Developing rice cultivars for high-fertility upland systems in the Asian tropics

www.elsevier.com/locate/fcr

Field Crops Research 97 (2006) 43–52

Developing rice cultivars for high-fertility upland

systems in the Asian tropics

G.N. Atlin a,*, H.R. Lafitte a, D. Tao b, M. Laza a, M. Amante a, B. Courtois c

a International Rice Research Institute, DAPO Box 7777, Metro Manila, Philippinesb Yunnan Academy of Agricultural Sciences, Kunming 650205, People’s Republic of China

c CIRAD-Biotrop TA40/03, Avenue Agropolis, 34398 Montpellier Cedex 5, France

Abstract

Traditionally, upland rice is grown in Asia in low-input, subsistence systems. More productive upland systems, using more fertilizer and

improved varieties, are emerging in China and Philippines, and could contribute to productivity increases in rainfed environments in other

countries. Here, we evaluate, on-station and on-farm, the yield under upland management of improved indica upland cultivars selected for

yield under high-fertility conditions. These cultivars are compared with traditional and improved tropical japonica upland varieties, and with

elite indica high-yielding varieties (HYV) developed for irrigated lowland production, to characterize the features of varieties that produce

high yields in favorable upland environments. Forty-four improved and traditional varieties and experimental lines were evaluated in irrigated

lowland, non-stressed upland, moderately stressed upland, severely water-stressed upland, and low-fertility upland environments in southern

Luzon, Philippines. Correlations between yields in non-stress and mild-stress environments were low but positive. Some cultivars, like

IR55423-01, were among the highest yielding under both conditions, indicating that high yield and moderate water-stress tolerance can be

combined. Upland-selected indica varieties yielded 3.56 t ha�1 in favorable upland environments on-station in southern Luzon, out-

performing improved tropical japonica and irrigated varieties by 23 and 69%, respectively. They were also the highest-yielding class in

infertile, acid soils. The improved upland indica cultivars are about 110 cm tall under favorable upland conditions and maintain a harvest

index of nearly 0.4, or about one-third higher than other cultivar types. The best upland-adapted rice varieties produced average yields on-farm

of 3.3 and 4.1 t ha�1 in southern Luzon and Yunnan, respectively, outyielding traditional checks by 30–50% with moderate N application.

Screening under both high-fertility, non-stress conditions and moderate reproductive-stage stress appears to be needed to develop cultivars

combining high-yield potential with drought tolerance. Upland-adapted indica cultivars offer a new approach to increasing productivity and

reducing risk in Asian rainfed rice systems.

# 2005 Elsevier B.V. All rights reserved.

Keywords: Rice cultivars; Upland environments; Harvest index

1. Introduction

Traditional upland rice crops are grown in unbunded,

unflooded fields, where soil conditions in the root zone

remain aerobic through most of the growing season. In most

traditional Asian upland rice-growing areas, soils are acidic

and infertile. Farmers usually treat upland rice as a

subsistence crop, investing little in inputs beyond family

labor. Because upland rice varieties are grown without

irrigation in unsaturated soils, they are considered to be

* Corresponding author. Tel.: +63 2 845 0563x2586; fax: +63 2 845 0606.

E-mail address: [email protected] (G.N. Atlin).

0378-4290/$ – see front matter # 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.fcr.2005.08.014

drought tolerant. However, upland rice yields in traditional

systems are low, averaging 1–2 t ha�1 in most growing

regions. Intensification of management of these systems

with currently available germplasm is difficult because most

traditional upland rice varieties are tall, low-tillering, and

prone to lodging when grown under conditions of favorable

moisture and high-soil fertility.

Recently, a new class of upland-adapted cultivars with

improved lodging resistance, harvest index, and input

responsiveness has been developed by breeding programs

in China, Brazil, and the Philippines. These varieties

combine some of the yield potential-enhancing traits of

lowland high-yield varieties with adaptation to aerobic soils.

G.N. Atlin et al. / Field Crops Research 97 (2006) 43–5244

They are often referred to as aerobic rice varieties to

differentiate them from traditional upland rice varieties

adapted to low-input systems. The distinguishing feature of

aerobic production systems is that crops are direct-seeded in

free-draining, non-puddled soils where no standing water

layer is maintained in the field, and roots grow in a mainly

aerobic environment. A commercial cropping system based

on aerobic rice has spread widely in Brazil as a result of the

development of improved varieties (Pinhiero et al., 2005).

Aerobic rice is also of interest in water-short lowland rice

production systems. Shifting from conventional flooded

systems to direct seeding in non-puddled, non-flooded fields

can reduce water requirements for rice production by over

50% by reducing percolation, seepage, and evaporation

losses (Castaneda et al., 2002). This gain is at the expense of

some reduction in yield relative to that obtainable in fully

irrigated systems, but may not result in significant yield loss

relative to conventional rainfed lowland management in

drought-prone areas. Both water-short irrigated and drought-

prone rainfed lowland systems may, therefore, benefit from

aerobic rice systems. Aerobic production systems based on

high-yield, input-responsive upland cultivars have already

been developed to replace conventional irrigated lowland

rice production on the water-short plains of northeastern

China, producing yields of 4.5–6.5 t ha�1, with substantial

water savings relative to conventional irrigated production

(Wang et al., 2002; Bouman et al., 2005).

Aerobic conditions can comprise a wide range of soil

hydrological environments, from near saturated to well

below field capacity. It is not known to what extent these

conditions should be considered separate breeding targets

(Atlin and Lafitte, 2002). There is also little information in

the literature regarding optimal characteristics of high-yield

upland rice varieties for the Asian tropics, and the relative

yield potential of the major germplasm groups under high-

input upland management. This information is needed to

identify promising parents and develop effective screens for

aerobic rice systems. There is also little information

available on the on-farm performance of improved upland

rice cultivars across locations and years under rainfed

conditions. This information is needed to assess the

productivity and stability of intensified upland rice-based

cropping systems. The objectives of this report, are

therefore, to:

� i
dentify cultivars and germplasm types with high-yield
potential under a range of upland hydrological conditions;

� d
ocument the yield of currently available upland rice
cultivars and breeding lines under purely rainfed

conditions and farmer management in emerging intensi-

fied upland rice-based systems in Philippines and Yunnan,

China;

� c
haracterize some of the common attributes of cultivars
with superior performance in aerobic rice systems;

� p
ropose screening methods that can be used to develop
superior cultivars for aerobic rice systems.

