Agriculturul S.wtems. Vol. 57, No. 1. yp. 101-l 14,...
Agriculturul S.wtems. Vol. 57, No. 1. yp. 101-l 14, 19%
6:) 1998 Elsevier Science Ltd. Al1 rights reserved
Printed in Great Britain
Pli: s0308-521x(97)00072-3
0308-521X/98 $19.00 t-O.00
Adaptability of Improved Rice Varieties in Senegalt
Samba Sall,” David Normaqh* & Allen M. Featherstoneb
“Senegalese Agricultural Research Institute, BP 34, Ziguinchor, Senegal
“Agricultural Economies Department, Waters Hall, Kansas State University, Manhattan.
KS 66506, USA
(Received 17 June 1997; accepted 20 October 1997)
ABSTRACT
Using yield gap and adapta ility (modzfîed stability) analyses, this study
evaluated the potential relej
nce of improved rice varieties released by, or
soon to ht? released by, th
Senegalese Agricultural Research Institute
(ISRA) in the Casamance ?egion of Senegal. Results from researcher-
managed, onfarm trials ina ated that level of fertiliser applied accounted
,for most cif the dtzerence
t Yield Gap II (i.e. the difference between
potential (experimental] an
actual yields ut the farm level), followed by
variety. Also, the adaptabil v analysis results indicated that most qf the
improvcd varieties performt
poorer than local varieties under poor pro-
duction environments but be er than local varieties under good production
environments. We conclude
that the rice breeding programme needs to
puy greater attention to de loping varieties better adapted to the varied
production environments ur er which farmers operate. {3 1998 Elsevier
Science Ltd. ,411 rights reser
?d
TRODUCTION
The Casamance area is locatel in southern Senegal. The climate is classified
as dry Guinean Savanna. Rit
is an important source of nutrition in this
traditional rain-fed rice growi ; region. The organization of rice production
*TO whom correspondence should b addressed.
Vontribution No. 98-117-J from the Lansas Agricultural Experiment Station.
101
- .,_*__. - .----. _- -.-- - --.--
102
S. Sali, D. Normun, A. M. Featherstone
is tied closely to the social and religious stratifications of the clom nmi.ty
(Linares, 1981). A wide range of indigenous rice varieties
1”
traditional y have
been grown in the area, although farmers recently have been very wi ling to
experiment with new varieties. Considerable diversity exists in the ar a both
in terms of farming/cropping systems and the conditions under whic d rice is
grown. Three types of rice-based systems occur (i.e. aquatic, phreaiic and
rain-fed), with the variety grown being determined by the position of the plot
on the toposequence. Superimposed on these differences is considerable
variability in soi1 quality (and management of the rice plots. Factors influ-
encing management include differences in resources between farming house-
holds and differences in the managerial ability of farmers. Differ’enoes also
result from stochastic events over which farmers have little control. For
example, major trends or constraints that have emerged since 19’73 include
persistent patterns of drought and salt intrusion, which brings into question
the continued suitability of many of the local varieties.
Because of farmers’ risk aversion, the probability of widespread adoption
of a new variety Will be enhanced if that variety shows stable yield superi-
ority over a range of production environments. The introduction of new rice
cultivars with high yields and short duration has been viewed as a strategy to
increase and secure rice production in the area. Accordingly, researah pro-
grammes and extension projects have focused on the development and dif-
fusion of improved rice varieties through multilocational, on-faim, and
adoption trials. The purposes of this study were to investigate the perfor-
mance of new rice varieti.es across the different production environments
faced by producers in the Casamance and to make recommendations for
their improvement.
A P P R O A C H
Three major techniques were used in this study: yield gap analysis, adapt-
ability analysis and fertilizer response analysis.
Yield gap analysi(
Yield gap analysis, developed by the International Rice Research Insi tute in
the 1970s has be n used extensively to measure and analyse the deterr iinants
of yield gaps in f rmers’ fields in Southeast Asia where high yielding v irieties
have been adopt d (De Datta et al., 1978). Because the main focu i is on
Yield Gap II (ie. the difference between potential [experimental] anc actual
yields at the fart level), it. is essentially on-farm testing after the fac t. This
yield gap cari be:nterpreted in either of two ways: first, as represent ng the
Adaptability of improved rice varieties in Senegal
103
potential production increment above farmers’ yield levels or, second, as an
indicator of more fundamental problems with the varieties themselves,
particularly their poor adaptation to farmers’ environmental or managerial
constraints. If the principal factors causing the yield gap cari be resolved
practically at the farm level, greater weight should be given to the first
interpretation, and steps to resolve these factors should be considered as
necessary complements for the successful extension of the new varieties.
However, if these factors cannot be resolved practically, then the second
interpretation is relevant and the research objectives and methods that are
producing such poorly adapted varieties must be reconsidered. The general
objective was to identify the factors that explain the difference between
actual and potential rice yields in selected environments. The contributions
of test factors (variety, fertilizer, pest control) to Yield Gap II were deter-
mined by means of a factorial tria1 using a modified version of the design
developed by De Datta et al. (1978). The experimental levels were the farm-
er’s level (Level 1), consisting of local variety, no fertilizer and no pest con-
trol, and the recommended level (Level 2), consisting of improved variety,
150 kg/ha of fertilizer (urea) and pest control. Factor levels for each treat-
ment are shown in Table 1, and the accompanying equations were used to
calculate the contributions of the three factors to Yield Gap II.
TABLE 1
Treatments [ncluded in the Yield Gap Tria1
Treatment
Variety
Fertilizer
Prst control
Al1 other fkctors
1
1
1
1
2
2
1
1
3
1
2
1
4
1
1
2
5
2
2
1
6
2
1
2
7
1
2
2
8
2
2
2
Based on these treatment descriptions and using Y, to represent the yield of treatment i, the
formula is:
Yield gap = Y, - Y1
U-1)
Contribution oj’variety = y2 + y5 + y6 + y8 Y1 + Y3 + Y4 + Y7
4
-
4
(1’2)
Contribution of fertilizer = y3 + y, + y7 + y8
y 1 + y 2 +
Y4 + Y6
-~
(1’3)
4
4
Contribution of pest contrbl = y4 + y6 + y7 + y, Y1 + Y, + Y 3 + Y5
4
-
4
o-4)
1 0 4
S. Sall, Il. Norrnan, A. M. Feathwstone
Adaptability analysis
Adaptability (formerly called modified stability) analysis techniques have
been developed to compare the performances of cultivars across djfyerent
environments. The techniqjue involves regressing the yield of each variety at
each site against the mean yield of a11 varieties at each site (Hildebrand, 1983;
Hildebrand and Russell, 1996). The mean yield then represents a type of
environmental index. A site where yields are low, due either to management
or to physical site characteristics, is considered a poor environment, and a
site with high yields is a good environment. With this definition, environment
is measured as a continuous proxy variable across the range of average
yields. In this study, we categorized and grouped the improved varieties
according to four standard stability types: Type A occurs when the yield of
an improved variety is superior to the local variety across a11 enviromnents;
Type B is when the improved variety is superior to the local variety i,n poor
environments but is inferior in good environments; Type C occurs when the
yield of an improved variety is inferior to that of the local variety in poor
environments but superior in good environments and Type D represents the
case in which the improved variety is inferior over a11 environments. Because
the level of fertilizer is likely to be one important factor determining the
quality of the environment, two regression models were estimated at two
levels of fertilizer use, namely one at 0 and one at 100 units of N. The
regression estimated was as follows:
where
Kiki= yields for the improved variety i and the local variety k at locat j on .j,
.Zi= the average yield of a11 varieties at 1ocation.j
Xi= a dummy variable that takes the value 1 for the improved varic $1 ly and
0 otherwis’ei
E= a random dariable with an assumed normal distribution
Fertilizer respons analysis
4
Local varieties habe evolved over generations and have become well a laptcd
to environments
generally have received little in the way of soi1 2 mend-
ic fertilizer). In contrast, most improved varieties end to
favourable environments in which applica ion of
norm. Therefore, we estimated fertilizer t-t sponse
Aduptability of improved rice varieties in Senegal
105
curves to answer two specific questions. The first is whether the improved
varieties are more responsive to inorganic fertilizer when other inputs reflect
farmers’ management. That is, are their fertilizer response curves steeper than
that of the local variety when facing on-farm stresses? Second, do the fertilizer
response curves cross, such that the ordering of varieties with respect to yield
changes significantly between low and high fertilizer levels (i.e. the so-called
cross-over effect)? A J-test was performed to determine the correct specific-
ation of the response function. Two specifications were tried, the quadratic
functional form and the three-halves functional form. The three-halves
functional form, which also permitted the use of nested tests of hypotheses
with respect to in-put use, was selected (Traxler and Byerlee, 1993). TO address
the two questions posed, we fitted the following regression model:
where
Y- yield measured in kg/ha
X= fertilizer application measured in kg/ha
Di -= 1 for improved variety and 0 for local variety
E - a random variable with an assumed normal distribution
We tested two hypotheses. First, that nitrogen responses are the same for
the improved varieties and the local varieties (h4 = h5 = 0) and, second, that
outputs without nitrogen are the same for improved and local varieties
(h3 = 0).
