Agronomy Journal Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (18)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Dobermann, A.
Right arrow Articles by Adviento, M. A. A.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Dobermann, A.
Right arrow Articles by Adviento, M. A. A.
Agricola
Right arrow Articles by Dobermann, A.
Right arrow Articles by Adviento, M. A. A.
Related Collections
Right arrow Soil Fertility and Productivity
Right arrow Tropical Soil Management
Right arrow Spatial Variability
Right arrow Rice
Right arrow Nutrient Cycling
Right arrow Nitrogen
Right arrow Site-Specific Analysis
Agronomy Journal 95:913-923 (2003)
© 2003 American Society of Agronomy

FERTILIZER MANAGEMENT

Soil Fertility and Indigenous Nutrient Supply in Irrigated Rice Domains of Asia

A. Dobermann{dagger},*,a, C. Witta, S. Abdulrachmanb, H. C. Ginesc, R. Nagarajanh, T. T. Sond, P. S. Tane, G. H. Wangf, N. V. Chiend, V. T. K. Thoad, C. V. Phunge, P. Staling, P. Muthukrishnang, V. Ravih, M. Babuh, G. C. Simbahana and M. A. A. Advientoa

a Int. Rice Res. Inst. (IRRI), DAPO Box 7777, Manila, Philippines
b Res. Inst. for Rice (RIR), Sukamandi, Indonesia
c Philippine Rice Res. Inst. (PhilRice), Maligaya, Nueva Ecija, Philippines
d Natl. Inst. for Soils and Fert. (NISF), Hanoi, Vietnam
e Cuu Long Delta Rice Res. Inst. (CLRRI), Omon, Cantho, Vietnam
f Zhejiang Univ. (ZU), Hangzhou, P.R. China
g Tamil Nadu Rice Res. Inst. (TNRRI), Aduthurai, Tamil Nadu, India
h Soil and Water Manage. Res. Inst. (SWMRI), Thanjavar, Tamil Nadu, India

* Corresponding author (adobermann2{at}unl.edu)

Received for publication July 2, 2002.

   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Knowledge-intensive approaches have been proposed to manage the variability in indigenous nutrient supplies (IS) in irrigated rice (Oryza sativa L.) systems. On-farm experiments were conducted at 155 locations in seven domains of Asia to quantify the variability of soil properties, grain yield, and nutrient uptake in N, P, and K omission plots (0-N, 0-P, and 0-K, respectively). Except for pH, coefficients of variation of soil properties within a domain ranged from 17 to 43%. Similar ranges were measured for grain yield and plant nutrient uptake in nutrient omission plots, which served as crop-based estimates of indigenous N, P, and K supply. Soil properties showed little association with plant nutrient uptake or grain yield in nutrient omission plots. Mean grain yields in nutrient omission plots increased in the order 0-N (3.9 Mg ha-1) < 0-K (5.1 Mg ha-1) <= 0-P (5.2 Mg ha-1). Soils, climate, and crop management caused large variability of IS among irrigated rice domains, years, growing seasons, and fields within a domain. Grain yield and nutrient uptake in omission plots were mostly higher in high-yielding than in low-yielding climatic seasons. No changes in indigenous N supply occurred for periods of 4 to 6 yr in the same seasons. Grain yields in nutrient omission plots were strongly correlated with each other and also with the yield in the fertilized farmers' fields. Fertilizer recommendations should be fine-tuned to spatial domains with relatively uniform agroecological characteristics, cropping practices, and socioeconomic conditions. Within such domains, season-specific management of the IS variability can include field-specific approaches.