2. Materials and methods

2.1. Performance of upland, lowland, and aerobic

varieties in five management regimes in southern Luzon,

Philippines

To clarify the degree of genotype-specific adaptation to

hydrologic and soil fertility conditions, and to characterize

differences between germplasm groups, 44 upland and

irrigated lowland varieties were screened in 14 three- or

four-replicate randomized complete-block trials classified

into five environment types differing in soil fertility and

hydrology:

(a) i
rrigated lowland fields established via transplanting at
IRRI in the dry season, in which the trials experienced

no water stress (NSL);

(b) u
pland, high-fertility fields established via direct
seeding at IRRI during the wet season, with soil

maintained at or near field capacity via occasional

irrigation, and in which trials experienced little stress

(NSU);

(c) u
pland, low-fertility fields established via direct seeding
at Siniloan, Philippines, a P-deficient acid upland site,

during the wet season (LFU);

(d) u
pland fields at IRRI established via direct seeding in
the dry season, with frequent irrigation resulting in mild

to moderate stress (MSU);

(e) u
pland fields at IRRI established via direct seeding in the
dry season, with infrequent irrigation after the maximum

tillering stage resulting in severe stress (SSU).

The 44 varieties are listed in Table 1. Thirty-six are

improved upland types, three are traditional Philippine

tropical japonica upland varieties, one is a traditional aus

variety and four are varieties from the IRRI irrigated rice-

breeding program. The improved upland varieties were of

diverse origin, representing both the indica and tropical

japonica groups. They are the output of Philippine, Indian,

Brazilian, Colombian, Indonesian, and West African

breeding programs targeted at a range of management

conditions. All varieties are photoperiod-insensitive or only

weakly sensitive, ranging in duration from approximately

90–130 days in duration. All varieties set seed well in both

the wet and dry seasons in Luzon.

Environments are described in Table 2. In all cases, these

environment types were represented by at least two trials in

different field locations. In all but the SSU environment type,

trials were conducted in at least two seasons. All upland

varieties were included in all trials, but trials conducted

before the dry season of 2002 did not include the four

irrigated varieties. In the dry season, all trials were planted

between January 3 and January 20. In the wet season, all

trials were planted between June 10 and July 10.

Trials were conducted at two sites in southern Luzon: Los

Banos and Siniloan. The Los Banos trials were conducted at

https://www.researchgate.net/publication/223360425_Sustainability_and_profitability_of_aerobic_rice_production_in_Brazil?el=1_x_8&enrichId=rgreq-2f34609f-afa5-4c94-b438-cae7257d6e49&enrichSource=Y292ZXJQYWdlOzIyMjk3NzM3MjtBUzoyMTg4NzQzNjk4NDMyMDBAMTQyOTE5NTEyNDU1MQ==

https://www.researchgate.net/publication/222138099_Performance_of_aerobic_rice_varieties_under_irrigated_conditions_in_North_China?el=1_x_8&enrichId=rgreq-2f34609f-afa5-4c94-b438-cae7257d6e49&enrichSource=Y292ZXJQYWdlOzIyMjk3NzM3MjtBUzoyMTg4NzQzNjk4NDMyMDBAMTQyOTE5NTEyNDU1MQ==

G.N. Atlin et al. / Field Crops Research 97 (2006) 43–52 45

Table 1

Description of 44 varieties evaluated in southern Luzon in five environments differing in soil fertility and hydrology

Cultivar Germplasm group Variety type Adaptation Origin

AUS 196 Aus Traditional Upland India

AZUCENA Tropical japonica Traditional Upland Philippines

B6144F-MR-6-0-0 Indica Improved Upland Indonesia

C 22 Tropical japonica Improved Lowland Philippines

CT13370-12-2-M Indica/tropical japonica Improved Upland Colombia

CT13377-4-2-M Tropical japonica Improved Upland Colombia

CT13382-8-3-M Indica/tropical japonica Improved Upland Colombia

CT6510-24-1-2 Indica Improved Upland Colombia

CT6516-24-3-2 Indica/tropical japonica Improved Upland Colombia

DINORADO Tropical japonica Traditional Upland Philippines

IR64 Indica Improved Lowland Philippines

IR72 Indica Improved Lowland Philippines

IR47686-30-3-2 Tropical japonica Improved Upland Philippines

IR55419-04 Indica Improved Upland Philippines

IR55423-01 Indica Improved Upland Philippines

IR60080-46A Tropical japonica Improved Upland Philippines

IR65261-09-1-B Tropical japonica Improved Upland Philippines

IR65907-116-1-B Indica/tropical japonica Improved Upland Philippines

IR66417-18-1-1-1 Tropical japonica Improved Upland Philippines

IR66421-062-1-1-2 Indica/tropical japonica Improved Upland Philippines

IR66424-1-2-1-5 Indica/tropical japonica Improved Upland Philippines

IR68702-072-1-4-B Tropical japonica Improved Upland Philippines

IR70358-84-1-1 Indica/tropical japonica/aus Improved Upland Philippines

IR70360-38-1-B-1 Tropical japonica Improved Upland Philippines


IR71525-19-1-1 Indica/tropical japonica Improved Upland Philippines


IRAT 170 Tropical japonica Improved Upland Ivory Coast




MARAVILHA Tropical japonica Improved Upland Brazil

PALAWAN Tropical japonica Traditional Upland Philippines

PRIMAVERA Unknown Improved Upland Brazil

PSBRC 80 Indica Improved Lowland Philippines

PSBRC 82 Indica Improved Lowland Philippines

UPL RI-5 Indica Improved Upland Philippines

UPL RI-7 Indica Improved Upland Philippines

VANDANA Aus/tropical japonica Improved Upland India

WAB 181-18 Tropical japonica Improved Upland Ivory Coast



WAB 96-1-1 Tropical japonica Improved Upland Ivory Coast

WAY RAREM Indica Improved Upland Indonesia

the IRRI Experiment Station (148110N, 1218150E, 21 masl),

where annual average rainfall was 2100 mm in 1979–2002.