DATA
Data used in the analysis were assembled during the 1982.-87 period under
the auspices of the Senegalese Institute for Agricultural Research (ISRA).
Data on farmers’ cultivation practices collected by the on-farm farming sys-
tems research (FSR) team indicated that the type of variety used, the level of
fertilizer used and the degree of protection against pests used in the rice
research programme differed significantly from those used by farmers. Dur-
ing 1986, the ISRA rice programme initiated a 2-year, researcher-managed
and farmer-implemented on-farm tria1 to estimate the major determinants of
Yield Gap II for rice. The tria1 was conducted on five farms in each of five
villages in the study area. The three factors were tested at the station and
farmers’ levels in a factorial design with three interna1 replications. Different
varieties (a total of 10) representing aquatic, phreatic and rain-fed rice and
106
S. Sali, D. Normun, A. M. Featherstone
,
local varieties were evaluated. The fertilizer levels were zero nitrogen and
150 kg/ha of urea topdressed (the recommended dose). Protection against
pests involved a single dose of the recommended fungicide (tricyclazole).
During the same period anld using the same research sites and the station for
reference purposes, the rice programme also conducted another $!-year,
researcher-managed and farmer-implemented tria1 to study the responses of
different rice varieties to nitrogen. For this purpose, a split-plot desigb with
four levels of urea topdressed (0, 50, 100 and 150 kg/ha) in three replications
was used. Not a11 results from this comprehensive tria1 progra.mme a:re
reported in this paper, only those relating to varieties that either had dlreacly
been recommended officially or were likely to be recommended in the near
future. The rice varieties considered in different parts of the analysis were
DJ684D (aquatic), DJ125 19, IKP and TOX728 (phreatic), and IRA.Tl12
and IRAT 10 (rain-fed).
RESULTS AND DISCUSSIONS
Yield gap analysis
Data in Table 2 collected by the FSR team show that most of the v rieties
tested by ISRA since 1982 have experienced yield losses between 35 % and
70% when transferred from research station trials (i.e. research-m naged
and research-implemented:) to farmers’ management and implemer :atio:n,
when no inorganic fertilizer was used. Inorganic fertilizer is used ra ely by
farmers on the rice trop. The relative decline in yield generally appe; red 1.0
be less for the twd local varieties (i.e. Ablaye Mano and Barafita).
In the yield gap trial, the three factors (i.e. fertilizer, variety ar d pest
control; Table 3) explained most of the difference between the far n and
TABLE 2
Total Yield Gap (kg/ha) for Rice Varieties between Trials at Djibelor Station and 1 a rmers’
Tests in Casamance, 1982-86
- - -
Rice type
Variety
Station
Farmer
Dlferencr (‘W)
- - -
Improved aquatic
DJ684D
2741
1551
43
Improved phreatic
DJ12519
4039
2564
36
IKP
2454
1026
69
TOX728
3208
1232
61
Local varieties
Ablayo Mano
1443
1433
7
Barafita
2513
2036
19
Station yields were es
under research-managed and research-implemented CO dlitions,
whereas farmers’
were taken under farmer-managed and farmer-implemented CO diitions.
Adaptability of improved rice varieties in Senegal
107
TABLE 3
Yield Gap II Tria1 and Contribution of Each Factor (kg/ha)
Rice type
Variety
Yield Gap II
Variety
Fertilization Pest Control
Residual
Aquatic
DJ684D
3731
1064 (28.51)
2300 (61.64)
305 (8.17)
62 (1.68)
Phreatic
DJ12519
3905
1117 (28.60)
2700 (69.14)
71 (1.81)
17 (0.45)
IKP
3683
483 (13.11)
2867 (77.84)
202 (5.48)
129 (3.57)
TOX728
3550
833 (23.46)
1980 (55.77)
214 (6.03)
520 (14.74)
The numbers in parentheses are the contributions of each factor in percentage terms.
recommended levels, thus confirming the findings of the FSR team. The
relative contributions of each factor were fairly constant across varieties. On
average, fertilizer explained 61--78% of yield variation, variety explained 13-
30% of yield variation and pest control explained only 2-8% of the yield
variation. Implications arising out of these results are as follows:
?? The results in Table 2 implied that with no fertilizer, the yield gap in
actual and relative terms, was lower for local than for improved vari-
eties. Because fertilizer was found to be the major determinant of the
yield gap for improved varieties (Table 3), this implies that local vari-
eties are relatively better adapted to zero fertilizer levels. There are a
number of reasons why farmers are reluctant to use chemical fertilizer
on their rice fields; for example, inadequacies in the credit programme,
production risks from recurrent drought periods and major concerns
about its toxic effect on fish. The obvious issue is the relevancy of a
research programme that focuses excessively on improved varieties
using only purchased inputs. Given the practical realities in the area,
there would be merit in determining the potential substitutability of
organic matter (i.e. manure, ashes, trop residues) for the expensive
inorganic fertilizer. Another possible practical implication is to screen
improved varieties under soi1 fertility conditions more typical of the
farmers’ level, a topic we Will discuss further.
?? The second most important determinant of the yield gap, in actual and
relative terms, was the rice variety used. The results indicated that
farmers theoretically could increase rice production by an average of
500-I 100 kg/ha through using an improved variety instead of their local
variety. Whether or not the superiority of improved varieties over local
varieties would be maintained under practical farming conditions is
another question. Once again, we Will discuss this issue further.
?? Pest control appeared to contribute least to the yield gap both in actual
and relative terms. In fact, the lack of major significance of this factor
seems to be confirmed by recent studies (ISRA-DERBAC, 1993)
108
S. Sd, D. Normun. A. M. Ftwtherstonc~
indicating that level of pest infestation only amounted to between 2%
and 8% of a11 rice plants on farmers’ fields.