Abbreviations: AD, Aduthurai • CV, coefficient of variation • df, degrees of freedom • FFP, farmers' fertilizer practice • HA, Hanoi • HYS, high-yielding season • IKS, indigenous potassium supply • INS, indigenous nitrogen supply • IPS, indigenous phosphorus supply • IS, indigenous supply of a nutrient • JI, Jinhua • LYS, low-yielding season • MA, Maligaya • OM, Omon • SSNM, site-specific nutrient management • SU, Sukamandi • TH, Thanjavur • 0-K, potassium omission (plot) • 0-N, nitrogen omission (plot) • 0-P, phosphorus omission (plot)


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
IRRIGATED DOUBLE- and triple-crop continuous rice and rice–upland crop systems in Asia present significant challenges for improving nutrient management because yield growth rates have slowed down in recent years (Dobermann and Cassman, 2002). Farm sizes in these environments typically range from about 0.3 ha to more than 5 ha, whereas individual fields are mostly 0.1 to 0.5 ha in size. Recent case studies have demonstrated that soil fertility, fertilizer use, and crop response to nutrient inputs may vary widely among regions, among rice fields within smaller irrigated and rainfed rice environments, and also from season to season in the same field (Angus et al., 1990; Cassman et al., 1996a; Olk et al., 1999; Adhikari et al., 1999). At present, however, many fertilizer recommendations are only given for larger areas, with little differentiation according to major agroecological zones, soil types, cropping systems, or field-specific information. Managing the location- and season-specific variability in nutrient supply is a key strategy to overcome the current mismatch of fertilizer rates and crop nutrient demand in irrigated rice environments (Dobermann and Cassman, 2002).

A new site-specific nutrient management (SSNM) approach for rice has been evaluated recently at numerous locations in Asia (Wang et al., 2001; Dobermann et al., 2002). In this work, field-specific, balanced amounts of N, P, and K were prescribed based on crop-based estimates of the indigenous supply of N (INS), P (IPS), and K (IKS) and by modeling the expected yield response as a function of nutrient interactions and climatic yield potential (Dobermann and White, 1999; Witt et al., 1999; Dobermann et al., 2002). In addition, timing and amount of N applications were fined-tuned to crop needs based on leaf chlorophyll content (Peng et al., 1996) measured at critical growth stages of rice. At issue is whether such knowledge-intensive nutrient management concepts can be simplified for wider-scale dissemination without loosing precision and gains in yields, profitability, and nutrient use efficiency.

Adoption of SSNM requires understanding and quantification of the indigenous supply of macronutrients for a larger fertilizer recommendation domain or for a specific rice field. Following Janssen et al. (1990), we define the effective IS as the cumulative amount of that nutrient originating from all indigenous (nonfertilizer) sources that circulates through the soil solution surrounding the entire root system during one complete crop cycle. In an irrigated system, IS includes plant available nutrients derived from (i) chemical and biological transformations of soil solids, (ii) biological N2 fixation in the floodwater–soil system, (iii) atmospheric deposition, and (iv) solutes and sediments deposited by irrigation and/or naturally occurring flooding (Cassman et al., 1998; Dobermann et al., 1998). For practical purposes, IS can be measured as plant nutrient accumulation at crop maturity in a nutrient omission plot under well-managed field conditions, i.e., when all other nutrients except one are amply supplied and other limitations to growth such as water or pests are absent (Janssen et al., 1990).

The objective of this paper is to quantify the spatial and temporal variability of soil properties and crop-based estimates of INS, IPS, and IKS among and within typical irrigated rice domains. Emphasis will be given to a comparative analysis of major irrigated rice areas in tropical and subtropical Asia. Several specific questions must be addressed. How variable is IS in irrigated rice domains, and what are the main sources of this variability? How do climate and crop management affect the variability of crop-based IS estimates? How does the IS change over time? Are IS of different nutrients correlated with each other and related to the inherent soil quality or crop management? In a second paper (Dobermann et al., 2003), we discuss sampling strategies and the use of soil tests or simple measurements such as grain yield in nutrient omission plots for estimating IS in SSNM approaches.


   MATERIALS AND METHODS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
General Site Characteristics
On-farm experiments were conducted in seven irrigated rice production domains of Asia located in five countries (Table 1). Five sites, Maligaya (MA), Omon (OM), Sukamandi (SU), Aduthurai (AD), and Thanjavur (TH), represent rice monoculture systems of the humid or subhumid tropics. Two sites, Hanoi (HA) and Jinhua (JI), were located in subtropical regions. Sites were located in large inland plains and basins (MA, SU, and JI) or river deltas (OM, HA, AD, and TH) with generally flat topography. Soil types included Entisols, Inceptisols, Alfisols, Vertisols, and Ultisols (Soil Survey Staff, 1999). At all sites, irrigated rice has been grown for a long time, and early adoption of intensive cropping had occurred. At least two crops of rice are grown annually. Cropping intensities of three crops per year were common at two sites [OM: rice–rice–rice; HA: rice–rice–maize (Zea mays L.)].