The soil is an Andaqueptic Haplaquol with pH of 6.5. The

Siniloan site is located approximately 60 km east of Los

Banos (148280N, 1218290E, 310 masl), on a light-textured,

acid (pH 3.7), P-deficient Typic Palehumult soil. Long-term

average annual rainfall data are not available for Siniloan,

but the annual mean for 1997–2000 was 4500 mm. Trials

were conducted only in the wet season at Siniloan, and in

both the wet and dry seasons at Los Banos. From 1 January

to 30 April, total rainfall was 257.2, 56.1, and 34.1 mm in the

dry seasons of 2001, 2002, and 2003, respectively; pan

evaporation was 699, 590, and 734 mm for the same three

periods. The vapor pressure deficit in the period from 60

days after sowing to harvest was 0.75 kPa in the dry season

experiments, compared to 0.50 kPa in the wet season. Free

water was not present within 1 m of the soil surface during

the dry season. In the wet season, free water was present

within 0.3 m of the surface during periods of high rainfall,

but usually fell below 0.5 m for several extended periods.

Wet season upland trials experienced little water stress at

either site, and were completely rainfed. Irrigated lowland

trials were maintained with constant standing water between

5 and 15 cm deep from transplanting to pre-harvest

drainage. Dry season upland trials were either sprinkler-

or basin-irrigated, with approximately 30 mm of water

applied at each sprinkler irrigation, and approximately

50 mm applied at each basin irrigation. Dry season irrigation


Table 2

Description of trials conducted in southern Luzon in which 44 varieties were evaluated for adaptation to five environments differing in soil fertility and

hydrology

Environment

typeaLocation Season Year N–P2O5–K2O Hydrology Establishment method Irrigation

method

Trial mean

yield (t ha�1)

NSL IRRI Dry 2001 60–60–60 Lowland Transplanting Continuous flood 2.70

NSL IRRI Dry 2002 120–30–30 Lowland Transplanting Continuous flood 2.82

NSL mean 2.76

NSU IRRI Wet 2001 130–100–100 Upland Direct seeding Rainfed 2.33

NSU IRRI Wet 2002 60–60–60 Upland Direct seeding Rainfed 2.15

NSU mean 2.23

MSU IRRI Dry 2002 110–60–60 Upland Direct seeding Sprinkler twice weekly 0.91



MSU IRRI Dry 2003 90–60–60 Upland Direct seeding Basin once every 10 days 1.07

MSU mean 1.05

LFU Siniloan Wet 2000 60–90–90 Upland Direct seeding Rainfed 1.42



LFU mean 1.00

SSU IRRI Dry 2002 90–60–60 Upland Direct seeding Sprinkler once weekly 0.55



SSU mean 0.46

a NSL, non-stressed lowlands; NSU, non-stressed uplands; MSU, moderately water-stressed uplands; LFU, low-fertility uplands; SSU, severely water-

stressed uplands.

regimes were constant from sowing to harvest. Because

irrigation frequencies and amounts were maintained con-

stant through the growing season, intermittent, repeated

stress between irrigations was experienced by all varieties,

regardless of flowering date and duration.

Weeds were controlled though a combination of early

season herbicide applications and manual weeding. Che-

mical insect control was used as required. A severe

infestation of mole crickets damaged seedlings in the DS

2002 trials at Los Banos. Heavy rainfall events shortly after

sowing resulted in severe localized erosion within the

Siniloan site in 2001 and 2002.

Plot size ranged from 4.5 to 7.5 m2. The entire plot was

harvested for grain yield. Yields given are for air-dried paddy,

at approximately 12% moisture content. Harvest index and

panicle number were determined from a 0.25 m2 sample.

To estimate cultivar correlations across environment

types for mean yield, least-squares means across trials

Table 3

Correlations among mean yields of 44 aerobic, upland, and irrigated varieties ac

Non-stressed uplands Moderately water-s

Non-stressed lowlands 0.72** 0.31*

Non-stressed uplands 0.44**

Moderately water-stressed uplands

Severely water-stressed uplands

* Significant at 0.05.** significant at 0.01.

within environment types were first estimated using a mixed

model in which trials and replicates were considered random

factors, and cultivars were considered fixed. The REML

algorithm of the SAS MIXED procedure (SAS Institute,

1992) was used for the analysis.

A subset of 15 cultivars (Table 3) could be unambigu-

ously classified into five types, based on their Glazsmann

(1987) isozyme genotype and selection history: irrigated

lowland high-yield varieties (HYV), improved upland-

adapted indica varieties selected under high-fertility

management (IHF), traditional Philippine tropical japonica

upland varieties (JTV), improved tropical japonica-derived

upland varieties targeted at low-fertility environments (JLF),

and early maturing, drought-tolerant upland varieties

targeted at drought-prone environments (EDT), with three

cultivars within each type group. Means for each cultivar

type were estimated within environment types, using a

model similar to the one described above, but subdividing

ross five environment types in southern Luzon, 2000–2003

tressed uplands Severely water-stressed uplands Low-fertility uplands

�0.10 0.41**

�0.04 0.57**

0.53** 0.50**

0.27


the cultivar effect into two components: cultivar types and

cultivars within types.

