Agronomy Journal 95:177-183 (2003)
© 2003 American Society of Agronomy
INSTRUMENTATION
Evaluation and Modification of a Domeless Net Radiometer
Douglas R. Cobos* and
John M. Baker
Department of Soil, Water, and Climate, Univ. of Minnesota, 1991 Upper Buford Circle, St. Paul, MN 55108
* Corresponding author (dcobos{at}soils.umn.edu)
Received for publication January 8, 2002.
 |
ABSTRACT
|
|---|
The use of net radiation as a critical variable in models of surface–atmosphere exchange is routine; however, its measurement is not. The discrete measurement of incoming and outgoing solar radiation with pyranometers and long-wave radiation with pyrgeometers is widely thought to be the most accurate method for estimating net radiation, but deployment at multiple sites is cost prohibitive. A recently developed single-sensor net radiometer differs from other commonly used net radiometers in that it has no domes covering the thermopile sensor. Long-term field data from this net radiometer were compared to reference net radiation calculated as the sum of the four independently measured components. The domeless net radiometer agreed well with the reference measurements, out-performing a similarly priced, widely used, domed net radiometer under most conditions. However, our analyses suggest that the domeless net radiometer has differing sensitivities to short and long-wave radiation, which is not accounted for in the calibration. Additionally, precipitation affected the domeless net radiometer much more severely than the domed model, which could be a significant limitation in areas with frequent rainfall. Despite these shortfalls, the domeless net radiometer evaluated here could be useful in a variety of field settings.
|
INTRODUCTION
|
|---|
ACCURATE MEASUREMENT of net radiation is essential to any study of the surface energy balance. Most surface transport models depend on net radiation as an input. In addition, many surface-atmosphere exchange measurements that are based on modified Bowen ratio techniques require the direct measurement of net radiation. Despite much effort to develop accurate, cost-effective instrumentation for net radiation measurement, there is still no widely accepted best method for field measurements of net radiation.
Instruments that directly measure net radiation are known as net radiometers. Net radiometers consist of a blackened thermopile sensor with one set of junctions facing upward and the other set facing downward, separated by a core, preferably of high thermal diffusivity. The net radiation is proportional to the temperature difference between the upper and lower thermopile junctions. The relation between Rn and the temperature difference between the upper and lower surfaces of net radiometers (after Tanner, 1963) is:
 | [1] |
where 
is the Stephan-Boltzman constant (W m-2 K-4); A and 
are the area (m2) and emissivity (dimensionless) of the sensor surfaces; Kp is the heat transfer coefficient through the net radiometer (W K-1); Kat and Kab are the (wind dependent) coefficients of heat transmission from the top and bottom radiometer surfaces to the air (W K-1); Ka = (Kat + Kab)/2; Ta, Tt, and Tb are the temperatures of air (K), the top surface of the net radiometer, and the bottom surface of the net radiometer; Tm = (Tt + Tb)/2; and 
= (Kat - Kab)/2. Because it is desired that (Tt - Tb) be proportional to the net radiation, it is necessary that be near zero, and that the sum of the terms associated with (Tt - Tb) be constant. The 2Kp/A term is constant and generally very large compared with 4T3m. The greatest challenge in net radiometry is keeping Ka constant and/or small compared to 2Kp/A (Tanner, 1963). One method that has been used to accomplish this is to force ventilate the radiometer, thus forcing to zero and making Ka constant. The disadvantages of forced ventilation include unwanted radiative effects from the ventilator on the net radiometer, and increased power requirements for the blower, making this method impractical at remote locations. A more commonly used method of keeping Ka constant and/or small compared to 2Kp/A is to equip the radiometer with hemispherical domes that shield it from variable ventilation by natural winds, once again forcing to zero and in this case making Ka very small in relation to 2Kp/A. The major disadvantages of domes are (i) they collect dust, (ii) the commonly used polyethylene dome material is not equally transparent to all wavelengths of the radiation spectrum, and (iii) they degrade over time periods as short as 3 mo, thus requiring frequent maintenance. The designers of some net radiometers have addressed the problem of differential radiation absorption by using extremely thin polyethylene domes. These domes, however, require the use of pressurized gas to stay inflated, and are also more susceptible to structural damage.