View this table:
[in this window]
[in a new window]
  		Table 1. Description of the experimental locations.

 
Each site represents a spatial domain in which on-farm experiments were conducted in 15 to 26 rice farms located within a radius of typically 15 to 25 km around a research station, resulting in a total number of 155 farms. The domains varied in size but were mostly in the 100 to 200 km2 range. Farms were typically clustered into several villages and selected to represent different socioeconomic conditions and farm sizes and the most common soil types, cropping systems, and farm management practices in the region. At each farm, a single rice field served as the principal experimental unit.

On-Farm Nutrient Omission Plots
All experiments followed a standardized experimental protocol with minor location-specific modifications to account for differences in field sizes, climatic seasons, and crop establishment techniques (Dobermann et al., 2002). Treatments in the on-farm experiments included 0-N, 0-P, and 0-K.

Nitrogen Omission Plot
Either no fertilizer (–F plot, sites AD, MA, OM, SU, and TH) or 30 kg P ha-1 and 50 kg K ha-1 (+PK plot, sites HA and JI) was applied to strip plots (40–100 m2) or 6- by 6-m plots embedded in the farmers' field. This treatment was sampled for eight consecutive rice crops grown from 1997 to 2001 at each site to estimate the INS defined as plant N accumulation in grain and straw at physiological maturity in a 0-N plot. In addition, from 1995 to 1996, paired –F and +PK plots were sampled for three to four rice crops in all farms at AD, MA, OM, and SU.

Phosphorus Omission Plot
Only N and K were applied to strip plots (40–100 m2) or 6- by 6-m plots embedded in the farmers' field to ensure that macronutrients other than P did not limit plant P uptake from indigenous sources. Depending on the site and climatic season, the N rates varied from 120 to 180 kg N ha-1; K rates varied from 100 to 150 kg K ha-1. The 0-P plots were used to estimate the IPS, defined as plant P accumulation in grain and straw at physiological maturity in a 0-P plot, for three to four consecutive rice crops grown from 1997 to 1999.

Potassium Omission Plot
Only N and P were applied to strip plots (40–100 m2) or 6- by 6-m plots embedded in the farmers' field to ensure that macronutrients other than K did not limit plant K uptake from indigenous sources. Depending on the site and climatic season, the N rates varied from 120 to 180 kg N ha-1; P rates varied from 25 to 40 kg P ha-1. The 0-K plots were used to estimate the IKS, defined as plant K accumulation in grain and straw at physiological maturity in a 0-K plot, for three to four consecutive rice crops grown from 1997 to 1999.

Omission plots were established as duplicate or triplicate sets in each field and separated from the surrounding field by bunds. Omission plots were moved to a different location within the same field after each crop to represent normal conditions of IS under typical crop management, avoid nutrient depletion, and obtain estimates of the average field-specific IS over time. Depending on initial soil test data, blanket doses of other nutrients were applied to all treatments at selected sites to prevent deficiencies other than N, P, or K. This included application of Zn at MA, AD, and TH and a single Mg application on degraded soils at some HA sites. Varieties grown were chosen by the farmer, but they were the same in all treatments within each field. Only modern, high-yielding rice varieties with a harvest index of 0.45 to 0.5 were grown. At the JI site in China, most farmers grew hybrid rice during the late rice season (Wang et al., 2001). Transplanting with hill densities ranging from about 20 to 50 hills m-2 was the predominant form of crop establishment at five sites (AD, HA, JI, SU, and TH), whereas most farmers at MA and OM used direct broadcast seeding with high seed rates (typically 100–200 kg ha-1). Farmers did all water management and pest control following the commonly recommended methods. Surface water from rivers or reservoirs was the main source of irrigation at all sites. Where pest problems were suspected or observed, measures to either control them in advance (prophylactic) or correct them were implemented under the guidance of researchers.