2.2. Performance of improved upland rice varieties

under aerobic management in farmers’ fields in southern

Yunnan, China

Variety trials were conducted on-farm by Yunnan

Academy of Agricultural Sciences research staff in southern

Yunnan, China, to compare improved upland tropical

japonica cultivars (IRAT 104, Yunlu 29, and Yunlu 52)

and the improved indica B6144-MR-6-0-0 with a traditional

upland tropical japonica landrace (Mengwanggu) under

high-input management. Trials were conducted under

upland conditions on rainfed terraces, in communities that

formerly relied upon upland rice production for food

security, but which now grow intensively managed aerobic

rice in rotation with other upland crops. Experiments were

seeded at a rate of 150 kg ha�1. Fertility management

consisted of a basal application of 15 kg ha�1 urea,

15 kg ha�1 zinc sulfate, and 300 kg ha�1 calcium magne-

sium phosphate, plus a top-dressing of 135 kg ha�1 urea at

the panicle initiation stage, for a total N application of

approximately 70 kg ha�1. Three-replicate trials with

harvested plot areas of 13.3 m2 were conducted annually

at 11 sites between 1993 and 2002. The 11 sites are at

elevations ranging from 830 to 1680 masl. They include four

low-elevation (�1250 masl) sites (Jinghong, Menglian,

Ximeng, and Mengla) at which both indica and japonica

cultivars can perform well, and seven short-season, higher-

elevation (>1250 masl) sites (Wengshan, Pingbian, Simao,

Menghai, Langcan, Genma, and Cangyuan) at which

japonica cultivars are usually favored. Means were

estimated from a combined residual maximum likelihood

(REML) analysis over years, using the REML algorithm of

the SAS MIXED procedure (SAS Institute, 1992), with

cultivars and elevation (low versus high) classes considered

fixed, and trials within elevation classes considered random.

Within the 1993–2002 period, each new line was subjected

to a testing cycle of at least 2 years. Mengwanggu was

common to all sites in all years. For the analysis, cultivars

were considered fixed.

2.3. Performance of improved upland rice varieties

under aerobic management in farmers’ fields in

Batangas, Philippines

Experiments were conducted in the province of Batangas,

southern Luzon, Philippines to compare the productivity of

improved indica and traditional tropical japonica varieties

under high-input upland management in farmers’ fields.

This region is characterized by light-textured upland soils of

volcanic origin. Farmers in the region grow a mix of

traditional and improved upland varieties in a double-

cropping system, in which rice is planted at the onset of the

rainy season (usually in May) followed in October by a

maize, sweet potato, or vegetable crop that exploits the late

rains. Varieties were evaluated using both researcher-

managed, replicated on-farm ‘‘Mother’’ trials and farmer-

managed ‘‘Baby’’ trials. In the wet season of 2002, two

traditional Philippine upland varieties (Dinorado and

Azucena) and two high-yielding improved upland rice

cultivars (IR55423-01 and IR74371-46-1-1) were grown in

farmers’ fields in two-replicate researcher-managed trials at

three sites near Lemery, Batangas, as part of a larger

experiment evaluating 15 varieties. A seeding rate of

60 kg ha�1 was used, with 90 kg N ha�1 applied in three

splits in the form of urea. The same varieties were also

screened in farmer-managed trials, in which farms were

treated as incomplete blocks. Farmers received 2 kg of seed

of each of three varieties, and were asked to plant and

manage them as they did their own crop. Farmer fertilizer

application rates are not precisely known, but ranged from

60 to 120 kg N ha�1. In those experiments, seed was

broadcast in May, shortly after the first monsoon rains, at a

rate of about 150 kg ha�1 on furrowed fields, then covered

by harrowing. In both researcher-managed and farmer-

managed trials, rows were plowed into the stand after

emergence. Weeds were managed via inter-row cultivation

using animal traction. At harvest, 5 m2 crop cuts were

harvested and yield determined for each of the introduced

varieties and for the farmer’s field adjacent to the Baby trial.

Cultivar means from both researcher- and farmer-managed

trials were estimated from a combined residual maximum

likelihood (REML) analysis over years, using the REML

algorithm of the SAS MIXED procedure (SAS Institute,

1992), with cultivars considered fixed and farms random.

3. Results

3.1. Performance of upland, lowland, and aerobic

varieties in five management regimes in southern Luzon,

Philippines

3.1.1. Correlations across environment types

Yields were determined for 44 diverse cultivars (Table 1)

in five environment types (Table 2) for at least two seasons.

Mean yields across cultivars for environments NSL, NSU,

MSU, LFU, and SSU were 2.76, 2.23, 1.05, 1.00, and

0.46 t ha�1, respectively. Average yields in all environments

were relatively low due to the large number of upland

varieties with poor yield potential under favorable condi-

tions. Correlations of variety mean yields across these

hydrological regimes are presented in Table 3. Cultivar

means in the NSL environment were highly correlated with

those from the NSU environment, indicating that similar

cultivars with high-yield potential were favored in both

lowland and near-saturated upland environments. Correla-

tions between yields in the well-watered NSL and NSU

environments and the MSU environment were positive but

considerably lower, indicating that selection in well-watered


Table 4

Subset of cultivars used in analysis of agronomic performance of cultivar

types in southern Luzon 2000–2003

Cultivar type Cultivar

Irrigated lowland high-yield

varieties (HYV)

IR64

IR72

PSBRC 82

Improved upland-adapted IR55423-01

environments is not adequate to identify genotypes that

perform well under moderate stress. Genotypic effects from

the severely stressed SSU environment were correlated with

those from the moderately stressed dry-season environment,

but were not positively associated with those estimated in

the irrigated lowland and aerobic wet season environments.

