Sward age and weather effects on alfalfa yield at a
semi-arid location in southwestern Saskatchewan
P .G. Jefferson and H.W. Cutforth
Agriculture and Agri-Food Canada, Semiarid Prairie Agricultural Research Centre, Box 1030, Swift Current,
Saskatchewan, Canada S9H 3X2. Received 8 July 1996, accepted 22 May 1997.
Jefferson, P. G. and Cutforth, H. W. 1997. Sward age and weather effects on alfalfa yield at a semiarid location in southwestern
Saskatchewan. Can. J. Plant Sci. 77: 595–599. Alfalfa (Medicago sativa L.) yield in the first and second years after establishment
is typically much greater than yield in subsequent years under dryland production systems in semiarid regions. Alfalfa is
a deep-rooted perennial that uses soil water stored at soil depths below the reach of shallow-rooted cereals and grasses. Since
alfalfa yield is positively related to evapotranspiration, this study was conducted to determine the relationship between historical
alfalfa yield data and weather variables as affected by sward age. Rambler alfalfa yields collated by sward age during cultivar yield
trials from 1951 to 1994 at Swift Current, Saskatchewan, were statistically related to monthly precipitation (April to August) and
monthly pan evaporation (May to September) during the growing season, and to the fall and winter total precipitation (September
to March). One-year-old swards yielded more than 3-, 4- or 5-yr-old swards. For 1- and 2-yr-old alfalfa swards, weather accounted
for 50% and 47% of the yield variability, respectively. However, weather accounted for 85, 87 and 96%, respectively, for 3-,
4- and 5-yr-old swards. We hypothesize that soil water stored deep in the profile accounted for much of the remaining yield variability
in one and two year old swards. Researchers must measure soil water use from soil depths to at least 3 m when assessing
dryland alfalfa yields.
Key words: Medicago sativa L., weather, modelling, forage yield
Jefferson, P. G. et Cutforth, H. W. 1997. Effet de l’âge du peuplement et effet de la température sur la productivité de la
luzerne en milieu semi-aride dans le sud-ouest de la Saskatchewan. Can. J. Plant Sci. 77: 595–599. Le rendement de la luzerne
(Medicago sativa L.) dans les première et deuxième années suivant l’installation est normalement plus abondant que dans les
années subséquentes dans les systèmes de production en culture sèche pratiqués dans les régions semi-arides. La luzerne est une
vivace à enracinement profond qui peut aller puiser l’eau du sol à des profondeurs inaccessibles aux racines superficielles des
céréales et des graminées. Comme le rendement de la luzerne est relié positivement à l’évapotranspiration, nous avons voulu
établir le lien existant entre les données accumulées au cours des années sur le rendement de la luzerne et les variables
météorologiques en fonction de l’âge du peuplement. Nous avons groupé par âge de peuplement les rendements du cultivar
Rambler obtenus dans les essais variétaux conduits de 1951 à 1994 à Swift Current en Saskatchewan. Les rendements étaient statistiquement
reliés aux mesures mensuelles de précipitations (avril à août) et de l’évaporation en bac (mai à septembre) durant la
saison de végétation ainsi que des précipitations totales de septembre à mars. Les peuplements d’un an produisaient plus que ceux
de 3, 4 ou 5 ans. Pour les peuplements d’un et de deux ans, les conditions météorologiques comptaient, respectivement pour 50 et
47 % de la variabilité du rendement, mais ces proportions passaient à 85, 87 et 98 %, respectivement, pour les peuplements de
trois, quatre et cinq ans. Une bonne partie du restant de la variabilité du rendement observée dans les peuplements de un et de deux
ans s’expliquerait par la quantité d’eau stockée dans les couches profondes du profil. L’évaluation des rendements de luzerne en
culture sèche devrait donc tenir compte de l’eau disponible jusqu’à au moins 3 m de profondeur.
Mots clés: Medicago sativa L., temps, modélisation, rendement fourrager
Alfalfa (Medicago sativa L.) is the most widely grown forage
legume in Western Canada due to its wide adaptation,
high forage quality, and high yield potential (Goplen et al.