Adaptability analysis
Turning to the adaptability analysis designed to assess the rob.ust ms o f
varieties across different environments, the results indicated that 1 ith no
inorganic fertilizer, a11 thle improved aquatic and phreatic rice 1 u-ieti es
yielded more under better rather than poor production environme ts (i.e.
Type C stability; Table 4). A similar relationship was found to exist or the
phreatic varieties under the high fertility level (i.e. 100 units of n rogen;
Table 5), although of course the average yield was much higher tha
when
no fertilizer was applied. Also, the statistical significance of the I levant
variables when no inorganic fertilizer was applied and the lack of sign icance
at the high fertilizer level implied that the improved varieties perfor n even
less satisfactorily under very poor production environments. Althol gh the
improved varieties obviously are very responsive to inorganic fertilize , other
elements also influence the quality of the production environment. .Jnfor-
tunately, data were not available to determine exactly what these w( -e, but
they could includie not only physical factors (e.g. soi1 type and inher nt soi1
quality including organic matter, weed problems), but also factors t lat are
more socioeconomic in nature such as managerial ability and diff’erc rces in
accessibility to resources (e.g. labour available for farm operation!
land/
labour ratio, availability of cash). Given that the adaptability analy is tria1
was implemented under researcher-managed and farmer-implementc 1 con-
ditions, the production environment likely was influenced more b:y I iysical
factors than by socioeconomic factors. However, under farmer-manai
:d and
farmer-implemented conditions, the relative influence of socioeconor ic fac-
tors likely would be greater. Indirect evidence in support of this
:an ‘be
obtained from comparing the average yields of the varieties that ap lear in
both Tables 2 and 4. The average incremental yield was 86% highe under
TABLE 4
Adaptab jlity Analysis for Improved Rice Varieties at Zero Fertilizer
_--
Rice ty,ve
Intempt (hz)
Slope (bl)
Stability type
I’ield (kg/ .r,J II’
_--
Aquatic
;g$)---
-t
C
1800
68
Phreatic
-***
+ ***
c
3066
98
IKP /
-***
+ ***
C
2717
97
TOX72$
-***
+ ***
c
3016
04
Rain-fed
IRATl 2
+
A
2280
‘15
IRATl P
2428
-75
***,o,G*
+
A
**p<o.o5;
Aduptuhility qf improved rice wrieties in Senegul
109
TABLE 5
Adaptability Analysis for Improved Rice Varieties at 100 N Fertilizer
Rice type
Variet-v
Intercept (b2)
S~O~C~ (bd
Stubi1it.v type
Y i e l d lkg/ha) R2
Aq uatic
DJ684D
-
+
c
4265
77
Phreatic
DJ12519
-
i-
c
4973
90
IKP
-
+
c
4682
75
TOX728
-
+
c
5014
55
Rain-fed
IRATll2
-
+
c
3331
55
IRATIO
+
-_
B
2423
56
the researcher-managed and farmer-implemented conditions (Table 4) than
under the farmer-managed and farmer-implemented conditions (Table 2).
Turning to the improved rain-fed varieties, their performance was superior
to that of the local variety with no fertilizer across a11 production environ-
ments (Le. Type A stability; Table 4). These relationships were not main-
tained under the high level of inorganic fertilizer (Table 5). However, these
results need to be interpreted with caution. Farmers traditionally have culti-
vated rice on newly cleared land. After 3 years or SO, farmers usually had to
begin contending with fertility and weed problems and, thus, tended to move
to other plots. However, with increasing population densities, this is no
longer feasible. Thus, the favourable situation depicted for improved rain-
fed rice varieties under zero fertilizer levels likely Will now be impossible to
emulate in practice. Thus, if rain-fed rice varieties are to be grown, they are
likely to experience conditions more analogous to those shown in Table 5.
Those results are much less promising, and, in fact, the Type B stability
shown by IRATlO arises because, at high levels of nitrogen application, it
becomes more sensitive to a particular type of rice blast (pyriculariose;
Mbodj, 1991). In any case, the Senegalese government has been very reluc-
tant to recommend widespread dissemination of rain-fed rice varieties in the
Casamance. The reason for this is to encourage greater diversification of the
farming systems away from rice, in order to reduce production risk. Rice is
obviously the most desirable trop for the lowlands (i.e. aquatic and phreatic
conditions), but the upland is suitable for other crops such as maize, sor-
ghum, cowpeas and groundnuts.
Fertilizer response analysis
The results shown in Tables 6 and 7 indicate rejection of the hypothesis that
improved and local varieties respond in an analogous manner to the appli-
cation of nitrogen fertilizer; that is, the F value was 10.40 with (2, 354)
degrees of freedom. Thus, improved varieties and local varieties under
farmers’ circumstances have different response curves. With respect to the
ZI -_I---
-
----
---
110
S. Sali, D. Norman, A. M. Featherstone
TABLE 6
Estimates of Response Functions for Rice Varieties”
Variable
Aquatic
Phrentic
Rnin-fed
-_
---+----
DJ684D
IKP
D/l2519
IRAT112
IRATIO
x
32.53
32.53
35.41
18.53
1.8.52
(4.16)
(2.49)
(9.11)
(1.47)
(1.61)
Xl.5
-- 1.46
-1.47
-1.67
-0.96
-10.96
w:;;’
(-1.40)
(-5.39)
(-0.95)
(-1.04)
Dl
- 187.03
1092.00
--157.56
-121.41
(0.69)
(-0.53)
(10.40)
(-0.46)
(-0.39)
DIX
9.58
--14.28
2.66
25.18
23.92
(0.86)
(-0.77)
(0.40)
(1.41)
(1.46)
D$f’ .5
-0.70
1.16
-0.09
-1.51
-1.07
(-0.79)
(0.78)
(-0.21)
(-- 1.06)
(~0.82)
Intercept
2417.50
2417.00
2535.60
1397.00
1397.00
(16.18)
(9.67)
(34.14)
(5.80)
66.35)
R2
92
84
92
30
46
“The t-statistics are in parentheses.
TABLE 7
Results of Hypothesis Testing for Fertilizer Response Curves
Nul1 hypothesis
Parumeter restriction
Test srrrtis!tic
Equal response to N
bq=b5=0
F= 10.40**
Equal response at N = 0
Aquatic rice
DJ684D
h3=0
t=z0.69
Phreatic rice
DJ12519
b,=O
t = 10.40**
IKP
h.3 = 0
t == -0.53
TOX728
h3=0
t=7.83’k*
Rain-fed rice
IRATl12
b3=0
t = -0.46
TRATIO
b3=0
t = -0.39
** Indicates rejection of hypothesis at p < 0.05.
a1 respon.ses of improved and local varieties at zero ikvel of
rejected in the case of two phreatic varieties, DJ12. 19 and
for the remainder. These results are somewhat arn. biguous
ary to what we showed earlier, that the yields of th :!Se t\\VO
are hig:her than those of the local varieties at zerc fertili-
zer level. Howe
for the remaining improved varieties, a cross-ov( r effect
improved varieties performing better than local 7 ariettes
as fertilizer leve
reased, thus confirming a Type C stability.
ate that the ordering of a set of varieties accor ling to
y identical at high and low levels of fertilizer app ication
Adaptability of improved rice varieties in Senegal
111
and may be affected by other physical and socioeconomic factors that
determine the quality of the production environment. Certainly, under
poorer production environments including a zero inorganic fertilizer level, the
two local rice varieties, Ablaye Mano and Barafita, are competitive in terms
of their yields and are especially vigorous at emergence (Posner et al., 1991).