These results indicate that, in hydrological terms, there are

three broad target environments for upland rice breeding:
indica varieties selected for
high-fertility environments (IHF)

IR55419-04

UPL RI-7
(1) e
Tabl

Mea

japo

in fiv

Varie

HYV

IHF

JLF

EDT

JTV

L.S.D

Improved upland tropical

japonica varieties selected for

low-fertility environments (JLF)

IR60080-46A

UPL RI-5

IR47686-30-3-2

nvironments in which there is little water stress, with

soil remaining near saturation throughout the growing

season, with soil moisture tension in the upper 15 cm of

soil never falling below �0.03 MPa;
Improved early maturing
varieties selected for

Vandana

IR70358-84-1-1
(2) m
drought tolerance (EDT) WAB 56-125

Traditional Philippine

upland tropical japonica

varieties (JTV)

Azucena

Palawan

Dinorado

oderate-stress environments in which soil moisture

falls below field capacity, either regularly or during the

critical reproductive period, with soil moisture tension

in the upper 15 cm layer often falling below�0.03 MPa,

reducing yields by 30–50% relative to saturated soil

environments;

(3) s
evere-stress environment where soil is rarely saturated
between irrigations or rainfall events, and in which soil

moisture tension frequently falls to�0.03 MPa at 30 cm

depths, resulting in yield reductions of 70% or more

relative to saturated soil environments.

The positive correlation between yields in non-stressed

and moderately stressed environments indicates that it is

possible to combine high-yield potential with tolerance to

moderate levels of water stress, and that the most effective

breeding approach for aerobic rice production systems that

are likely to be subject to intermittent stress is to screen in

both non-stressed and moderately stressed environments,

selecting on the basis of mean yield across these regimes.

Correlations between line means at the low-pH LFU site

at Siniloan and the more fertile NSU site at Los Banos were

positive and significant. These results indicate that

responsive genotypes performing well in favorable environ-

ments can also produce relatively high yields in unfavorable

environments. Selection for yield potential under favorable

conditions, required for gains in intensively managed

aerobic rice systems, is therefore, also likely to produce

yield gains in less fertile environments.

e 5

n yield (t ha�1) of irrigated high-yield varieties (HYV), improved indica upl

nica varieties (JTV), and improved tropical japonica types selected for low-fert

e environment types in southern Luzon, 2000–2003

ty type Environment type

Non-stressed

lowland

Non-stressed

upland

Modera

upland

4.04 2.12 0.84

3.62 3.56 1.47

3.31 2.89 1.10

1.71 1.44 1.11

2.29 1.63 0.81

.0.05 0.82 0.47 0.30

3.1.2. Performance of cultivar groups

Among the varieties listed in Table 1 were representatives

of distinct cultivar groups. Table 4 lists the subset of three

cultivars comprising the sample for each group (HYV, IHF,

JTV, JLF, and EDT) Mean yields of variety groups within

environments are presented in Table 5. The HYV types

outyielded all others under lowland conditions, although the

difference was not significant relative to the IHF and JLF

groups. In the high-fertility, well-watered IRRI wet season

environment, IHF varieties significantly outyielded all other

types, surpassing the HYV types by nearly 70%. The IHF

types also outyielded all others under conditions of moderate

stress in the dry season at IRRI and under low-fertility, acid-

soil conditions at Siniloan. The responsiveness of IHF

varieties to improved fertility conditions is illustrated by the

contrast in their performance in the NSU and LFU

environments; when grown in high-fertility (NSU) environ-

ments at IRRI, the improved indica upland types yielded

2.30 t ha�1 more than when grown in the low-fertility (LFU)

Siniloan environment. The equivalent responses were only

1.75, 0.60, and 0.87 t ha�1 for the JLF, EDT, and JTV

cultivar groups, respectively. The advantage of the EDT

and types selected in high-fertility environments (IHF), traditional tropical

ility and drought-prone environments (JLF and EDT, respectively) evaluated

tely water-stressed Severely water-stressed

upland

Infertile

upland

0.47 0.91

0.60 1.26

0.31 1.14

1.00 0.84

0.26 0.76

0.30 0.38


types was evident only under conditions of extreme stress,

when they significantly outyielded all other cultivar groups.

They were the lowest-yielding cultivar group in well-

watered environments. The JTV landraces were the lowest-

yielding under conditions of water stress and low fertility. In

general, the aerobic-adapted IHF group combined respon-

siveness to favorable conditions with a high degree of

tolerance to both water stress and low fertility.

The superiority of aerobic-adapted improved indica

cultivars under favorable and moderately stressful upland

conditions appears to be a function of their higher harvest

index (HI) in these environments relative to other cultivar

types (Table 6). Both IHF andHYVvariety types achievedHI

of nearly 50% in irrigated lowland environments, but theHI of

HYV declined precipitously in all upland environments,

whereas IHF varieties retained a relatively high HI. JLF

varieties targeted at low-input systems usually producedmore

biomass than aerobic rice varieties (Table 7), but partitioned

much less drymatter to grain.Only in themost severelywater-

stressed environment was the HI of IHF varieties exceeded by

that of the drought-tolerant EDT group. Drought-tolerant

tropical japonica varieties retained much higher levels of

grain-set under extreme stress than other types, achieving HI

of over 30%. However, their low biomass production in all

environments resulted in much reduced yields relative to the

IHF group in all other environments.