A problem that affects some net radiometers is differing sensitivities to solar and thermal radiation (Duchon and Wilk, 1994; Field et al., 1992; Halldin and Lindroth, 1992; Wright and Oliver, 1990). This can be avoided by separately measuring these components, using upward and downward facing pyranometers to measure incident and reflected solar radiation and pyrgeometers to measure long-wave radiation. Because these instruments are designed and calibrated specifically for the measurement of their respective wavebands, which together encompass essentially all of the radiation contributing to the surface energy balance, the problem of differing sensitivities to different wavelengths of radiation is minimized. However, the instrumentation cost of a four-component net radiation measurement system can be prohibitive. Several studies have been conducted in which net radiation measurements by different instruments were compared. Field et al. (1992) used a four-component system as a reference for comparing several domed radiometers from different manufacturers. Subsequent analyses showed that instruments with double layer hemispherical domes and instruments with excessively thick domes had significantly lower sensitivities to long-wave radiation than to short-wave radiation. Additionally, 10 to 15% daytime differences and even greater nighttime differences in measured net radiation were found among the instruments. Halldin and Lindroth (1992) conducted a similar comparison and also found that differences existed between the sensitivities to long-wave and short-wave radiation, and that up to 30% errors were present in factory calibration. Kustas et al. (1998) compared several domed net radiometers from two manufacturers and determined that design differences prevented instruments from the two manufacturers from agreeing under changing environmental conditions, despite attempts to cross calibrate. This suggests that to properly assess spatial differences across a landscape, cross calibrated instruments from a single manufacturer should be used.
A recently developed net radiometer (NR Lite, Kipp & Zonen, Delft, the Netherlands) differs from others in that the domes covering the thermopile sensor have been replaced with a black Teflon coating. This is an advantage from the standpoint of maintenance, but raises questions about accuracy and sensitivity to wind speed. With domeless, nonventilated net radiometers, Ka varies with wind speed. In addition, asymmetrical thermal convection from the upper and lower surfaces of the net radiometer and asymmetrical turbulent effects on natural wind due to the mounting structures of the net radiometer both tend to make nonzero. The designers of the NR Lite took steps to minimize these errors. The thin sensor maximizes Kp, while the relatively unobtrusive sensor housing minimizes . The conical shape of the thermopile is designed to improve the cosine response of the instrument, and the Teflon coating allows the sensor to remain relatively clean.
To our knowledge, the NR Lite is the only currently available domeless net radiometer. The manufacturer is cautious to the point of self-deprecation, including in the manual such statements as "The instrument does not have a level because it is not considered accurate enough to require leveling." The purpose of this investigation was to compare direct measurements of net radiation by the NR Lite to a reference net radiation computed by the summation of the four separately measured components to see if its performance was as poor as advertised. A similarly priced, widely used domed net radiometer (model Q*7.1, REBS, Seattle, WA) was also included in the comparison. It should be noted that the manufacturers of the Q*7.1 believe the standard Eppley instrument calibrations to be flawed and instead calibrate the Q*7.1 to modified Eppley instruments, which have been recalibrated (C. Fritschen, personal communication, 2000), thereby decreasing agreement between the Q*7.1 and the four component Eppley system used for comparison here.
Brotzge and Duchon (2000) conducted a comparison among several NR Lite net radiometers, a Q*7.1 net radiometer, and two four-component systems. This investigation revealed large calibration differences between the NR Lite and the reference four-component system. The authors also found that even after standardization, the NR Lite exhibited significant errors when wind speeds exceeded 5 m/s, and when condensation or precipitation were present on the thermopile sensor of the instrument. Additionally, a cosine response error was suspected at solar elevation angles of <20°.
|
MATERIALS AND METHODS
|
|---|
The time constant of the NR Lite was determined by tracking its signal as it was alternately shaded and exposed to a light source in a still-air laboratory setting. Accuracy and uniformity of spectral response were determined from an extended set of field measurements. The data used in this analysis were collected over a 335-d period (DOY 31–365) in 1999 at the University of Minnesota's Rosemount Agricultural Experiment Station, 24 km south of St. Paul (44°44' N, 93°05' W). A permanent instrument tower is positioned in the center of a 17-ha agricultural field with approximately 200 m of fetch. A full description of the site and instrumentation can be found in Baker et al. (1999). During the course of this study the surface beneath the radiometers ranged from snow, to soil (Waukegan silt loam) with approximately 25% corn residue, to full corn canopy. The equipment pertinent to this study included two net radiometers (the NR Lite and Q*7.1), and upward and downward facing pyranometers and pyrgeometers (models 8-48 and PIR, Eppley Laboratories, Newport, RI) mounted in close proximity at a height of 3 m from the soil surface. To help minimize the effects of dew or frost, the upward-facing pyranometers and pyrgeometers were force ventilated by fans. Wind speed was measured by a cup anemometer (Gill, model 12102) mounted at a height of 3 m on the same mast as the radiation sensors.