The rice programme has used a high level of inorganic fertilizer application
in its screening process. As a result, lodging problems have been minimized,
although efforts to eliminate the problem of pyriculariose do not always
appear to have been successful. Another very important implication borne
out by the results of this study is that the variety screening process also
appears to be inadequate in providing improved rice varieties well adapted to
less favourable production environments, including the farmer’s common
strategy of not using any inorganic fertilizer. On balance, most of the existing
improved rice varieties appear to be most appropriate for those farmers
working in favourable production environments, including the extensive use
of inorganic fertilizer. We conclude that, because less than 5% of the farmers
applied inorganic fertilizer (ISRA-DERBAC, 1993), the production envi-
ronments under which much of the rice is produced are less than ideal.
Therefore, widespread adoption of improved rice varieties Will require vari-
eties well adapted to such suboptimal growing conditions. Thus, some
adjustments are needed in the approach of the rice breeding programme,
particularly with respect to the screening process and greater collaboration
with agronomists, soi1 scientists and the FSR team in order to identify
acceptable ways of improving the production environments of farmers.
CONCLUSION
In this paper, three different methods have been used to assess the potential
suitability of rice varieties under varying production environments in the
Casamance region of Senegal. Al1 the approaches bave their place. The yield
gap approach, although intuitively appealing to station-based scient,ists in
helping to highlight the major factors (especially physical/technical factors)
contributing to the difference between experiment station and farm level
yields, does not by itself indicate differences that arise as a result of variation
in actuai production environments. It in essence represents a single point on
a production function. The production function (i.e. fertilizer response)
approach cari help assess the major determinants of yield (e.g. fertilizer and
variety) but the values of the coefficients on the variables in the function Will
reflect the levels and qualities of the non-experimental variables. However, it
is particularly sui ted to assessing the economic optima for different inputs for
specific production environments. In terms of ability to assess robustness of
112
S. Sall, B. Norman, ,4. M. Featlzerstone
the varieties across different practical production environments and; when
necessary, develop multiple recommendations, adaptability analysis is t’he
most suitable. However, to fully exploit the potential of adaptability, analy-
sis, it would be desirable to go one step further than was possible in this
paper, namely to identify Imore specifically the determinants of the different
production environments ;as represented by different values of the environ-
mental index.
The results of this study demonstrate the irrelevancy of a single approach
to recommending the same rice varieties for a11 production environments
found in the region. The analysis shows the need for multiple recommenda-
tions appropriate for fanmers operating in different production environ-
ments. In the long run, major breakthroughs in rice production Will be
realized only through substantial improvements in the physical and socio-
economic production environments of the majority of small farmers. How-
ever, this is likely to require a sustained incremental approach. Therefore, we
believe the rice programme needs to focus its activities on two com-
plementary strategies:
?? The rice programme should produce varieties adapted to thel varied
production environments under which farmers operate. In FSR par-
lance, this requires the recognition of more than one research or
recommendation domain for the improved rice varieties and a. scieening
process that takes the different production environments into account.
Because, for reasons discussed subsequently, breeders likely will have
difficulty developing improved rice varieties with Type A stability across
the range of productilon environments found in the region, a strategy of
breeding some with a Type B stability and some with a Type C smbility
might be more appropriate.
?? The rice programme should develop close collaborative relationships
with agronomists, soils scientists and the FSR team to identify the
determinants of the different production environments under pi-actical
farming con’ itions and, in cooperation with farmers, design an evalu-
P
ate relevant: strategies for their improvement. Also, improve ent in
organic mat’er and nlot relying too heavily on purchased inputs ‘s likely
to be more elevant for farmers operating in poor production t nviron-
ments. In developing high yielding varieties very responsive to ir 3rganic
fertilizer, brteders have tended to emphasize grain yield at the :xpense
of stover (i.4. biomass). Varieties with Type A stability are likely to have
such charac eristics. Under poor production environments, rice 1arieties
1
that have a higher stover/grain ratio are likely to be more effc rtive in
contributin
organic matter to the soil. This is more likely tc be the
”
case with v, rieties of Type B stability. Also under poor prc luction
Adaptability qf’improved rire varieties in Senegal
113
environments, seeking practical ways of incorporating legumes into the
cropping system would have merit.
In essence, what we are advocating is that ISRA should move away from a
single blanket recommendation for rice varieties to one that emphasizes a
smorgasbord of different rice varieties, accompanied by information as to
when they work best (i.e. conditional and targeting information; Norman et
al., 1995), from which farmers cari Select. Such an approach recognizes that
limited-resource farmers live and work on farms characterized by a high
degree of both biophysical and socioeconomic diversity. It also recognizes
that farmers are rational and have the best knowledge about their own pro-
duction environments. Consequently, we believe the close collaboration that
we are advocating between station- and farm-based researchers (i.e. com-
modity-based programmes and FSR teams) and farmers is critically impor-
tant, not only in developing relevant improved rice varieties but also in
identifying practical strategies for improving the farmers’ production envi-
ronments. Also, such collaboration is needed because our research (Sall,
1997) has shown that farmers’ decisions as to whether or not they Will adopt
improved varieties also involve criteria other than yield.
REFERENCES
De natta, S.K., Gomez, K.A., Herdt, R.W. and Barker, R. (1978) A Hundbook on
the MethodologJl
for an Integrated Experiment-Survey on Rice Yield Constraints.
The International Rice Research Institute, Los Banos, Laguna, Philippines.
Hildebrand, P.E. (1983) Modified stability analysis of farmer managed, on-farm
trials. Agronomy Journal 76, 27 l-274.
Hildebrand, P.E. and Russell, J.T. (1996) .4daptability
Analysis: A Methodfor the
Design and Interpretation of On-Farm Research E.xtension. Iowa State University
Press, Ames, IA.
ISRA-DERBAC (1993) Rapport d’Activite,
Campagne Agricole 1991-1992, Conven-
tion Recherche d’Accompagnement.
Institut Senegalais Recherche Agronomique
and Projet de Development Integre de la Basse Casamance SRA, Ministere du
Development Rural et de 1’Hydraulique.
Linares, O.F. (1981) From tidal swamp to inland valley: on the social organization
of wet rice cultivation among the Diola of Senegal. Ajiica 51, 21-32.
Mbodj, Y. (1991) Les Recherches Rizicoles en Cusamance: Situation en 1991. Centre
Recherche Agronomique, Djibelor.
Norman, D.W., Worman, F.D., Siebert, J.D. and Modiakgotla, E. (1995) The
farming systems approach to development and appropriate technology genera-
tion. FAO Farm Systems Management Series No. 10. FAO, Rome. Italy.
Posner, J., Kamuanga, M. and Mamadou L. (1991) Lowland cropping systems in the
Lower Casamance of Senegal: results of four years of agronomie research.
M.S.U. International Development Paper No. 30. Department of Agricultural
Economies, Michigan State University, East Lansing, MI.
114
S. Sar’l, D. Norman, A. M. Featherstone
Sall, S. (1997) A technology-characteristic approach to adoption: the ase of
improved rice varieties in Southern Senegal. Ph.D. Dissertation. Depart kent of
Agricultural Eqonomics, Kansas State University, Manhattan, KS.
I
,,
Traxler, G. and Byerlee, D. (1993) A joint-product analysis of the adoptron of
modern cereal varieties in developing countries. Americun Journwl of’ Ag.C-
cultural Economies 75, 98 l-989.