1980). Some alfalfa cultivars, such as Beaver, Barrier,
Anchor and Algonquin, are grown for high yield production
in areas where irrigation is available. Other cultivars with
greater drought and cold temperature tolerance are grown
under dryland conditions. In the Brown soil zone of Western
Canada drought stress occurs frequently (Cutforth et al.
1993) and commonly grown cultivars are creeping-rooted in
plant type, such as Rambler, Roamer, Heinrichs and
Rangelander.
The drought tolerance of alfalfa may be related to its ability
to withdraw water from deep in the soil profile and below
the root zone of annual crops (Kiesselbach et al. 1929). On
the northern Great Plains, alfalfa has been reported to
deplete soil water to 3 to 4 m deep in the profile (Brun and
Worcester 1975; Cutforth et al. 1991; Campbell et al. 1994).
Another drought tolerance strategy for alfalfa may be its
ability to deplete soil water to –4 MPa tension (Brun and
Worcester 1975; Cutforth et al. 1991), which is well below
the laboratory lower limit of soil water extraction at –1.5
MPa. Studies of alfalfa root architecture indicate that the
largest proportion of alfalfa root mass is located in the upper
30 cm of the soil profile (Sheaffer et al. 1988). Thus, the
595
Abbreviations: APR, MAY, JUN, JUL, AUG, monthly
precipitation from April to August, respectively; EVAPMAY,
EVAPJUN, EVAPJUL, EVAPAUG, EVAPSEP,
monthly evaporation for May to September, respectively;
FWPPT, fall and winter (September to March) total precipitation.
596 CANADIAN JOURNAL OF PLANT SCIENCE
relative importance of deep soil water to alfalfa grown in
semiarid environments might be overlooked by agronomists,
plant breeders, and crop modellers.
Alfalfa yield has shown a positive linear relationship to
evapotranspiration (Sheaffer et al. 1988). Previous alfalfa
water relation research done at this location has shown a
relationship between pan evaporation and potential evapotranspiration
(Pelton and Korven 1969). Prediction of alfalfa
forage yield from environmental variables that are closely
associated with evapotranspiration would be desirable in
deterministic or mechanistic crop modelling. However, the
accuracy of such predictions may be greatly reduced if
water from deep soil layers contributes significantly to alfalfa
production, particularly of newly seeded swards, and this
water is not accounted for. Keisselbach et al. (1929)
observed that production of newly seeded alfalfa swards
was significantly higher than older stands and attributed this
to the crop’s use of deep soil water. Further, large reserves
of soil water at depths below approximately 1.2 to 1.5 m
which are unavailable to relatively shallow rooted annual
cereals or perennial grasses would be available to deep rooted
crops such as alfalfa (Brun and Worcester 1975;
Campbell et al. 1993, 1994; Kiesselbach et al. 1929).
Sward age is not considered in the long-term forage yield
evaluation of alfalfa. The literature and our experience suggest
that often sward age is confounded with growing season
environmental conditions, and this exerts considerable
influence on the variation in forage yields that are observed.
Thus to consider the effect of sward age over a large number
of observations (growing seasons), a historical database
for the Swift Current location was developed.
The objective of this preliminary research was to evaluate
the relationship of historical alfalfa yield data with weather
variables and sward age (years since establishment) for a
semiarid site in southwestern Saskatchewan as affected by
alfalfa. The inclusion of sward age in the above relationship
has not been developed previously by agronomists, plant
breeders or crop modellers but might have significant
impacts on interpretation of their results.
MATERIALS AND METHODS
The database of historical alfalfa yields (1951 to 1994) consisted
of the cultivar Rambler. Rambler was chosen as the
benchmark variety for several reasons. It was registered in
1955 but continues to be included in dryland alfalfa trials
since it is a widely known cultivar. This resulted in the
largest number of years of available data. It is well adapted
and winter-hardy and data from long-term trials were available.
The database was gleaned from old Research Centre
reports, Uniform Alfalfa Cultivar Trials (pre-registration
tests) and a few agronomy trials. All experiments were
spring seeded on previously fallowed soil (Swinton loam;
fine, mixed, mesic aridic Haploboroll) located on the dryland
research area of the Semiarid Prairie Agricultural
Research Centre at Swift Current, Saskatchewan (50o 16?N,
107°44?W, 825 m elevation). Each data point was the mean
over the number of replications for that year. If two or more
experiments of the same sward age were reported in the
same calender year, we took the mean for Rambler over
experiments. In 1962 and 1985, there were no alfalfa trials
harvested due to winter injury of existing experiments. Data
from 1955 were discarded as the samples contained a large
proportion of weeds.