The pyrgeometers were factory calibrated at the time of purchase, within 5 yr of this study. Concerns over the accuracy of the model 8-48 pyranometers prompted the purchase of Eppley's first tier pyranometers (PSP) after the conclusion of this study. A side-by-side comparison of upward and downward facing 8-48's and PSPs over a 20-d period yielded no statistically significant differences (p < 0.01) in slope and offset of the regression from unity and zero, respectively. Net radiation was calculated from these instruments by summing net long-wave and net short-wave radiation. The NR Lite net radiometer was supplied with a factory calibration at the time of purchase within 1 yr of the beginning of this study. Because it has no domes, the NR Lite signal is a function of both net radiation and wind speed. To remedy this, a factory-supplied wind-correction factor was applied to the net radiation data from the NR Lite during postprocessing when wind speed exceeded 5 m s-1. The Q*7.1 net radiometer was factory-supplied with two calibration factors for use during periods of positive and negative net radiation. A factory-supplied wind correction factor was also applied to the net radiation data from the Q*7.1. The polyethylene domes of the Q*7.1 are routinely changed every 6 mo, and were changed previous to the onset of this study. Data from all the radiation sensors were recorded every 10 s and averaged over 30-min intervals by a datalogger (model CR 23X, Campbell Scientific, Logan, UT).
After the conclusion of the initial study, seven NR Lite net radiometers were mounted within 2 m of the pyrgeometers and pyranometers. Intercomparison data were collected from these radiometers to compare their relative calibration and sensitivity under field conditions.
The resulting data are presented as linear regressions, with the four-component net radiation data taken as reference values (x axis). Despite inherent limitations, we have included coefficient of determination (R2) values for the linear regressions, because R2 values are more easily interpreted by the general scientific audience than more meaningful measures of "goodness of fit" (Legates and McCabe, 1999). In addition, standard errors, slopes, and y intercepts are reported for the regressions.
|
RESULTS AND DISCUSSION
|
|---|
The dynamic response is expressed as the time required for the instrument error to drop below 
e-1 following a step change of . We measured a value for the NR Lite of approximately 13 s (data not shown), consistent with the stated specification of <20 s. By comparison, manufacturer's specifications for the Eppley instruments and the Q*7.1 are 2 to 4 s and 30 s, respectively.
Figure 1a shows the entire field data set of net radiation as determined with the NR Lite factory calibrations plotted against net radiation calculated by component summation. The relationship exhibits a high coefficient of determination, but also a relatively high standard error (Table 1). The R2 value of the Q*7.1 regressed against the same component summation data (Fig. 1b) is identical to that of the NR Lite, with a somewhat improved standard error. We attempted to identify causes for discrepancies between the NR Lite and the summed measurements. One obvious potential source of problems is precipitation, so all data collected during intervals when precipitation was recorded were subsequently purged. This improved the comparison, but a systematic lack of agreement between the two methods was still noted immediately preceding and following precipitation. The cause of the poor agreement during periods immediately before recorded precipitation is presumably slight, unmeasured precipitation, sufficient to wet the net radiometers without tipping the bucket in the precipitation gauge. This phenomenon was also reported by Brotzge and Duchon (2000). The presence of such moisture likely affects the domed pyranometer and pyrgeometer differently than the undomed NR Lite net radiometer. Similarly, differences in the evaporation of residual precipitation from the domed Eppley and undomed NR Lite are probably the cause of inconsistencies after the cessation of precipitation. Because the domes on the pyranometers and pyrgeometers are thermally isolated, evaporation of precipitation should have less effect on the output from their thermopile sensors than on the domeless NR Lite. Thus, the inconsistencies observed in the measurements of net radiation between the two methods during rainfall events are attributed to the NR Lite. Consistent with this notion, the standard error for the domed Q*7.1 was essentially unaffected by the exclusion of the data collected during precipitation periods (Table 1).