6:) 1998 Elsevier Science Ltd. Al1 rights reserved
Printed in Great Britain
Pli: s0308-521x(97)00072-3
0308-521X/98 $19.00 t-O.00
Adaptability of Improved Rice Varieties in Senegalt
Samba Sall,” David Normaqh* & Allen M. Featherstoneb
“Senegalese Agricultural Research Institute, BP 34, Ziguinchor, Senegal
“Agricultural Economies Department, Waters Hall, Kansas State University, Manhattan.
KS 66506, USA
(Received 17 June 1997; accepted 20 October 1997)
ABSTRACT
Using yield gap and adapta ility (modzfîed stability) analyses, this study
evaluated the potential relej
nce of improved rice varieties released by, or
soon to ht? released by, th
Senegalese Agricultural Research Institute
(ISRA) in the Casamance ?egion of Senegal. Results from researcher-
managed, onfarm trials ina ated that level of fertiliser applied accounted
,for most cif the dtzerence
t Yield Gap II (i.e. the difference between
potential (experimental] an
actual yields ut the farm level), followed by
variety. Also, the adaptabil v analysis results indicated that most qf the
improvcd varieties performt
poorer than local varieties under poor pro-
duction environments but be er than local varieties under good production
environments. We conclude
that the rice breeding programme needs to
puy greater attention to de loping varieties better adapted to the varied
production environments ur er which farmers operate. {3 1998 Elsevier
Science Ltd. ,411 rights reser
?d
TRODUCTION
The Casamance area is locatel in southern Senegal. The climate is classified
as dry Guinean Savanna. Rit
is an important source of nutrition in this
traditional rain-fed rice growi ; region. The organization of rice production
*TO whom correspondence should b addressed.
Vontribution No. 98-117-J from the Lansas Agricultural Experiment Station.
101
- .,_*__. - .----. _- -.-- - --.--
102
S. Sali, D. Normun, A. M. Featherstone
is tied closely to the social and religious stratifications of the clom nmi.ty
(Linares, 1981). A wide range of indigenous rice varieties
1”
traditional y have
been grown in the area, although farmers recently have been very wi ling to
experiment with new varieties. Considerable diversity exists in the ar a both
in terms of farming/cropping systems and the conditions under whic d rice is
grown. Three types of rice-based systems occur (i.e. aquatic, phreaiic and
rain-fed), with the variety grown being determined by the position of the plot
on the toposequence. Superimposed on these differences is considerable
variability in soi1 quality (and management of the rice plots. Factors influ-
encing management include differences in resources between farming house-
holds and differences in the managerial ability of farmers. Differ’enoes also
result from stochastic events over which farmers have little control. For
example, major trends or constraints that have emerged since 19’73 include
persistent patterns of drought and salt intrusion, which brings into question
the continued suitability of many of the local varieties.
Because of farmers’ risk aversion, the probability of widespread adoption
of a new variety Will be enhanced if that variety shows stable yield superi-
ority over a range of production environments. The introduction of new rice
cultivars with high yields and short duration has been viewed as a strategy to
increase and secure rice production in the area. Accordingly, researah pro-
grammes and extension projects have focused on the development and dif-
fusion of improved rice varieties through multilocational, on-faim, and
adoption trials. The purposes of this study were to investigate the perfor-
mance of new rice varieti.es across the different production environments
faced by producers in the Casamance and to make recommendations for
their improvement.
A P P R O A C H
Three major techniques were used in this study: yield gap analysis, adapt-
ability analysis and fertilizer response analysis.
Yield gap analysi(
Yield gap analysis, developed by the International Rice Research Insi tute in
the 1970s has be n used extensively to measure and analyse the deterr iinants
of yield gaps in f rmers’ fields in Southeast Asia where high yielding v irieties
have been adopt d (De Datta et al., 1978). Because the main focu i is on
Yield Gap II (ie. the difference between potential [experimental] anc actual
yields at the fart level), it. is essentially on-farm testing after the fac t. This
yield gap cari be:nterpreted in either of two ways: first, as represent ng the
Adaptability of improved rice varieties in Senegal
103
potential production increment above farmers’ yield levels or, second, as an
indicator of more fundamental problems with the varieties themselves,
particularly their poor adaptation to farmers’ environmental or managerial
constraints. If the principal factors causing the yield gap cari be resolved
practically at the farm level, greater weight should be given to the first
interpretation, and steps to resolve these factors should be considered as
necessary complements for the successful extension of the new varieties.
However, if these factors cannot be resolved practically, then the second
interpretation is relevant and the research objectives and methods that are
producing such poorly adapted varieties must be reconsidered. The general
objective was to identify the factors that explain the difference between
actual and potential rice yields in selected environments. The contributions
of test factors (variety, fertilizer, pest control) to Yield Gap II were deter-
mined by means of a factorial tria1 using a modified version of the design
developed by De Datta et al. (1978). The experimental levels were the farm-
er’s level (Level 1), consisting of local variety, no fertilizer and no pest con-
trol, and the recommended level (Level 2), consisting of improved variety,
150 kg/ha of fertilizer (urea) and pest control. Factor levels for each treat-
ment are shown in Table 1, and the accompanying equations were used to
calculate the contributions of the three factors to Yield Gap II.
TABLE 1
Treatments [ncluded in the Yield Gap Tria1
Treatment
Variety
Fertilizer
Prst control
Al1 other fkctors
1
1
1
1
2
2
1
1
3
1
2
1
4
1
1
2
5
2
2
1
6
2
1
2
7
1
2
2
8
2
2
2
Based on these treatment descriptions and using Y, to represent the yield of treatment i, the
formula is:
Yield gap = Y, - Y1
U-1)
Contribution oj’variety = y2 + y5 + y6 + y8 Y1 + Y3 + Y4 + Y7
4
-
4
(1’2)
Contribution of fertilizer = y3 + y, + y7 + y8
y 1 + y 2 +
Y4 + Y6
-~
(1’3)
4
4
Contribution of pest contrbl = y4 + y6 + y7 + y, Y1 + Y, + Y 3 + Y5
4
-
4
o-4)
1 0 4
S. Sall, Il. Norrnan, A. M. Feathwstone
Adaptability analysis
Adaptability (formerly called modified stability) analysis techniques have
been developed to compare the performances of cultivars across djfyerent
environments. The techniqjue involves regressing the yield of each variety at
each site against the mean yield of a11 varieties at each site (Hildebrand, 1983;
Hildebrand and Russell, 1996). The mean yield then represents a type of
environmental index. A site where yields are low, due either to management
or to physical site characteristics, is considered a poor environment, and a
site with high yields is a good environment. With this definition, environment
is measured as a continuous proxy variable across the range of average
yields. In this study, we categorized and grouped the improved varieties
according to four standard stability types: Type A occurs when the yield of
an improved variety is superior to the local variety across a11 enviromnents;
Type B is when the improved variety is superior to the local variety i,n poor
environments but is inferior in good environments; Type C occurs when the
yield of an improved variety is inferior to that of the local variety in poor
environments but superior in good environments and Type D represents the
case in which the improved variety is inferior over a11 environments. Because
the level of fertilizer is likely to be one important factor determining the
quality of the environment, two regression models were estimated at two
levels of fertilizer use, namely one at 0 and one at 100 units of N. The
regression estimated was as follows:
where
Kiki= yields for the improved variety i and the local variety k at locat j on .j,
.Zi= the average yield of a11 varieties at 1ocation.j
Xi= a dummy variable that takes the value 1 for the improved varic $1 ly and
0 otherwis’ei
E= a random dariable with an assumed normal distribution
Fertilizer respons analysis
4
Local varieties habe evolved over generations and have become well a laptcd
to environments
generally have received little in the way of soi1 2 mend-
ic fertilizer). In contrast, most improved varieties end to
favourable environments in which applica ion of
norm. Therefore, we estimated fertilizer t-t sponse
Aduptability of improved rice varieties in Senegal
105
curves to answer two specific questions. The first is whether the improved
varieties are more responsive to inorganic fertilizer when other inputs reflect
farmers’ management. That is, are their fertilizer response curves steeper than
that of the local variety when facing on-farm stresses? Second, do the fertilizer
response curves cross, such that the ordering of varieties with respect to yield
changes significantly between low and high fertilizer levels (i.e. the so-called
cross-over effect)? A J-test was performed to determine the correct specific-
ation of the response function. Two specifications were tried, the quadratic
functional form and the three-halves functional form. The three-halves
functional form, which also permitted the use of nested tests of hypotheses
with respect to in-put use, was selected (Traxler and Byerlee, 1993). TO address
the two questions posed, we fitted the following regression model:
where
Y- yield measured in kg/ha
X= fertilizer application measured in kg/ha
Di -= 1 for improved variety and 0 for local variety
E - a random variable with an assumed normal distribution
We tested two hypotheses. First, that nitrogen responses are the same for
the improved varieties and the local varieties (h4 = h5 = 0) and, second, that
outputs without nitrogen are the same for improved and local varieties
(h3 = 0).