Harvests were taken as close to 10% bloom stage as possible.
Total annual forage yield data were used for years
where more than one harvest was recorded.
All alfalfa experiments were located within 0.65 km of the
meterological station. The weather data used in this study
included monthly precipitation (mm) from April to August
(APR, MAY, JUN, JUL, AUG). Monthly precipitation for the
previous September to March were accumulated as Fall and
Winter precipitation (FWPPT; water equivalent (mm)).
Monthly evaporation (mm) from a Class A pan was used for
May through September (EVAPMAY, EVAPJUN, EVAPJUL,
EVAPAUG, EVAPSEP). Freezing temperatures from
October to April make it impossible to determine evaporation
from a Class A pan during that period. Evaporation was considered
to reflect several weather variables, such as temperature,
wind speed and relative humidity. Water deficit was
calculated as the sum of monthly evaporation from May to
July minus the sum of monthly precipitation from April to
July.
Sward age from 1 to 5 yr was treated as a factor for statistical
analysis. Since there were fewer observations for
sward ages 6 to 11 yr, all these observations were combined
into one category, namely >5 yr sward age. The sample size
for each sward age category is listed in Table 1. The GLM
procedure and pair-wise t-testing of least square means
(SAS Institute, Inc. 1985) were used to compare sward age.
The STEPWISE procedure (SAS Institute, Inc. 1985) was
used for forward multiple regression analysis to predict
yield from weather data. The threshold probability for
including variables in the model was 0.500. Pearson correlation
coefficients between yield and water deficit were calculated
with the REG procedure (SAS Institute, Inc. 1985).
RESULTS AND DISCUSSION
Forage yield varied from 120 kg ha–1 in 1968 to 4805 kg
ha–1 in 1978 (Table 1). The largest variation within a calender
year due to sward age occurred in 1956 with yields ranging
from 633 to 4057 kg ha–1. The 1-yr old swards yielded
the highest on average; forage yield generally declined with
age of sward to 3 yr old but recovered slightly for the >5-yrold
stands. The difference among the sward ages was significant
(P ??0.011) The high average yield for 1- and
2-yr-old swards would support the hypothesis that alfalfa is
extracting deep soil water early in its life. This water reserve
would decline with use, barring a significant recharge event,
and subsequently yields would decline. The recovery of
average yield in >5-yr-old stands may be related to the
demographics of these data. These old swards must have
exhibited reasonable production and plant density to be
maintained for such a long period while other trials of similar
age had been discarded. Perhaps the persistence of these
swards contributed to the recovery of yields observed. These
long-term stands were not harvested in droughty years due
to low productivity. Had low production years been includJEFFERSON
AND CUTFORTH — SWARD AGE AFFECTS ALFALFA YIELD 597
ed, this mean value might approach that for the 3- to 5-yrold
swards. We considered the implications of including the
yield data for >5-yr-old swards and recognize the bias that
selective harvesting has caused in the database. However,
we concluded that the value of including yields from old
swards was greater than the risk of bias.
As sward age increased from 1 to 5 yr, the ability of the
weather variables to explain the variation in yield increased
from 50 to 96% (Table 2). The >5-yr-old sward yield prediction
was 75%, less than that for the 5-yr-old swards but
better than the 1- and 2-yr sward predictions. Again, these
data may be skewed by the vigor of the plants. Fall and winter
precipitation, April, May, June and July monthly precipitation
totals were included in at least four of the regression
formulae as were monthly evaporation for May, June, July
and August.
Alfalfa forage yield was correlated to annual water deficit
(Evaporation less precipitation) for each sward age (Fig. 1).