Although the drought-tolerant japonica varieties are

much shorter in duration than other cultivar classes,

Table 7

Mean biomass production (t ha�1) of irrigated high-yield varieties (HYV), imp

traditional tropical japonica varieties (JTV), and improved tropical japonica type

respectively) evaluated in five environment types in southern Luzon, 2000–2003

Variety type Environment type

Irrigated

lowland

Non-stressed

upland

Moderate

upland

HYV 8.59 7.96 7.15

IHF 7.57 9.59 5.92

JLF 8.48 9.28 5.93

EDT 4.41 6.37 4.25

JTV 6.89 7.44 6.46

L.S.D.0.05 1.63 1.37 2.00

Table 6

Mean harvest index of irrigated high-yield varieties (HYV), improved indica upl

japonica varieties (JTV), and improved tropical japonica types selected for low-fert

in five environment types in southern Luzon, 2000–2003


Irrigated

lowland

Non-stressed

upland

Moderate

upland

HYV 0.47 0.27 0.21

IHF 0.48 0.37 0.28

JLF 0.39 0.31 0.21

EDT 0.39 0.25 0.29

JTV 0.34 0.22 0.16

L.S.D.0.05 0.07 0.05 0.03

flowering in 64 days under stress, their improved harvest

index and yield under severe stress appears to result from

drought tolerance rather than avoidance. In the MSU and

SSU environments, stress was imposed at least 15 days

before the earliest varieties flowered, causing a reduction in

biomass production of over 30% relative to the favorable

NSU environment. This reduction is proportionately less for

the EDT types than for other variety types but is not

consistent with the hypothesis that the drought-tolerant class

merely escaped stress at flowering. The mechanism of

tolerance may be related to a flowering program in which

peduncle elongation and floret opening occur in spite of

stress. Drought-tolerant varieties exhibited a slight reduction

in days to flower under severe stress relative to the irrigated

lowland environment, whereas for all other cultivar classes,

flowering was delayed by 2 weeks or more as a result of

severe stress (Table 8).

Other features of the aerobic rice cultivars studied in this

experiment include intermediate plant height and panicle

number under favorable upland conditions relative to

irrigated and improved upland varieties. The IHF varieties

had a mean height of approximately 110 cm (data not

shown) compared to means of 93 cm and 126 cm for HYV

and JLF varieties, respectively. The semi-dwarf stature of

irrigated HYVs appears to be detrimental to aerobic

adaptation, but the tall, weak-strawed improved JLF types

are not sufficiently lodging-resistant to support the relatively

high levels of N application needed to achieve high-aerobic

roved indica upland types selected in high-fertility environments (IHF),

s selected for low-fertility and drought-prone environments (JLF and EDT,

ly water-stressed Severely water-stressed

upland

Infertile

upland

3.24 4.15

3.63 4.55

2.24 4.88

3.31 3.05

3.23 3.45

1.54 0.97

and types selected in high-fertility environments (IHF), traditional tropical

ility and drought-prone environments (JLF and EDT, respectively) evaluated

ly water-stressed Severely water-stressed

upland

Infertile

upland

0.17 0.25

0.15 0.28

0.10 0.25

0.32 0.23

0.09 0.20

0.08 0.09


Table 9

Mean number of productive tillers m�2 of irrigated high-yield varieties (HYV), improved indica upland types selected in high-fertility environments (IHF),

traditional tropical japonica varieties (JTV), and improved tropical japonica types selected for low-fertility and drought-prone environments (JLF and EDT,

respectively) evaluated in five environment types in southern Luzon, 2000–2003


Irrigated

lowland

Non-stressed

uplands

Moderately water-stressed

uplands

Severely water-stressed

uplands

Infertile

uplands

HYV 354 465 576 428 N/A

IHF 237 318 388 353 183

JLF 255 319 389 276 222

EDT 265 327 367 389 236

JTV 182 213 251 186 99

L.S.D.0.05 63 51 36 59 204

Table 10

Grain yield (t ha�1) of upland and aerobic rice cultivars under high-input

management at short or long-season locations in Yunnan, 1993–2002

Cultivar Type Grain yield (t ha�1)

Short-season Long-season

Mengwanggu Traditional japonica 2.57 � 0.10 2.92 � 0.12

IRAT 104 Improved japonica 2.37 � 0.26 3.37 � 0.39

B6144F-MR-6-0-0 Improved indica 1.70 � 0.17 4.13 � 0.20

Yunlu 29 Improved japonica 3.34 � 0.20 3.29 � 0.24

Yunlu 52 Improved japonica 2.86 � 0.23 4.00 � 0.28

Table 8

Mean days to flower of irrigated high-yield varieties (HYV), improved indica upland types selected in high-fertility environments (IHF), traditional tropical

japonica varieties (JTV), and improved tropical japonica types selected for low-fertility and drought-prone environments (JLF and EDT, respectively) evaluated

in irrigated lowland and severely stressed upland environments at IRRI, 2000–2003

Variety type Days to flower in

irrigated lowland

Days to flower in severely

water-stressed upland

Flowering delay

due to stress

HYV 82.6 98.1 15.5

IHF 80.7 94.3 13.6

JLF 90.1 108.0 17.9

EDT 68.8 64.4 �4.4

JTV 90.6 106.0 15.4

rice yields. The panicle number of all improved upland

variety types was significantly higher than those of JTV

types, but lower than that of irrigated lowland types

(Table 9). IHF and JLF varieties produced 100–200 fewer

panicles m�2 than HYVs in all environments. The high-

yielding Brazilian aerobic cultivars are also characterized by

moderate tillering (250 tillers m�2; Pinheiro and de Castro,

2000). However, increased panicle number may be an

important component of higher harvest index and yield

potential under aerobic management. High panicle number

is a feature that distinguishes the high-yielding semi-dwarf

hybrid Magat, which can produce over 700 panicles m�2 in

aerobic experiments (George et al., 2002). Selection for

increased panicle number appears to be a promising avenue

for increasing aerobic rice grain yield.

In general, these results indicate that the main feature

distinguishing improved aerobic-adapted rice varieties of the

IHF type from other cultivar types is their ability to retain

both high-biomass production and relatively high-harvest

index under favorable and moderately stressful upland

conditions. Irrigated rice varieties produced less biomass

and partitioned less of it to grain under upland management.

Traditional tropical japonicas and improved JLF varieties

produced as much or more biomass as aerobic varieties

under upland conditions, but partitioned much less to grain.

The high harvest index of IHF varieties may result from a

lower sensitivity to reproductive-stage water stress than all

but the most drought-tolerant upland materials, without

sacrifice of the biomass production needed for high-yield

potential.