DATA
Data used in the analysis were assembled during the 1982.-87 period under
the auspices of the Senegalese Institute for Agricultural Research (ISRA).
Data on farmers’ cultivation practices collected by the on-farm farming sys-
tems research (FSR) team indicated that the type of variety used, the level of
fertilizer used and the degree of protection against pests used in the rice
research programme differed significantly from those used by farmers. Dur-
ing 1986, the ISRA rice programme initiated a 2-year, researcher-managed
and farmer-implemented on-farm tria1 to estimate the major determinants of
Yield Gap II for rice. The tria1 was conducted on five farms in each of five
villages in the study area. The three factors were tested at the station and
farmers’ levels in a factorial design with three interna1 replications. Different
varieties (a total of 10) representing aquatic, phreatic and rain-fed rice and
106
S. Sali, D. Normun, A. M. Featherstone
,
local varieties were evaluated. The fertilizer levels were zero nitrogen and
150 kg/ha of urea topdressed (the recommended dose). Protection against
pests involved a single dose of the recommended fungicide (tricyclazole).
During the same period anld using the same research sites and the station for
reference purposes, the rice programme also conducted another $!-year,
researcher-managed and farmer-implemented tria1 to study the responses of
different rice varieties to nitrogen. For this purpose, a split-plot desigb with
four levels of urea topdressed (0, 50, 100 and 150 kg/ha) in three replications
was used. Not a11 results from this comprehensive tria1 progra.mme a:re
reported in this paper, only those relating to varieties that either had dlreacly
been recommended officially or were likely to be recommended in the near
future. The rice varieties considered in different parts of the analysis were
DJ684D (aquatic), DJ125 19, IKP and TOX728 (phreatic), and IRA.Tl12
and IRAT 10 (rain-fed).
RESULTS AND DISCUSSIONS
Yield gap analysis
Data in Table 2 collected by the FSR team show that most of the v rieties
tested by ISRA since 1982 have experienced yield losses between 35 % and
70% when transferred from research station trials (i.e. research-m naged
and research-implemented:) to farmers’ management and implemer :atio:n,
when no inorganic fertilizer was used. Inorganic fertilizer is used ra ely by
farmers on the rice trop. The relative decline in yield generally appe; red 1.0
be less for the twd local varieties (i.e. Ablaye Mano and Barafita).
In the yield gap trial, the three factors (i.e. fertilizer, variety ar d pest
control; Table 3) explained most of the difference between the far n and
TABLE 2
Total Yield Gap (kg/ha) for Rice Varieties between Trials at Djibelor Station and 1 a rmers’
Tests in Casamance, 1982-86
- - -
Rice type
Variety
Station
Farmer
Dlferencr (‘W)
- - -
Improved aquatic
DJ684D
2741
1551
43
Improved phreatic
DJ12519
4039
2564
36
IKP
2454
1026
69
TOX728
3208
1232
61
Local varieties
Ablayo Mano
1443
1433
7
Barafita
2513
2036
19
Station yields were es
under research-managed and research-implemented CO dlitions,
whereas farmers’
were taken under farmer-managed and farmer-implemented CO diitions.
Adaptability of improved rice varieties in Senegal
107
TABLE 3
Yield Gap II Tria1 and Contribution of Each Factor (kg/ha)
Rice type
Variety
Yield Gap II
Variety
Fertilization Pest Control
Residual
Aquatic
DJ684D
3731
1064 (28.51)
2300 (61.64)
305 (8.17)
62 (1.68)
Phreatic
DJ12519
3905
1117 (28.60)
2700 (69.14)
71 (1.81)
17 (0.45)
IKP
3683
483 (13.11)
2867 (77.84)
202 (5.48)
129 (3.57)
TOX728
3550
833 (23.46)
1980 (55.77)
214 (6.03)
520 (14.74)
The numbers in parentheses are the contributions of each factor in percentage terms.
recommended levels, thus confirming the findings of the FSR team. The
relative contributions of each factor were fairly constant across varieties. On
average, fertilizer explained 61--78% of yield variation, variety explained 13-
30% of yield variation and pest control explained only 2-8% of the yield
variation. Implications arising out of these results are as follows:
?? The results in Table 2 implied that with no fertilizer, the yield gap in
actual and relative terms, was lower for local than for improved vari-
eties. Because fertilizer was found to be the major determinant of the
yield gap for improved varieties (Table 3), this implies that local vari-
eties are relatively better adapted to zero fertilizer levels. There are a
number of reasons why farmers are reluctant to use chemical fertilizer
on their rice fields; for example, inadequacies in the credit programme,
production risks from recurrent drought periods and major concerns
about its toxic effect on fish. The obvious issue is the relevancy of a
research programme that focuses excessively on improved varieties
using only purchased inputs. Given the practical realities in the area,
there would be merit in determining the potential substitutability of
organic matter (i.e. manure, ashes, trop residues) for the expensive
inorganic fertilizer. Another possible practical implication is to screen
improved varieties under soi1 fertility conditions more typical of the
farmers’ level, a topic we Will discuss further.
?? The second most important determinant of the yield gap, in actual and
relative terms, was the rice variety used. The results indicated that
farmers theoretically could increase rice production by an average of
500-I 100 kg/ha through using an improved variety instead of their local
variety. Whether or not the superiority of improved varieties over local
varieties would be maintained under practical farming conditions is
another question. Once again, we Will discuss this issue further.
?? Pest control appeared to contribute least to the yield gap both in actual
and relative terms. In fact, the lack of major significance of this factor
seems to be confirmed by recent studies (ISRA-DERBAC, 1993)
108
S. Sd, D. Normun. A. M. Ftwtherstonc~
indicating that level of pest infestation only amounted to between 2%
and 8% of a11 rice plants on farmers’ fields.