The correlation coefficients were lower at sward ages one
and two than for older swards and this was similar to the
multiple regression results. It is noteworthy that a cluster of
high yields at high water deficits is evident in both 1- and
2-yr-old swards but is not evident in the other ages. We
interpret these values as representing swards where soil
water has contributed to the yields observed. One datum (n)
was excluded from the regression and correlation calcula-
Table 1. Historical dryland alfalfa yield at Swift Current, Saskatchewan from 1951 to 1994 by age of stand
Stand age (yr)
Year 1 2 3 4 5 > 5 SE
(kg ha–1)
1951 1475 –
1952 1681 –
1953 2744 3002 3943 631
1954 3832 3956 3674 3016 419
1956 4057 633 2421
1957 952 –
1958 1120 605 364
1959 1596 1255 241
1960 896 1101 740 181
1961 340 112 (6)z 161
1963 2554 –
1964 638 605 23
1965 2958 3181 158
1966 4551 2246 2509 1262
1967 2274 864 1026 772
1968 255 120 95
1969 1693 2604 1178 (6,7) 722
1970 3812 4321 3841 3388 (7,8) 381
1971 1607 1224 733 438
1972 1396 1231 1100 593 (10) 346
1973 2116 2157 1330 2007 1310 (11) 427
1974 3743 2444 2494 2415 (6) 647
1975 2699 4082 3193 2733 644
1976 3360 4184 3132 3176 3412 (6) 425
1977 4486 2502 2864 3260 2450 3841 (7) 803
1978 4805 2376 1600 2096 1941 (6,8) 1284
1979 1419 3487 2538 1754 2034 1992 (7) 728
1980 2078 1302 1450 1964 380
1981 2942 1701 2109 632
1982 3277 1119 3759 (10) 1406
1983 3467 2843 2907 (11) 343
1984 1394 –
1986 988 –
1987 350 –
1988 200 –
1989 2552 1606 669
1990 2404 2874 –
1991 3698 3735 3808 1063 3558 (6) 1183
1992 1290 724 2259 2900 675 (7) 980
1993 1788 3077 (6) 911
1994 4065 –
Mean 2642A 2400AB 1977C 2087BC 2085B 2259ABC
SEM 270 250 252 225 286 286
n 23 23 22 19 15 18
zIn > 5 yr column, stand age is listed in parentheses after the yield value. Where two ages are listed, a mean yield is presented for brevity. Individual yields
were used in data analysis.
A–C Means followed by the same letter are not significantly different as tested by paired t-tests (P < 0.05).
598 CANADIAN JOURNAL OF PLANT SCIENCE
tion for age four. In 1982, a 36-cm snowfall from 27 to
29 May resulted in frost damage to all alfalfa cultivars and
reduced the yield potential of Rambler. If this datum were
included in the correlation, the coefficient would be –0.51
(P < 0.05).
When taken together, these results suggest that yields
from 1- and 2-yr-old swards of dryland alfalfa do not accurately
reflect long-term forage productivity at this semiarid
site. This is consistent with Brun and Worcester’s (1975)
observation that 1- and 2-yr-old swards of alfalfa had 120 to
150 mm of available soil water (at tension ??1.5 MPa) at
1.22 to 2.44 m deep in the profile, while 6-yr-old swards had
zero available soil water at that depth. They also reported
that 3- and 4-yr-old swards had 50 mm of additional soil
water stored. However, from our results of long-term yields
three year old swards were as low in productivity as 5-yr-old
swards. Using the 56 to 73 mm H2O t–1 dry matter conversion
factor determined by Sheaffer et al. (1988), our yield
data suggest that between 37 and 48 or 24 and 31 mm more
soil water was used by 1 or 2-yr-old swards, respectively,
compared to 3-yr-old alfalfa swards. The preceding crop
management used on areas included in our database ranged
from wheat-fallow rotation (slow recharge of deep soil
water) to 2 or more years of continuous fallow (fast recharge
of deep soil water). If alfalfa had been grown previously on
a trial site, there may have been little or no recharge of deep
soil water (Kiesselbach et al. 1929). This may explain, in
part, the wide range of variation within individual years to
sward age.
The practical importance of these results is twofold.