3.2. Performance of aerobic rice varieties over sites and

years in southern Yunnan, China, and in Batangas,

Philippines

Cultivar means for the experiments conducted in Yunnan

are presented in Table 10. Means over all cultivars, seasons,

and locations were 3.36 t ha�1 for the longer season sites,

and 2.72 t ha�1 for the shorter-season sites. F-tests of both

cultivar and cultivar � elevation group interaction were

highly significant (data not shown), indicating that different

cultivars were favored in the two types of environment.

Yunlu 29 was the highest-yielding cultivar in the short-

season, high-elevation environments. With a mean yield of

3.34 t ha�1 in these environments, it outyielded the

traditional check by about 30%. B6144F-MR-6-0-0, an

improved indica introduction from Indonesia, and Yunlu 52,

a locally bred improved tropical japonica upland rice, were

the highest-yielding cultivars in longer season environments.


Table 11

Mean grain yield for aerobic and traditional upland rice varieties grown under farmer and researcher management in a traditional rice–maize double-cropping

system in Batangas Province, Philippines, 2002

Cultivar Cultivar

type

No. of farmer-

managed

trials

Grain yield

under farmer

management (t ha�1)

Yield under

researcher

management (t ha�1)

Duration in

researcher-managed

trials (days)

IR55423-01 Aerobic 16 3.41 � 0.26 3.90 123

IR74371-46-1-1 Aerobic 8 3.42 � 0.33 3.81 112

Azucena Traditional 18 2.04 � 0.25 2.15 125

Dinorado Traditional 21 2.28 � 0.24 2.04 131

Farmers’ variety Mixed 58 3.14 � 0.19 N/A N/A

L.S.D.0.05 0.62 1

With mean yields of over 4.0 t ha�1, they outyielded the

traditional check by about 40%.

Means for yield and crop duration in on-farm upland rice

trials in Batangas are presented in Table 11. Mean yields of

traditional tropical japonica varieties averaged over 3 t ha�1

for 66 reporting farms. The improved indica varieties, with

mean yields of approximately 3.4 t ha�1, outyielded the

traditional tropical japonica varieties by over 50% under

farmer management, and by nearly 100% in researcher-

managed trials. They also outyielded the mixture of

traditional and improved varieties used by farmers, and

were preferred because of their reduced duration and

improved lodging resistance.

4. Discussion

4.1. Key attributes of aerobic cultivars, and

implications for aerobic rice breeding

Rice cultivars for aerobic systems need to combine high-

biomass production, harvest index, and lodging resistance

with moderate drought tolerance, particularly at the highly

sensitive reproductive stage (Garrity and O’Toole, 1994).

Because aerobic systems depend on direct seeding in dry

soil, without accumulation of standingwater, vigorous early

growth is also needed to compete with weeds and to root

deeply to avoid early season drought. The results of this

study indicate that, for the Asian tropics, these traits are

most likely to be found in varieties that are primarily of

indica germplasm. Traditional tropical japonica varieties

and improved varieties derived predominantly from them

are low-tillering and have low HI; they are, therefore,

unlikely to have the yield potential or early vegetative vigor

needed for intensively managed upland systems. High-

yielding cultivars that have emerged from Asian breeding

programs for the favorable uplands are often rather closely

related to elite tropical irrigated varieties, usually through

IR8 and its relatives. For example, examination of the

pedigree of the upland rice variety IR55423-01, one of the

highest-yielding pure-line varieties under our aerobic

management trials, shows that its coefficient of coparentage

of with IR8 is only slightly less than that of the irrigated

varieties IR64 and IR72 (Atlin and Lafitte, 2002). IR55423-

01 does not, however, appears to carry the sd-1 allele from

its irrigated parents, and is of intermediate height. Varieties

like IR55423-01, which is capable of maintaining HI under

moderate water deficit, should be evaluated more thor-

oughly in order to discover mechanisms that underlie this

ability.

Recent research has shown that drought tolerance and

high spikelet fertility under reproductive-stage stress in

aerobic-adapted germplasm are strongly associated with

short duration and minimal flowering delay under stress.

These are likely to be key attributes of upland-adapted

cultivars targeted at drought-prone or severely water-

restricted production environments (Lafitte and Courtois,

2002), and may be useful as screening criteria for drought

tolerance. However, the most efficient approach to the

identification of cultivars adapted to any target environ-

ment is usually to screen directly for performance in that

environment, unless the repeatability (H) of grain yield

estimates in the target system is low (Atlin, 2001). In rice,

H of grain yield estimates in variety trials does not appear

to be much affected by soil water status per se (Atlin and

Lafitte, 2002), although uneven application of irrigation

water to the experimental field can greatly increase error

variances and reduce precision. Direct selection under

high-input aerobic management, is therefore, likely to be

required to optimize performance under aerobic manage-

ment. Because the results of this study indicate that yield in

well-watered environments is not highly correlated with

yield under moderate stress, early screening of breeding

lines under both high-fertility aerobic management in the

wet season and, where possible, moderate water stress in

the dry season appears to be the best approach to

developing cultivars combining high-yield potential and

moderate stress tolerance.

4.2. Implications for aerobic rice target environments

The Yunnan and Batangas results confirm that cultivars

suited to aerobic management systems using moderate N

fertilization rates are currently available, both for hilly areas


of southeast Asia that have traditionally relied on shifting

cultivation, and for permanent arable upland cropping

systems of the type found in many parts of south and

southeast Asia, where rice is rotated annually or double-

cropped with other upland crops. About 100,000 ha of

upland rice are grown annually in southern Yunnan, and the

system in Batangas covers several thousand hectares.

Currently available improved upland/aerobic rice varieties

can consistently deliver wet-season yields of between 3

and 4 t ha�1 without supplemental irrigation in these

systems. Such cultivars are the basis for more productive

and sustainable upland cropping systems that are evolving

in several parts of southeast Asia. The Batangas results,

which were generated in farmers’ fields in an indigenous

cropping system in which upland rice is grown once a year,

with one intervening non-rice crop, show that high-aerobic

rice yields can be sustainable in annual double-crop

rotations.