Measurements
Initial soil samples for determination of general soil properties in 0- to 0.15-m depth were collected from each field in spring 1997 and analyzed following standard procedures (Ponnamperuma et al., 1981; van Reeuwijk, 1992). Thereafter, in two to four cropping seasons at each site, soil samples were also collected at 20 to 30 d after planting from the nutrient omission plots and analyzed for organic C (Walkley, 1947) and total N (Bremner, 1996) in 0-N plots, 0.5 M NaHCO3 extractable P (Olsen et al., 1954) and Bray-I P (Bray and Kurtz, 1945) in 0-P plots, and 1 M ammonium acetate extractable K (van Reeuwijk, 1992) in 0-K plots. Initial soil data and those from different cropping seasons were pooled to obtain averages of extractable nutrients for each field (Table 2).


View this table:
[in this window]
[in a new window]
  		Table 2. Status and variability of soil fertility in 0- to 0.15-m depth in 155 irrigated rice fields of Asia.

 
Plant sampling procedures followed a standard procedure at all experimental sites (Witt et al., 1999; Dobermann and Fairhurst, 2000). In each omission plot, a 12-hill plant sample (or two 0.5-m2 samples in direct-seeded rice) was collected at physiological maturity of rice for determination of components of yield, harvest index, and nutrient concentrations in plant tissue. Grain yields were obtained from a central 5-m2 harvest area in each plot at harvestable maturity and are reported at standard moisture content of 0.14 g H2O g-1 fresh weight. Grain and straw subsamples from the 12-hill sample were dried to constant weight at 70°C. Straw yields were estimated from the oven-dry grain yield of the 5-m2 harvest area and the grain/straw ratio of the 12-hill sample. Nitrogen concentrations in grain and straw were measured by micro-Kjeldahl digestion, distillation, and titration; tissue P by the molybdenum-blue colorimetric method; and tissue K by atomic adsorption spectrophotometer after wet digestion (Walinga et al., 1995). Note that at the OM site, no 0-P and 0-K plots were established in the 1999 dry season (only three crops with 0-P and 0-K plots), whereas at MA, no plant P and K determinations were conducted in 0-P and 0-K plots in the 1997 dry season. At all sites, plant samples were also collected from two to three 6- by 6-m sampling plots within the farmers' field surrounding the nutrient omission plots (Dobermann et al., 2002).

Data Analysis
For statistical analysis, the two main rice seasons at each site were classified as either high-yielding seasons (HYS) or low-yielding seasons (LYS). At the tropical sites, the HYS is usually a dry season with low rainfall and high solar radiation, whereas the LYS is a wet (monsoon) season with lower yield potential due to cloudy conditions and high rainfall. Sukamandi was the only exception to this rule because dry-season yields at this site are lower than wet-season yields so that the latter was defined as HYS. For the two subtropical sites, the HYS is the late rice crop at JI and the early (spring) rice crop at HA (Table 1).

The PROC MIXED procedure of the SAS/STAT software (SAS Inst., 1999) was used to perform analysis of variance (ANOVA) for plant nutrient uptake and grain yield measured in omission plots across all sites and seasons. For 0-N plots, this analysis included measurements of eight crops sampled from 1997 to 2001. For 0-P and 0-K plots, the ANOVA was based on three to four crops sampled at each site from 1997 to 1999. A mixed ANOVA model (Littell et al., 1996) with site [degrees of freedom (df) = 6], season (df = 1), and site x season (df = 6) as fixed effects and farm within site (subsequently designated as farm[site]), year, year x season, season x farm[site], year x site, year x farm[site], year x season x site, and year x season x farm[site] as random effects was used to assess the overall variability of IS measurements. Note that site represents one of the seven rice production domains sampled, season refers to HYS and LYS (Table 1), and farm represents single rice fields nested within each site. To test the fixed effects, PROC MIXED uses the generalized F ratio. In the computation of F ratios, the restricted maximum likelihood (REML) estimates of the variance components were used while the generalized least squares were used to estimate fixed effects. The containment method was used to compute for the denominator df of the F value. Under this method, the denominator df for a test of a fixed effect are calculated by searching for random terms that syntactically contain the fixed effect of interest. The df of the random effect that has the minimum rank contribution to the XZ matrix are used as the denominator df. (X is the design matrix for the fixed effects while Z is the design matrix for the random effects.)