Adaptability analysis
Turning to the adaptability analysis designed to assess the rob.ust ms o f
varieties across different environments, the results indicated that 1 ith no
inorganic fertilizer, a11 thle improved aquatic and phreatic rice 1 u-ieti es
yielded more under better rather than poor production environme ts (i.e.
Type C stability; Table 4). A similar relationship was found to exist or the
phreatic varieties under the high fertility level (i.e. 100 units of n rogen;
Table 5), although of course the average yield was much higher tha
when
no fertilizer was applied. Also, the statistical significance of the I levant
variables when no inorganic fertilizer was applied and the lack of sign icance
at the high fertilizer level implied that the improved varieties perfor n even
less satisfactorily under very poor production environments. Althol gh the
improved varieties obviously are very responsive to inorganic fertilize , other
elements also influence the quality of the production environment. .Jnfor-
tunately, data were not available to determine exactly what these w( -e, but
they could includie not only physical factors (e.g. soi1 type and inher nt soi1
quality including organic matter, weed problems), but also factors t lat are
more socioeconomic in nature such as managerial ability and diff’erc rces in
accessibility to resources (e.g. labour available for farm operation!
land/
labour ratio, availability of cash). Given that the adaptability analy is tria1
was implemented under researcher-managed and farmer-implementc 1 con-
ditions, the production environment likely was influenced more b:y I iysical
factors than by socioeconomic factors. However, under farmer-manai
:d and
farmer-implemented conditions, the relative influence of socioeconor ic fac-
tors likely would be greater. Indirect evidence in support of this
:an ‘be
obtained from comparing the average yields of the varieties that ap lear in
both Tables 2 and 4. The average incremental yield was 86% highe under
TABLE 4
Adaptab jlity Analysis for Improved Rice Varieties at Zero Fertilizer
_--
Rice ty,ve
Intempt (hz)
Slope (bl)
Stability type
I’ield (kg/ .r,J II’
_--
Aquatic
;g$)---
-t
C
1800
68
Phreatic
-***
+ ***
c
3066
98
IKP /
-***
+ ***
C
2717
97
TOX72$
-***
+ ***
c
3016
04
Rain-fed
IRATl 2
+
A
2280
‘15
IRATl P
2428
-75
***,o,G*
+
A
**p<o.o5;
Aduptuhility qf improved rice wrieties in Senegul
109
TABLE 5
Adaptability Analysis for Improved Rice Varieties at 100 N Fertilizer
Rice type
Variet-v
Intercept (b2)
S~O~C~ (bd
Stubi1it.v type
Y i e l d lkg/ha) R2
Aq uatic
DJ684D
-
+
c
4265
77
Phreatic
DJ12519
-
i-
c
4973
90
IKP
-
+
c
4682
75
TOX728
-
+
c
5014
55
Rain-fed
IRATll2
-
+
c
3331
55
IRATIO
+
-_
B
2423
56
the researcher-managed and farmer-implemented conditions (Table 4) than
under the farmer-managed and farmer-implemented conditions (Table 2).
Turning to the improved rain-fed varieties, their performance was superior
to that of the local variety with no fertilizer across a11 production environ-
ments (Le. Type A stability; Table 4). These relationships were not main-
tained under the high level of inorganic fertilizer (Table 5). However, these
results need to be interpreted with caution. Farmers traditionally have culti-
vated rice on newly cleared land. After 3 years or SO, farmers usually had to
begin contending with fertility and weed problems and, thus, tended to move
to other plots. However, with increasing population densities, this is no
longer feasible. Thus, the favourable situation depicted for improved rain-
fed rice varieties under zero fertilizer levels likely Will now be impossible to
emulate in practice. Thus, if rain-fed rice varieties are to be grown, they are
likely to experience conditions more analogous to those shown in Table 5.
Those results are much less promising, and, in fact, the Type B stability
shown by IRATlO arises because, at high levels of nitrogen application, it
becomes more sensitive to a particular type of rice blast (pyriculariose;
Mbodj, 1991). In any case, the Senegalese government has been very reluc-
tant to recommend widespread dissemination of rain-fed rice varieties in the
Casamance. The reason for this is to encourage greater diversification of the
farming systems away from rice, in order to reduce production risk. Rice is
obviously the most desirable trop for the lowlands (i.e. aquatic and phreatic
conditions), but the upland is suitable for other crops such as maize, sor-
ghum, cowpeas and groundnuts.
Fertilizer response analysis
The results shown in Tables 6 and 7 indicate rejection of the hypothesis that
improved and local varieties respond in an analogous manner to the appli-
cation of nitrogen fertilizer; that is, the F value was 10.40 with (2, 354)
degrees of freedom. Thus, improved varieties and local varieties under
farmers’ circumstances have different response curves. With respect to the
ZI -_I---
-
----
---
110
S. Sali, D. Norman, A. M. Featherstone
TABLE 6
Estimates of Response Functions for Rice Varieties”
Variable
Aquatic
Phrentic
Rnin-fed
-_
---+----
DJ684D
IKP
D/l2519
IRAT112
IRATIO
x
32.53
32.53
35.41
18.53
1.8.52
(4.16)
(2.49)
(9.11)
(1.47)
(1.61)
Xl.5
-- 1.46
-1.47
-1.67
-0.96
-10.96
w:;;’
(-1.40)
(-5.39)
(-0.95)
(-1.04)
Dl
- 187.03
1092.00
--157.56
-121.41
(0.69)
(-0.53)
(10.40)
(-0.46)
(-0.39)
DIX
9.58
--14.28
2.66
25.18
23.92
(0.86)
(-0.77)
(0.40)
(1.41)
(1.46)
D$f’ .5
-0.70
1.16
-0.09
-1.51
-1.07
(-0.79)
(0.78)
(-0.21)
(-- 1.06)
(~0.82)
Intercept
2417.50
2417.00
2535.60
1397.00
1397.00
(16.18)
(9.67)
(34.14)
(5.80)
66.35)
R2
92
84
92
30
46
“The t-statistics are in parentheses.
TABLE 7
Results of Hypothesis Testing for Fertilizer Response Curves
Nul1 hypothesis
Parumeter restriction
Test srrrtis!tic
Equal response to N
bq=b5=0
F= 10.40**
Equal response at N = 0
Aquatic rice
DJ684D
h3=0
t=z0.69
Phreatic rice
DJ12519
b,=O
t = 10.40**
IKP
h.3 = 0
t == -0.53
TOX728
h3=0
t=7.83’k*
Rain-fed rice
IRATl12
b3=0
t = -0.46
TRATIO
b3=0
t = -0.39
** Indicates rejection of hypothesis at p < 0.05.
a1 respon.ses of improved and local varieties at zero ikvel of
rejected in the case of two phreatic varieties, DJ12. 19 and
for the remainder. These results are somewhat arn. biguous
ary to what we showed earlier, that the yields of th :!Se t\\VO
are hig:her than those of the local varieties at zerc fertili-
zer level. Howe
for the remaining improved varieties, a cross-ov( r effect
improved varieties performing better than local 7 ariettes
as fertilizer leve
reased, thus confirming a Type C stability.
ate that the ordering of a set of varieties accor ling to
y identical at high and low levels of fertilizer app ication
Adaptability of improved rice varieties in Senegal
111
and may be affected by other physical and socioeconomic factors that
determine the quality of the production environment. Certainly, under
poorer production environments including a zero inorganic fertilizer level, the
two local rice varieties, Ablaye Mano and Barafita, are competitive in terms
of their yields and are especially vigorous at emergence (Posner et al., 1991).