Regional cultivar trials are conducted for 3 harvest years in
Saskatchewan prior to recommendation to producers. Yield
decline with stand age should be considered in the evaluation
of new cultivars to avoid over estimation of a cultivar’s
yield potential. Deep soil water utilization (to 3 m soil
depth) by alfalfa must be documented in alfalfa productivity
research such as agronomy or modelling. If soil water use
is not documented, dryland alfalfa yields will be skewed
high when yield trials are grown on land not previously
seeded to deep rooted crops, i.e. the yields claimed will be
misleading and not a true reflection of long-term dryland
yields. Once deep soil water reserves have been exhausted,
dryland alfalfa yields will depend upon annual soil water
reserves added by infiltration of snow melt from the previous
winter and growing season precipitation — a true dryland
situation. One suggestion would be to report yields, or
yield advantages, in water use efficiency terms, i.e., dry
matter yield per unit of water used. Assessments of the longterm
productivity of dryland alfalfa must consider yield
decline with stand age and the longevity of the stand as factors
in the analysis.
Brun, L. J. and Worcester, B. K. 1975. Soil water extraction by
alfalfa. Agron. J. 67: 586–589.
Campbell, C. A., Zentner, R. P., Selles, F. and Akinremi, O. O.
1993. Nitrate leaching as influenced by fertilization in the Brown
soil zone. Can. J. Soil Sci. 73: 387–397.
Table 2. Weather variables selected by forward stepwise multiple regression equations for historical alfalfa yield prediction from weather data at
Swift Current, Saskatchewan by stand age
Stand Partial correlation coefficient by variable Model
age (yr) FWPPT APR MAY JUN JUL AUG EVAPMAY EVAPJUN EVAPJUL EVAPAUG R2 Prob. RSD
1 – 0.018 0.268 – 0.06 0.03 – – 0.083 0.045 0.498 0.056 1076
2 0.022 0.023 0.035 0.048 – – 0.123 0.195 0.023 – 0.469 0.142 1058
3 0.033 0.126 0.08 0.016 0.016 – 0.396 0.119 – – 0.852 0.001 663
4 0.036 0.041 0.083 0.040 0.126 – 0.149 0.241 – 0.023 0.869 0.002 731
5 – 0.076 0.204 0.282 0.105 – 0.056 0.080 0.020 0.049 0.955 0.006 341
> 5 0.092 – 0.054 0.104 0.046 – – 0.309 0.041 0.110 0.756 0.017 782
Fig. 1. Alfalfa forage yield as affected by water deficit
(Evaporation less precipitation) by age of sward. Lines shown are
linear regressions for reference. One datum (n) was excluded from
regression and correlation for sward age four. *,**, *** denote
P < 0.05, P < 0.01 and P < 0.001, respectively.
JEFFERSON AND CUTFORTH — SWARD AGE AFFECTS ALFALFA YIELD 599
Campbell, C. A., Lafond, G. P., Zentner, R. P. and Jame, Y. W.
1994. Nitrate leaching in a Udic Haploboroll as influenced by fertilization
and legumes. J. Environ. Qual. 23: 195–201.
Cutforth, H. W., Jefferson, P. G. and Campbell, C. A. 1991.
Lower limit of available water for three plant species grown on a
medium-textured soil in southwestern Saskatchewan. Can. J. Soil
Sci. 71: 247–252.
Cutforth, H. W., Jones, K. and Lang, T-A. 1993. Agroclimate of
the Brown soil zone of southwestern Saskatchewan. Research
Branch, Agriculture Canada Research Station, Swift Current, SK.
Publ. no. 379MOO88.
Goplen, B. P., Baenziger, H., Bailey, L. D., Gross, A. T. H.,
Hanna, M. R., Michaud, R., Richards, K. W. and Waddington,
J. 1980. Growing and managing alfalfa in Canada. Agriculture
Canada Publ. 1705. Ottawa, ON.
Kiesselbach, T. A., Russel, J. C. and Anderson, A. 1929. The
significance of subsoil moisture in alfalfa production. J. Amer.
Soc. Agron. 21: 241–268.
Pelton, W. L. and Korven, H. C. 1969. Evapotranspiration estimates
in a semiarid climate. Can. Agric. Eng. 11(2): 50–61.
Sheaffer, C. C., Tanner, C. B. and Kirkham, M. B. 1988.
Alfalfa water relations and irrigation. Pages 373–409 in A. A.
Hanson, D. K. Barnes, and R. R. Hill, Jr., ed. Alfalfa and alfalfa
improvement. Agronomy no. 29. American Society Agronomy,
Madison, WI.