To analyze whether trends in INS over a period of 8 to 13 successive rice crops (4–6 yr) were significantly different from zero, data of each field sampled were analyzed by ordinary least squares linear regression of measured INS against a time trend variable:

[1]
where y is the INS (kg N ha-1), a is a constant, t is the number of the successive rice crop grown, and b is the slope or magnitude of the INS trend (kg N ha-1 per crop). A statistically significant positive or negative yield trend was recorded only if the null hypothesis of a zero slope for the time trend variable could be rejected at a 5% level of significance (two-tail test).

Using fuzzy-k-means clustering (McBratney and De Gruijter, 1992), farms (fields) were classified into different categories based on field-specific average grain yields in 0-N, 0-P, and 0-K plots measured from 1997 to 1999. Fuzzy classification was conducted using Mahalanobis distance, a fuzzy exponent of 1.3, and a range of 2 to 10 classes (Minasny and McBratney, 2002). The optimum number of classes was established on the basis of minimizing the fuzziness performance index and the modified partition entropy (Roubens, 2001). The efficacy of the classification in partitioning the variance of various soil and crop characteristics was then evaluated using the intraclass correlation, {rho}i, defined as (Webster and Oliver, 1990):

[2]

Analogous to the r2 value in a regression analysis, {rho}i measures the proportion of variance explained by the classification in relation to the total variance between ({sigma}2b) and within ({sigma}2w) classes. Analysis of variance was used to estimate {sigma}2b and {sigma}2w.


   RESULTS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Soil Fertility Status in Irrigated Rice Domains
The sample of 155 rice farms in this study represents major irrigated rice environments in south and southeast Asia. In addition to the mean values of soil properties shown in Table 2, it should be noted that soil nutrient levels were below the commonly used critical levels of 10 mg kg-1 P and 0.2 cmolc kg-1 K (Dobermann and Fairhurst, 2000) in 49% of all farms for P and 39% for K, respectively.

Soils varied from coarse-textured soils at TH to heavy clays at OM, MA, and AD sites, and large differences in soil fertility occurred among sites. Average soil organic C was only 7 to 10 g kg-1 at AD and TH, whereas it was 18 to 19 g kg-1at OM and JI sites. Average Olsen P was only about 5 mg kg-1 at MA, OM, and SU, whereas site averages where in the 21 to 28 mg kg-1 range at JI, AD, and TH. Average extractable soil K was 0.18 to 0.19 cmolc kg-1 at HA, MA, and OM but ranged from 0.28 to 0.51 cmolc kg-1 at all other sites. Soil organic C (19 g kg-1), Olsen P (21 mg kg-1), and extractable K (0.28 cmolc kg-1) were all relatively high at JI in southeast China, which probably reflects a history of farmyard and green manure use as well as high rates of P and K fertilizer promoted in recent years (Wang et al., 2001). Except for soil pH, average coefficients of variation (CVs) of soil fertility characteristics within a domain ranged from 17 to 43% (Table 2).

Variability of Crop-Based Estimates of Indigenous Nutrient Supplies
Mean grain yields in nutrient omission plots are shown in Table 3. Yield increased in the order 0-N (3.9 Mg ha-1) < 0-K (5.1 Mg ha-1) <= 0-P (5.2 Mg ha-1). For comparison, average rice yield was 5.2 Mg ha-1 in the fertilized areas of the same farmer's fields. Indigenous nutrient supplies in all omission plots sampled from 1997 to 2001 ranged from 7 to 130 kg N ha-1, 3 to 34 kg P ha-1, and 27 to 213 kg K ha-1.


View this table:
[in this window]
[in a new window]
  		Table 3. Variability of grain yield and plant nutrient accumulation in nutrient omission plots in 155 irrigated rice fields of Asia. Descriptive statistics are based on eight (0-N) or three to four (0-P and 0-K) successive rice crops sampled in each field from 1997 to 2001.

