteorology 104 2000 303–313 night was estimated by dividing the weight increment of the blotting paper at the moment of cover removal by the time elapsed since sunset. The total amount of guttation water intercepted by the blotting paper in- side the canopy was calculated as this mean guttation rate multiplied by night length 12 h. Interception of guttation drops is subject to stochas- tic variability which can be estimated from the obser- vation data themselves. Intercepted water is not the same as the true quantity of guttation, and rather it tends to be an underestimate for the following rea- sons. In the first place, only drops beyond a certain size will get detached and fall. Secondly, the vertically projected area of the blotting paper is what counts, and thirdly only water from the leaf area above the blotting paper will be collected. Only if a leaf physi- cally touches the blotting paper, may it directly deliver water, thereby raising the collected amount.

3. Results and discussion

3.1. Effects of cover removal on formation Cutting off nocturnal net radiation loss by shield- ing resulted in a large reduction in dew amount in the rice canopy Fig. 1 because no dew occurred in the Fig. 1. Relation between dew amount mm at sunrise and the time of removal of the screens. The points are mean values of data measured on top leaves during five, five and six nights, for removal at 03:00, 04:00 and 05:00 h, respectively, and for the control without cover. canopy during the covered period. The accumulated dew amount on the top leaves at sunrise after 1–3 h exposure to nocturnal radiative loss was less than half that in the control canopy, which had been exposed to 12 h of nocturnal radiative loss. As expected, dew du- ration on the top leaves with cover was much shorter than that without cover Fig. 2. In Figs. 1 and 2 the error bars refer to the standard error in 150 and 50 data points for uncovered and covered canopies, respectively. On average, the dew duration on top leaves when temporarily covered was 83 of that without cover, measured from the time the cover was removed Fig. 3. On top leaves, dew duration after sunrise ranged between 1.4 and 3.4 h, which was reduced by between 0 and 2 h by shielding. Because of the rapid increase of solar radiation after sunrise, the moment of drying was only slightly delayed for leaves that had been exposed all night. 3.2. Simulation of the dew formation For calculated and observed dew amount and du- ration on top leaves of the rice canopy, the r 2 values were quite high 0.80 and 0.79, respectively as shown in Fig. 4. This indicates that the MICROWEATHER model behaved reasonably well. The remaining stan- W. Luo, J. Goudriaan Agricultural and Forest Meteorology 104 2000 303–313 307 Fig. 2. Relation between daily dew duration h and the time of removal of the screens. The points are mean values of data observed on top leaves during five, five and six nights for removal at 03:00, 04:00 and 05:00 h, respectively, and for the control without cover. Fig. 3. Relation between dew duration hour on the top leaves in the covered canopy and that in the control, in both cases measured from the time of the removal of the screens from the covered canopy. The points are observation data during five, five and six nights for removal at 03:00, 04:00 and 05:00 h, respectively. 308 W. Luo, J. Goudriaan Agricultural and Forest Meteorology 104 2000 303–313 Fig. 4. Relation between measured and simulated a dew amount and b duration on top leaves control: d , covered: s . dard errors were 0.033 mm and 2.33 h for dew amount and dew duration, respectively. To investigate the rea- son of these deviations of the simulation results from the observed ones, a sensitivity analysis was made for the effects of net radiative loss, water vapour pressure deficit VPD and wind speed as the most relevant weather variables for dew formation. The sensitivity analysis results showed that the model results were sensitive first to the nocturnal net radiative loss, sec- ond to water vapour pressure deficit, and third to wind Table 2 Sensitivity analysis of the simulation of dew amount DEW and duration DEWT on top leaves a Date R n,mean b W m − 2 RH max c m s − 1 u mean d Observed weather data R n e ± 10 VPD f ± 10 u g ± 10 DEW mm DEWT h DEW mm DEWT h DEW mm DEWT h DEW mm DEWT h 22–23 February −43.9 0.96 0.70 0.218 12.5 ± 0.042 ±0.3 ± 0.021 −0.1 – +0.3 ±0.019 0 – +0.23 28 February– 1 March − 46.5 0.91 0.72 0.148 12.8 ± 0.049 −0.3 – +0.4 ±0.039 −0.3 – +0.4 ±0.039 −0.2 – +0.3 28–29 March − 38.8 0.90 0.40 0.041 11.5 ± 0.016 −1.0 – +1.7 ±0.014 −1.5 – +2.1 ±0.011 −1.0 – +1.7 2–3 April − 41.5 0.90 0.64 0.063 7.9 ± 0.019 −1.4 – +2.1 ±0.017 −1.8 – +2.3 ±0.010 −1.4 – +1.9 a The amounts are on leaf area basis. b R n,mean is the nightly mean net radiative loss. c RH max is the daily maximum relative humidity. d u mean is the wind speed. e R n is the net radiative loss. f VPD is water vapour pressure deficit. g u is the wind speed. speed Table 2. The smaller sensitivity of the model to wind speed may be attributed to the low wind speed during the dew