View larger version (30K):
[in this window]
[in a new window]
|
Fig. 1. Comparison of net radiation measured by the NR Lite and Q*7.1 net radiometers vs. summation of the four components for all sky conditions. Best-fit regressions are plotted as solid lines.
|
|
View larger version (24K):
[in this window]
[in a new window]
|
Fig. 6. Net radiation measured by the NR Lite net radiometer versus summation of the four components in two fields, for nighttime periods, before and after addition of Mylar tape.
|
|
Efforts were made to purge the data set of pre- and postprecipitation effects. Following many rains that occurred during the 1999 measurement period, the effects of residual moisture on the NR Lite were readily apparent for several hours after the rainfall. After nighttime rainfall, when subsequent evaporation was slow, the NR Lite yielded apparently erroneous results for up to 6 h, often until midmorning. Residual rainfall effects disappeared more quickly after many of the daytime rains. The exclusion of these rain-affected data increased the linearity of the NR Lite regression against the summed net radiation, and reduced the standard error (Table 1).
It is suspected that the routine presence of morning dew or frost on both the NR Lite and Q*7.1 introduced much error into the measurements compared with the summed components, which were force-ventilated to reduce dew and frost formation on the domes. Evaporation of dew or frost directly from the thermopile sensor of the NR Lite would be expected to introduce greater error than evaporation from the dome of the Q*7.1, similar to the observed effects of rainfall. Unfortunately, during this comparison, a dew/frost sensor was not present at the site of the experiment. Efforts were made to utilize environmental data to determine time periods when dew or frost were likely to be present. These attempts proved to be largely useless, and were not included in subsequent analyses. The efforts made to purge the data set of precipitation and condensation-affected data underscore the susceptibility of the NR Lite to errors associated with these conditions.
Many models use daily totals of net radiation rather than the 30-min averages that have been reported. To help evaluate the applicability of the NR Lite for use with such applications, daily totals of net radiation were computed over the 335-d measurement period for the NR Lite and Q*7.1 (Table 1, Fig. 2). Interestingly, the Q*7.1 outperformed the NR Lite when data were integrated in this fashion.
View larger version (25K):
[in this window]
[in a new window]
|
Fig. 2. Comparison of daily totals of net radiation measured by the NR Lite and Q*7.1 net radiometers vs. summation of the four components with data affected by precipitation excluded.
|
|
To determine if the NR Lite has similar sensitivities to direct beam and diffuse short-wave radiation, 30-min data from daytime periods with clear and overcast sky conditions were segregated and plotted (Fig. 3a and 4a, respectively). The clear sky regression exhibits a high coefficient of determination, but a high standard error. This is due mostly to the large net radiation values expected under clear skies, but could be partially due to the increased likelihood of dew and frost formation on clear mornings. The regression of the Q*7.1 data against the summed components also exhibited a higher standard error, under clear sky, daytime conditions than under all sky conditions (Fig. 3b), and generally poorer performance than the NR Lite.
View larger version (24K):
[in this window]
[in a new window]
|
Fig. 3. Comparison of net radiation measured by the NR Lite and Q*7.1 net radiometers vs. summation of the four components for all clear sky daytime conditions with data affected by precipitation excluded.
|
|
View larger version (23K):
[in this window]
[in a new window]
|
Fig. 4. Comparison of net radiation measured by the NR Lite and Q*7.1 net radiometers vs. summation of the four components for all overcast daytime conditions with data affected by precipitation excluded.
|
|
The regression of the NR Lite data against the summed components from overcast time periods shows a lower coefficient of determination, and lower standard error (due to the generally lower values of net radiation) when compared with all sky conditions (Fig. 4a). The presence of cloud cover increases the contribution of the incoming long-wave component to the radiation balance, suggesting that the NR Lite is less sensitive to long-wave radiation. The Q*7.1 agreed better with the summed components under daytime, overcast sky conditions than during clear sky conditions (Fig. 4b), and exhibited significantly better performance than the NR Lite.
Figure 5a shows the nighttime data from the NR Lite plotted against the summation of the components. The agreement between the two measurement methods is rather poor with substantial scatter. These data are consistent with several other studies that conclude net radiometer performance is poorest when the radiation balance is dominated by the long-wave component (Baker, 1998; Duchon and Wilk, 1994; Field et al., 1992). This strengthens the suspicion that the NR Lite has differing sensitivity to long-wave and short-wave radiation. The regression of Q*7.1 nighttime data against the summed components yielded a lower R2 value and a higher standard error than the NR Lite regression, suggesting poorer nighttime performance (Fig. 5b).