 
Highly significant differences in grain yield or nutrient uptake in omission plots occurred among sites (Table 4 and Fig. 1). Average INS was lowest at OM and TH (38 kg N ha-1), whereas soils at the JI site provided on average 69 kg N ha-1 from indigenous sources. Indigenous P supply was lowest at OM and SU (11–12 kg P ha-1), whereas farms at JI averaged about 21 kg P ha-1. Average IKS ranged from 58 kg K ha-1 on coarse-textured soils of the Cauvery Delta (TH) to 120 kg K ha-1 at JI in China. Year x site effects contributed little to the overall variation (0–6%), suggesting that the year-to-year variation at individual sites was too small to significantly affect the across-site comparison.


View this table:
[in this window]
[in a new window]
  		Table 4. Variation in indigenous nutrient supplies among cropping seasons and sites (domains) in irrigated rice domains of Asia, and relative contribution of different sources of spatial and temporal variation to the total variance observed.

 


View larger version (20K):
[in this window]
[in a new window]
  		Fig. 1. Effect of climatic season on grain yield and plant N accumulation in N omission (0-N) plots sampled from 1997 to 2001 in seven irrigated rice domains. Closed circles represent high-yielding seasons (HYS) with favorable climate, whereas open circles represent low-yielding seasons (LYS) at each site. Error bars indicate the standard error of the means. AD, Aduthurai; HA, Hanoi; JI, Jinhua; MA, Maligaya; OM, Omon; SU, Sukamandi; TH, Thanjavur.

 
Grain yield and nutrient uptake in omission plots were consistently higher in HYS than in LYS crops for all three nutrients. The average difference between HYS and LYS was 27 and 29% for grain yield measurements compared with 15 and 20% for N, P, or K uptake measured in omission plots, but means were only statistically significant (P < 0.05) for 0-N plots (Table 4). Whereas N data included 4 yr of consecutive sampling, 0-P and 0-K plots were only sampled for a 2-yr period, and this period was associated with significant climatic variability, including an El Niño cycle and other climatic events (Dobermann et al., 2002). Although, across all sites, site x season effects were not statistically significant at the P < 0.05 level, site differences occurred (Fig. 1 and 2). At HA, JI, and MA, for example, average plant N accumulation in 0-N plots was similar in HYS and LYS crops, but HYS grain yields were significantly higher than those in LYS crops because nonnutrient-related yield limitations occurred during the LYS, particularly during the grain-filling period. At other sites (OM and SU), both N uptake and grain yield in 0-N plots were much lower in the LYS than in the HYS (Fig. 1 and 2), which reflects broader growth limitations in LYS crops during the entire growth period due to less solar radiation and humid conditions but also differences in INS that are associated with the length and aeration of the soil during fallow periods preceding crops.



View larger version (95K):
[in this window]
[in a new window]
  		Fig. 2. Seasonal and spatial variability of plant N accumulation and rice grain yield measured in N omission (0-N) plots in farmers' fields at Omon (Mekong Delta, Vietnam) and Maligaya (Central Luzon, Philippines). The boxplots show medians (horizontal lines), 25 to 75% quartile ranges (white boxes), and the 10 and 90% percentiles (error bars) at each site. HYS, high-yielding season; LYS, low-yielding season.

 
Spatial variation among farms within a domain contributed 14 to 25% of the overall variation in grain yield or nutrient uptake (Table 4). This proportion was highest for INS (23–25%) and lowest for IPS (14–18%). Year x farm effects contributed another 7 to 9% to the total variation in crop-based INS estimates, 0 to 0.3% for IPS, and 2.4 to 4.5% for IKS (Table 4). This suggests that in some years, variability among farms may be less or more than in others, particularly for INS (Fig. 2) because of the very dynamic nature of N in the soil and the many climatic and management factors affecting it (Cassman et al., 1996a; Cassman et al., 1998). Average CVs among fields within a small domain sampled ranged from 12 to 37% for grain yield or 18 to 36% for nutrient uptake in omission plots (Table 3). Some differences occurred among the domains sampled. For example, average CVs of grain yield and N uptake in 0-N plots were lowest at AD and JI (13–14%) and highest at SU (37%, data not shown). Farms at the SU site were clustered in three villages, each representing a different soil type and belonging to a different irrigation scheduling group, which caused significant variation in fallow periods, N mineralization, and planting dates.