nights. The dew amount esti- mation error caused by 10 of measurement error in net radiative loss varied from about 20 under clear and wet conditions 22–23 February to about 40 under dry and very quiet conditions 28–29 March. Under dry conditions, the sensitivity of the model to VPD increased and became as large as to net radiative loss. For dew duration, the model was not sensitive W. Luo, J. Goudriaan Agricultural and Forest Meteorology 104 2000 303–313 309 to the three weather variables under clear and wet conditions whereas its sensitivity to the three weather variables greatly increased under dry and very quiet 28–29 March and 2–3 April conditions. The dew duration estimation error caused by 10 of measure- ment error in net radiative loss and VPD could be over 2 h under clear, dry and very quiet conditions. There was a slight tendency for the simulation to underestimate dew amount and duration Fig. 4, prob- ably because of a measurement error in nocturnal net radiative loss. When dew forms on the dome of the net radiometer, some of the radiation seen by the sensor will come from the dew rather than from the sky above. Since dew is always at a higher temperature than the sky under clear conditions, the measured nocturnal net radiative loss is always less negative than the real value. According to a study of Halldin and Lindroth 1992, the radiometer type used in our experiment un- derestimated the nocturnal net radiative loss by about 30 W m − 2 in Ostby, Sweden. In the course of the night, the measured net radiative loss tended to become less strong. Part of this reduction may have been caused by dew formed on the radiometer dome. This portion should not exceed the difference 1R n between the maximum nocturnal net radiative loss usually before dew occurred and that at the time of the nightly mini- mum temperature. According to our observations, this Table 3 Dew amount and duration, R n,max just before dew occurred, R n,T min at the moment of minimum temperature and their difference a Date Observed Simulated R n,max W m − 2 R n,T min W m − 2 R n,max − R n,T min W m − 2 DEW mm DEWT h DEW mm DEWT h 22–23 February 0.212 12.7 0.218 12.5 − 51.6 − 34.2 − 17.4 23–24 February 0.110 11.0 0.157 13.2 − 48.2 − 41.2 − 6.9 24–25 February 0.172 10.0 0.174 13.1 − 42.8 − 41.5 − 1.3 28 February–1 March 0.142 11.0 0.148 12.8 − 49.8 − 45.8 − 4.0 1–2 March 0.173 12.0 0.204 13.0 − 46.5 − 42.6 − 3.9 28–29 March 0.123 11.8 0.041 11.5 − 42.3 − 37.3 − 5.0 30–31 March 0.141 12.6 0.081 7.9 − 41.2 − 34.6 − 6.6 31 March–1 April 0.180 12.8 0.120 8.3 − 43.3 − 38.8 − 4.5 1–2 April 0.162 12.8 0.130 12.3 − 49.0 − 34.7 − 14.2 2–3 April 0.118 12.7 0.063 7.9 − 46.3 − 38.4 − 7.9 6–7 April 0.120 11.4 0.133 10.3 − 33.4 − 18.9 − 14.6 7–8 April 0.127 9.0 0.075 6.1 − 37.7 − 29.4 − 8.3 8–9 April 0.144 9.2 0.112 8.4 − 35.2 − 34.1 − 1.2 9–10 April 0.126 11.0 0.138 9.4 − 28.2 − 26.1 − 2.2 Mean − 7.0 a All values refer to heavy dew nights. latter value ranged between 0 and 18 W m − 2 with a mean value of 7 W m − 2 during the heavy dew nights as is shown in Table 3. According to the model sen- sitivity analysis, a measurement error of 7 W m − 2 in net radiative loss can cause a dew amount estimation error of about 0.07 mm per night under clear and dry conditions. Therefore, the measurement error of noc- turnal net radiative loss caused by dew formation on the radiometer dome, although it was not as large as that in Sweden, could well have been a reason for the deviation of the simulated results from the observa- tions, together with errors in other relevant weather variables, such as VPD and wind speed. 3.3. Effects of nocturnal net radiative loss, water vapour pressure deficit and wind speed on dew formation To see the direct effect of nocturnal net radiative loss on both dew amount and duration, the total noc- turnal net radiation loss R nt was calculated based on the observed hourly data. The experimental treatment ensured a larger range of R nt than occurs naturally in the field. Relations between R nt and dew amount and duration observed as well as simulated on the top leaves are shown in Fig. 5a and Fig. 6. The slope of the regression line −0.0808 mm m 2 MJ − 1 in Fig. 5a 310 W. Luo, J. Goudriaan Agricultural and Forest Meteorology 104 2000 303–313 Fig. 5. a Relation between daily dew amount, both observed control: d , covered: s and simulated control: +, covered: △, for top leaves and total nocturnal net radiative loss R nt , MJ m − 2 observed: r 2 = 0.66, S.E.=0.033 mm; simulated: r 2 = 0.54, S.E.=0.046 mm. b Similar, but for the effective total nocturnal net radiative loss R nt,effective observed: r 2 = 0.84, S.E.=0.022 mm; simulated: r 2 = 0.64, S.E.=0.040 mm. shows that at crop height, the condensation energy 2.5×10 6 J m − 2 mm − 1 in dew formation had, on average, balanced about 20 of the nocturnal net ra- diative loss. The additional negative effect of vapour pressure deficit in this relation will be investigated below. Compared to the relations between R nt and the observed results for the top leaves, the relations Fig. 6. Relation between daily dew duration both observed con- trol: d , covered: s and simulated control: +, covered: △ on top leaves and the total nocturnal net radiative loss R nt , MJ m − 2 ob- served: r 2 = 0.86, S.E.