View larger version (36K):
[in this window]
[in a new window]
|
Fig. 5. Comparison of net radiation measured by the NR Lite and Q*7.1 net radiometers vs. summation of the four components for nighttime conditions with data affected by precipitation excluded.
|
|
Multiple regression of the voltage output from the NR Lite against net short-wave and net long-wave values from the Eppley instruments for all rain-free data showed that the NR Lite sensitivity to long-wave radiation is less than that for solar radiation by about 15%. This suggests that the performance of the NR Lite could be improved by coating 15% of each surface with a substance that reflects solar radiation and absorbs long-wave radiation, to reduce the short-wave sensitivity of the thermopile sensor. To accomplish this, Mylar tape (850, 3M, St. Paul, MN) was applied to both the upward and downward faces of the NR Lite in three pie slices, which covered a total of 15% of each thermopile sensor. At the time of application, it was assumed that the long-wave emissivity of the Mylar tape was near 1, but subsequent investigation revealed a value of 0.59. The short wave reflection coefficient of Mylar is 0.9. Data were collected from two NR Lite net radiometers in different fields from DOY 159 to 199, 2001. Each field was equipped with a four-component net radiation measurement system as described above. The reflective tape was placed on each radiometer midway through each measurement period (DOY 179). Throughout the course of this trial, the raw mV output from the NR Lite was recorded. Best-fit regression of the NR Lite mV output against the summed components was used to determine calibration factors both before and after the addition of the reflective tape.
The addition of Mylar tape to the radiometer surfaces increased the standard error slightly in both fields (Table 1). With only three slices of Mylar tape present, the movement of the sun across the horizon probably caused variable exposure of the individual slices to direct sunlight throughout the course of the day. The application of more and thinner slices, or else concentric circles of reflective material, could remedy this problem.
Figure 6 (a and b) shows the regressions of the NR Lites in both fields against the summed components, both before and after modification, during nighttime conditions only. Separate nighttime calibration coefficients were not determined, but instead, the best-fit all-condition calibration coefficients were utilized. As can be clearly seen from the regression statistics (Table 1), and Fig. 6, the addition of the Mylar tape decreased the standard error and brought the slope of each regression nearer unity during nighttime periods. The addition of the reflective tape allows the user to obtain more accurate nighttime net radiation measurements while applying only one calibration coefficient, rather than the separate daytime and nighttime calibration coefficients advocated by Brotzge and Duchon (2000).
A cursory investigation was conducted into the effects of wind speed on the relative accuracy of the uncorrected NR Lite sensors. Data were separated into three classes based on wind speed: low (U < 2.5 m/s), medium (2.5 m/s < U < 5 m/s), and high (U > 5 m/s). The standard error increased with increasing wind speed among the three classes for both daytime and nighttime periods (Table 2). This suggests that the accuracy of the NR Lite decreases with increasing wind speed, even at moderate wind speeds (Fig. 7). The wind speed correction recommended by Kipp and Zonen and a wind speed analysis conducted by Brotzge and Duchon (2000) both only call for a correction to be applied at wind speeds >5 m/s. Our data indicate that this correction needs to be applied at lower wind speeds as well.
View this table:
[in this window]
[in a new window]
|
Table 2. Regression statistics for wind speed analysis. Data were sorted into three categories based on wind speed, during all conditions, daytime only, and nighttime only. Net radiation data from the NR Lite were regressed against net radiation from component summation for all wind speed and times of day.
|
|
View larger version (23K):
[in this window]
[in a new window]
|
Fig. 7. Error in NR Lite measurement of net radiation vs. wind speed. Error is defined as the difference between net radiation from the NR Lite and component summation.