The yields obtained with tropical aerobic varieties are

comparable to those obtained in favorable rainfed lowland

environments in south and southeast Asia, indicating that

aerobic rice-based production systems are likely to be

attractive to farmers in drought-prone rainfed lowlands as

well as in uplands. In drought-prone lowland rice-producing

areas like the Bastar Plateau in Chhattisghar, India, or the

light-soil areas of northeastern Thailand and southern Laos,

yields average less than 3 t ha�1 and drought risk is high. In

these areas, failure to accumulate standing water in bunded

fields can delay transplanting (Rickman et al., 2001) or may

severely disrupt water-dependent weed control operations,

such as beusani (Tomar, 2002). Dry direct-seeded aerobic

rice systems, which do not require standing water to

accumulate either for establishment or weed management,

result in early establishment and more reliable weed control,

improving the synchrony between crop development and

seasonal rainfall. Weed pressure is a major constraint to the

adoption of direct seeding with conventional lowland

germplasm. However, upland-adapted indica rice varieties

tend to be moreweed-competitive than either lowland HYVs

or tropical japonica types (unpublished data), and may

facilitate the adoption of risk-avoiding direct-seeding

technology.

5. Conclusion

Medium-height, aerobic-adapted cultivars consistently

yielding approximately 4 t ha�1 under farmer management

in rainfed conditions are now available to rice growers in the

Asian tropics. These cultivars, which are usually of the

indica type, combine a moderate level of drought tolerance

with input responsiveness. They could permit the replace-

ment of shifting cultivation systems with more productive,

more sustainable, and less extensive systems in southeast

Asia, and may permit rainfed lowland rice growers in

drought-prone areas to adopt risk-avoiding direct-seeding

systems. Mechanisms underlying the high-harvest index of

these cultivars under moderate water stress require further

study, but breeding programs focusing on screening

improved, indica-derived germplasm under both favorable

and water-short conditions are making rapid progress in

developing a new class of rice cultivar for water-short

environments.

References

Atlin, G.N., 2001. Breeding for suboptimal environments. In: Fukai, S.,

Basnayaka, J. (Eds.), Increased Lowland Rice Production in theMekong

Region. ACIAR Proceedings 101, 245–251.

Atlin, G.N., Lafitte, H.R., 2002. Developing and testing rice varieties for

water-saving systems in the tropics. In: Bouman, B.A.M., Hengsdijk,

H., Hardy, B., Bindraban, P.S., Tuong, T.P., Ladha, J.K. (Eds.), Water-

Wise Rice Production. Proceedings of the International Workshop on

Water-Wise Rice Production. International Rice Research Institute, Los

Banos, 356 pp., 8–11 April.

Bouman, B.A.M., Yang, X., Wang, H., Wang, Z., Zhao, J., Chen, B., 2005.

Performance of Aerobic Rice Varieties Under Irrigated Conditions in

North China (this volume).

Castaneda, A.R., Bouman, B.A.M., Peng, S., Visperas, R.M., 2002. The

potential of aerobic rice to reduce water use in water-scarce irrigated

lowlands in the tropics. In: Bouman, B.A.M., Hengsdijk, H., Hardy, B.,

Bindraban, P.S., Tuong, T.P., Ladha, J.K. (Eds.), Water-Wise Rice

Production. Proceedings of the International Workshop on Water-Wise

Rice Production. International Rice Research Institute, Los Banos, 356

pp., 8–11 April.

Garrity, D.P., O’Toole, J.C., 1994. Screening rice for drought resistance at

the reproductive phase. Field Crops Res. 39, 99–110.

George, T., Magbanua, R., Laza, M., Atlin, G., Virmani, S., 2002. Magat, a

wetland semi-dwarf hybrid rice for high yielding production on irrigated

dryland. Int. Rice Res. Newslett. 27, 26–28.

Glazsmann, J.C., 1987. Isozymes and classification of Asian rice varieties.

Theor. Appl. Genet. 74, 21–30.

Lafitte, H.R., Courtois, B., 2002. Interpreting cultivar � environment

interactions for yield in upland rice: assigning value to drought-adaptive

traits. Crop Sci. 42, 1409–1420.

Pinhiero, B.da S., de Castro, E.daM., Guimaraes, C.M., 2005. Sustainability

and profitability of aerobic rice production in Brazil. Field Crops Res.

(this volume).

Rickman, J.F., Pyseth, M., Bunna, S., Sinath, P., 2001. Direct seeding of rice

in Cambodia. In: Proceedings of an International Workshop, ACIAR

Proceedings No. 101, Vientiane, Laos, 30 October–2 November.

SAS Institute, 1992. SAS/STATUser’s Guide, Version 6, 4th ed., vols. 1 and

2. SAS Inst., Cary, NC.

Tomar V.S., 2002. The beushening system of rice crop establishment in

Eastern India. In: Pandey, S., Mortimer, M., Wade, L., Tuong, T.P.,

Lopez, K., Hardy, B. (Eds.), Direct Seeding in Asian Rice Systems:

Strategic Issues and Opportunities. Proceedings of a Workshop, 25–28

January 2000. Bangkok, Thailand. IRRI.

Wang, H., Bouman, B.A.M., Zhao, D., Wang, C., Moya, P.F., 2002.

Aerobic rice in northern China: challenges and opportunities. In:

Bouman, B.A.M., Hengsdijk, H., Hardy, B., Bindraban, P.S.,

Tuong, T.P., Ladha, J.K. (Eds.), Water-Wise Rice Production. Pro-

ceedings of the International Workshop on Water-Wise Rice Produc-

tion. International Rice Research Institute, Los Banos, 356 pp., 8–11

April.

Developing rice cultivars for high-fertility upland systems in the Asian tropics

Documents