=1.55 h; simulated: r 2 = 0.72, S.E.=2.48 h. between R nt and simulated results show more scatter Fig. 5a and Fig. 6. To assess the direct effect of vapour pressure deficit and wind speed on dew formation, total night VPD R D dt, mean night VPD D mean , nightly minimum VPD D min and nightly mean wind speed u night were calculated based on the hourly observed data. The ex- posed fraction of the night f, calculated as the ratio of the exposed period to night length from 1800 to 600 equals 12 h, was taken into account when calculating the VPDs because no dew occurred under the covered canopy during the screened period. The value of f was 0.25, 0.167 and 0.083 for the time of removal at 03:00, 04:00 and 05:00, respectively. According to both the experimental data and the simulation results however, there was only a weak direct relation between f R D dt or f D mean or f D min or u night and dew amount or du- ration. But when VPD and u night were added to the regression with the nocturnal net radiative loss, a con- siderable improvement was found for the relation of observed and simulated dew amount Y on top leaves to R nt and f D min and u night Fig. 5b. The multi-linear regression expression for dew amount was: Y = aR nt + bf D min + cu night 2 where a, b and c are regression coefficients. Eq. 2 can be simplified as a single regression equation by introducing a variable defined as corrected nocturnal net radiative loss: W. Luo, J. Goudriaan Agricultural and Forest Meteorology 104 2000 303–313 311 R nt,corr = R nt + b a fD min + c a u night 3 Therefore, Y = aR nt,corr Based on the observed and simulated results, the val- ues of ba and ca were determined as 2.0 kJ m − 2 Pa − 1 and 0.2 MJ m − 3 s when R nt is in MJ m − 2 , VPD in kPa and u night in m s − 1 as shown in Fig. 5b. Com- pared to Eq. 1, the value of ba in Eq. 3 must be ρc p sr a 1t for dew amount when R nt instead of R n is used. In this study, 1t is exposed period in hour. Taking s as 0.2 kPa K − 1 and r a as 50 s m − 1 , the aver- age value of ρc p sr a 1t is about 2.2, which is the same order of magnitude as observed 2.0. The relation between simulated dew amount and R nt,corr r 2 = 0.64, S.E.=0.040 mm shows more scat- ter than that between the observed dew amount and R nt,corr r 2 = 0.84, S.E.=0.023 mm Fig. 5b. The latter relation appears to be even better than the one between the simulated and the observed dew amount Fig. 4 r 2 = 0.80, S.E.=0.033 mm. The above re- sults further confirmed that measurement errors in net radiative loss, VPD and wind speed may be the main causes of the scatter of the simulation results from the observed results. Fig. 8. Relation between measured and simulated dew amount at a 23H crop height and b 12H. The closed dots refer to the control canopy and the open dots to the covered canopy. Fig. 7. Seasonal course of nightly total guttation amount mm per night per leaf area. 3.4. Guttation In our experiment, guttation by the rice plants was quite heavy and its nightly total amount depended on the crop development stage Fig. 7. The nightly total amount of guttation intercepted by the blotting pa- per installed at 12H and 23H decreased from about 0.25 mm per leaf blotting paper area at tillering to 0.05 mm per leaf area at the dough ripe stage. At the tillering stage, the guttation water exuded from the rice leaves was more than the maximum dew amount 312 W. Luo, J. Goudriaan Agricultural and Forest Meteorology 104 2000 303–313 measured on top leaves 0.212 mm during the experi- mental period. These results indicate that guttation, as one of the crop surface wetness contributors, can sup- ply as much water to the rice crop surface as dew. For short grass, Hughes and Brimblecombe 1994 also found that guttation was of the same importance as dewfall. In the paddy rice canopy, the difference between guttation and dewdrops was easily visible because gut- tation drops were much bigger than dew drops. They were suspended along the leaf edge whereas dewdrops were distributed homogeneously on both sides of the leaf surface. Guttation does not contribute to the leaf wetness at the top of the canopy. Inside the canopy however, it might contribute to the leaf wetness in the same way as precipitation does, i.e. the lower layer of the canopy could intercept the guttation drops. This was proved by the fact that the water amount col- lected by the blotting paper at both 12H and 23H was much more than the dew amount simulated at these two heights Fig. 8. Therefore, the contribution of guttation to leaf wetness in paddy rice crops cannot be ignored. Unfortunately, only sophisticated methods such as measuring the isotopic oxygen ratio in water can distinguish guttation water from dew. Our shield- ing experiment was not designed in this way. Yet, it provided a feasible method for allowing an estima- tion of guttation water amount intercepted by different layers inside the rice canopy.

4. Conclusions