|
|
After the conclusion of the initial study, seven NR Lite net radiometers were compared to test their relative calibration and sensitivity under field conditions. A comparison of the seven NR Lite net radiometers with the summed components showed a consistent underestimation of the magnitude of both positive and negative net radiation by the NR Lites when the factory calibrations were used. The average daytime underestimation was -49.4 W m-2, whereas the average nighttime underestimation was 4.88 W m-2. This error was present and approximately equal in all seven NR Lites. The difference between net radiation measured by each NR Lite and the group mean was determined with factory calibrations to assess the agreement among the NR Lites. The average difference between the individual sensors and the group mean was 3.60 W m-2, indicating good agreement among the sensors. Linear regressions between the raw mV output from the NR Lites and net radiation measured as the sum of the components were developed, and best-fit calibrations were computed for each NR Lite. The best-fit calibrations increased by 0.49 W m-2/µV, or 14% over the factory calibrations. The average difference between the individual sensors and the group mean remained essentially unchanged with the alteration of the calibration factors.
|
CONCLUSIONS
|
|---|
Our data indicate that the NR Lite agrees well with the summation of the four independently measured components of net radiation under most field conditions. In fact, under some conditions the NR Lite appears to agree more closely with the summed components than does the Q*7.1. However, because it has no shielding domes, precipitation and condensation adversely affect the NR Lite for extended periods of time. Although many of the micrometeorological measurement techniques that are dependent on measurements of net radiation routinely fail during precipitation events, eliminating the need for net radiation data, the significant amount of postprecipitation down time exhibited by the NR Lite can be problematic. The absence of domes also allows wind to affect the net radiation measurement at speeds as low as 2.5 m/s, although not severely. Under all daytime sky conditions, the agreement was good between the NR Lite and summed components, but, like other commercially available net radiometers, the accuracy of nighttime net radiation measurements by the NR Lite was not as good, primarily because the instrument is less sensitive to long-wave radiation than it is to solar radiation, by about 15%. This problem was addressed by covering 15% of each surface with film that is highly reflective in the solar spectrum while retaining high emissivity in the thermal portion of the spectrum. This resulted in improved performance at night, and negated the need for separate day and night calibration factors. Further improvement is probably possible with optimization of the configuration of reflective material. Overall, our results indicate that the performance of the NR Lite is better than one might expect after reading the manual, and with minor modifications it could be a useful research instrument, especially in arid areas.
|
REFERENCES
|
|---|
- Baker, J.M. 1998. Comparison of net radiometer measurements against summed radiation components. Symp. on Meteorological Observations and Instrumentation, 10th, Phoenix, AZ. 11–16 Jan. 1998. Am. Meteorological Soc., Boston, MA.
- Baker, J.M., K.J. Davis, and G.C. Liknes. 1999. Surface energy balance and boundary layer development during snowmelt. J. Geophys. Res. 104:19611–19621.
- Brotzge, J.A., and C.E. Duchon. 2000. A field comparison among a domeless net radiometer, two four-component net radiometers, and a domed net radiometer. J. Atmos. Ocean. Technol. 17: 1569–1582.
- Duchon, C.E., and G.E. Wilk. 1994. Field comparisons of direct and component measurements of net radiation from clear skies. J. Appl. Meteorol. 33:245–251.
- Field, R.T., L.J. Fritschen, E.T. Kanemasu, E.A. Smith, J.B. Stewart, S.B. Verma, and W.P. Kustas. 1992. Calibration, comparison, and correction of net radiation instruments used during FIFE. J. Geophys. Res. 97:18681–18695.
- Halldin, S., and A. Lindroth. 1992. Errors in net radiometry: Comparison and evaluation of six radiometer designs. J. Atmos. Ocean. Technol. 9:762–783.
- Kustas, W.P., J.H. Prueger, L.E. Hipps, J.L. Hatfield, and D. Meek. 1998. Inconsistencies in net radiation estimates from use of several models of instruments in a desert environment. Agric. Forest Meterol. 90:257–263.
- Legates, D.R., and G.J. McCabe, Jr. 1999. Evaluating the use of "goodness-of-fit" measures in hydrologic and hydroclimatic model validation. Water Resour. Res. 35:233–241.
- Tanner, C.B. 1963. Basic instrumentation and measurements for plant environment and micrometeorology. Soils Bull. 6. Dep. of Soil Science, Univ. of Wisconsin, Madison, WI.
- Wright, I.R., and H.R. Oliver. 1990. Correction of errors associated with measurement of net-all wave radiation with double-domed radiometers. Boundary Layer Meteorol. 53:401–407.
This article has been cited by other articles:
|
|
|
|
 
J. L. Steiner and J. L. Hatfield
Winds of Change: A Century of Agroclimate Research
Agron. J.,
May 7, 2008;
100(Supplement_3):
S-132 - S-152.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