The rice programme has used a high level of inorganic fertilizer application
in its screening process. As a result, lodging problems have been minimized,
although efforts to eliminate the problem of pyriculariose do not always
appear to have been successful. Another very important implication borne
out by the results of this study is that the variety screening process also
appears to be inadequate in providing improved rice varieties well adapted to
less favourable production environments, including the farmer’s common
strategy of not using any inorganic fertilizer. On balance, most of the existing
improved rice varieties appear to be most appropriate for those farmers
working in favourable production environments, including the extensive use
of inorganic fertilizer. We conclude that, because less than 5% of the farmers
applied inorganic fertilizer (ISRA-DERBAC, 1993), the production envi-
ronments under which much of the rice is produced are less than ideal.
Therefore, widespread adoption of improved rice varieties Will require vari-
eties well adapted to such suboptimal growing conditions. Thus, some
adjustments are needed in the approach of the rice breeding programme,
particularly with respect to the screening process and greater collaboration
with agronomists, soi1 scientists and the FSR team in order to identify
acceptable ways of improving the production environments of farmers.
CONCLUSION
In this paper, three different methods have been used to assess the potential
suitability of rice varieties under varying production environments in the
Casamance region of Senegal. Al1 the approaches bave their place. The yield
gap approach, although intuitively appealing to station-based scient,ists in
helping to highlight the major factors (especially physical/technical factors)
contributing to the difference between experiment station and farm level
yields, does not by itself indicate differences that arise as a result of variation
in actuai production environments. It in essence represents a single point on
a production function. The production function (i.e. fertilizer response)
approach cari help assess the major determinants of yield (e.g. fertilizer and
variety) but the values of the coefficients on the variables in the function Will
reflect the levels and qualities of the non-experimental variables. However, it
is particularly sui ted to assessing the economic optima for different inputs for
specific production environments. In terms of ability to assess robustness of
112
S. Sall, B. Norman, ,4. M. Featlzerstone
the varieties across different practical production environments and; when
necessary, develop multiple recommendations, adaptability analysis is t’he
most suitable. However, to fully exploit the potential of adaptability, analy-
sis, it would be desirable to go one step further than was possible in this
paper, namely to identify Imore specifically the determinants of the different
production environments ;as represented by different values of the environ-
mental index.
The results of this study demonstrate the irrelevancy of a single approach
to recommending the same rice varieties for a11 production environments
found in the region. The analysis shows the need for multiple recommenda-
tions appropriate for fanmers operating in different production environ-
ments. In the long run, major breakthroughs in rice production Will be
realized only through substantial improvements in the physical and socio-
economic production environments of the majority of small farmers. How-
ever, this is likely to require a sustained incremental approach. Therefore, we
believe the rice programme needs to focus its activities on two com-
plementary strategies:
?? The rice programme should produce varieties adapted to thel varied
production environments under which farmers operate. In FSR par-
lance, this requires the recognition of more than one research or
recommendation domain for the improved rice varieties and a. scieening
process that takes the different production environments into account.
Because, for reasons discussed subsequently, breeders likely will have
difficulty developing improved rice varieties with Type A stability across
the range of productilon environments found in the region, a strategy of
breeding some with a Type B stability and some with a Type C smbility
might be more appropriate.
?? The rice programme should develop close collaborative relationships
with agronomists, soils scientists and the FSR team to identify the
determinants of the different production environments under pi-actical
farming con’ itions and, in cooperation with farmers, design an evalu-
P
ate relevant: strategies for their improvement. Also, improve ent in
organic mat’er and nlot relying too heavily on purchased inputs ‘s likely
to be more elevant for farmers operating in poor production t nviron-
ments. In developing high yielding varieties very responsive to ir 3rganic
fertilizer, brteders have tended to emphasize grain yield at the :xpense
of stover (i.4. biomass). Varieties with Type A stability are likely to have
such charac eristics. Under poor production environments, rice 1arieties
1
that have a higher stover/grain ratio are likely to be more effc rtive in
contributin
organic matter to the soil. This is more likely tc be the
”
case with v, rieties of Type B stability. Also under poor prc luction
Adaptability qf’improved rire varieties in Senegal
113
environments, seeking practical ways of incorporating legumes into the
cropping system would have merit.
In essence, what we are advocating is that ISRA should move away from a
single blanket recommendation for rice varieties to one that emphasizes a
smorgasbord of different rice varieties, accompanied by information as to
when they work best (i.e. conditional and targeting information; Norman et
al., 1995), from which farmers cari Select. Such an approach recognizes that
limited-resource farmers live and work on farms characterized by a high
degree of both biophysical and socioeconomic diversity. It also recognizes
that farmers are rational and have the best knowledge about their own pro-
duction environments. Consequently, we believe the close collaboration that
we are advocating between station- and farm-based researchers (i.e. com-
modity-based programmes and FSR teams) and farmers is critically impor-
tant, not only in developing relevant improved rice varieties but also in
identifying practical strategies for improving the farmers’ production envi-
ronments. Also, such collaboration is needed because our research (Sall,
1997) has shown that farmers’ decisions as to whether or not they Will adopt
improved varieties also involve criteria other than yield.
REFERENCES
De natta, S.K., Gomez, K.A., Herdt, R.W. and Barker, R. (1978) A Hundbook on
the MethodologJl
for an Integrated Experiment-Survey on Rice Yield Constraints.
The International Rice Research Institute, Los Banos, Laguna, Philippines.
Hildebrand, P.E. (1983) Modified stability analysis of farmer managed, on-farm
trials. Agronomy Journal 76, 27 l-274.
Hildebrand, P.E. and Russell, J.T. (1996) .4daptability
Analysis: A Methodfor the
Design and Interpretation of On-Farm Research E.xtension. Iowa State University
Press, Ames, IA.
ISRA-DERBAC (1993) Rapport d’Activite,
Campagne Agricole 1991-1992, Conven-
tion Recherche d’Accompagnement.
Institut Senegalais Recherche Agronomique
and Projet de Development Integre de la Basse Casamance SRA, Ministere du
Development Rural et de 1’Hydraulique.
Linares, O.F. (1981) From tidal swamp to inland valley: on the social organization
of wet rice cultivation among the Diola of Senegal. Ajiica 51, 21-32.
Mbodj, Y. (1991) Les Recherches Rizicoles en Cusamance: Situation en 1991. Centre
Recherche Agronomique, Djibelor.
Norman, D.W., Worman, F.D., Siebert, J.D. and Modiakgotla, E. (1995) The
farming systems approach to development and appropriate technology genera-
tion. FAO Farm Systems Management Series No. 10. FAO, Rome. Italy.
Posner, J., Kamuanga, M. and Mamadou L. (1991) Lowland cropping systems in the
Lower Casamance of Senegal: results of four years of agronomie research.
M.S.U. International Development Paper No. 30. Department of Agricultural
Economies, Michigan State University, East Lansing, MI.
114
S. Sar’l, D. Norman, A. M. Featherstone
Sall, S. (1997) A technology-characteristic approach to adoption: the ase of
improved rice varieties in Southern Senegal. Ph.D. Dissertation. Depart kent of
Agricultural Eqonomics, Kansas State University, Manhattan, KS.
I
,,
Traxler, G. and Byerlee, D. (1993) A joint-product analysis of the adoptron of
modern cereal varieties in developing countries. Americun Journwl of’ Ag.C-
cultural Economies 75, 98 l-989.