Temporal variation among crops sampled contributed most to the overall variation of either grain yield or nutrient uptake in omission plots, but crop effects varied among sites and farms. About 30 to 45% of the overall variance was contributed by year x season x farm interactions, indicating that season effects varied randomly from year to year and farm to farm. Another 13 to 27% of the overall variability was due to year x season x site interactions (Table 4), indicating that season x site interaction varied randomly from year to year. Residual effects due to spatial variation among omission plots within the same field and measurement error accounted for 9 to 12% (grain yield) or 16 to 22% (nutrient uptake) of the overall variance.

Crop-based estimates of IS are also affected by the quality of crop management, which affects stresses other than the nutrient of interest. Effects of the overall crop management quality on estimates of INS, IPS, and IKS were assessed for sites at which crop management problems were more commonly observed in some farms (AD, MA, OM, and SU). A crop management quality score was assigned to each crop–farm data set collected, based on observations of land preparation, water supply and management, weeds, rats, snails, insect pests, diseases, and other problems such as lodging, seed quality, or typhoons (Dobermann et al., 2002). During the 1997 to 1999 period, mean INS was 57 kg N ha-1 in fields with high management quality (no problems observed), 55 in fields with medium quality (some problems observed), and 51 in fields with low quality (severe problems observed). Both IPS (16, 14, and 12 kg P ha-1, respectively) and IKS (86, 90, and 73 kg K ha-1, respectively) declined in similar order.

Trends in Indigenous Nitrogen Supply over Time
Despite seasonal fluctuations and large spatial variability within seasons at each site (Fig. 2), average INS trends over periods of 4 to 6 yr were not statistically different from zero (Table 5). On an individual farm basis, a negative trend in INS was fitted for 68% of all farms, but statistically significant negative trends (P < 0.05) were only observed in nine farms at AD and JI sites. Similarly, a positive trend was fitted for 32% of all farms, but positive trends were statistically significant in just three cases (data not shown).


View this table:
[in this window]
[in a new window]
  		Table 5. Medium-term trends in indigenous N supply (INS) in irrigated rice domains of Asia. Statistics shown are based on field-specific linear regression functions fitted to INS measurements conducted in 8 to 13 rice crops grown in each farm.

 
The average slope was -0.5 kg N ha-1 per rice crop grown and varied among sites from -1.6 to +0.9 kg N ha-1 crop-1. However, at all sites, the average p value of the t statistic was 0.28 to 0.5, suggesting no significant changes of INS over time. There was no relationship between the initial INS values or fitted intercepts and the trend in INS over time (data not shown).

Relationships between Grain Yields in Different Treatments
Grain yields in 0-N, 0-P, and 0-K plots were strongly correlated with each other (Fig. 3). The closest relationship was found between grain yield in 0-P and 0-K plots (R2 = 0.86). Yields were generally similar in 0-P and 0-K plots but higher than in 0-N plots (Fig. 3). Grain yields in 0-N, 0-P, and 0-K plots were also strongly correlated with the yield in the field surrounding the omission plots [Fig. 4, farmers' fertilizer practice (FFP)], which received varying amounts of the three macronutrients. Fertilizer rates in the FFP averaged 117, 18, and 31 kg ha-1 N, P, and K, respectively, per crop. Of the 155 farmers, only five did not apply any P fertilizer, and 14 did not apply K during the experimental period. In general, grain yields were highest in farms with high indigenous supply of all three nutrients. A stepwise linear regression based on field-specific yield averages of four crops (1997–1999) resulted in the following model:

where Y represents the different treatment yields (Mg ha-1) and SEE is the standard error of estimate. Partial correlation coefficients in this model were 0.31 for Y0-N, 0.10 for Y0-P, and 0.39 for Y0-K, suggesting, in most cases, greater importance of INS and IKS with the present approaches of farmers' fertilizer management.



View larger version (52K):
[in this window]
[in a new window]
  		Fig. 3. Correlation between grain yield measured in N (0-N), P (0-P), and K (0-K) omission plots in the same field. Values shown are data of three to four consecutive rice crops grown from 1997 to 1999 in 155 farms at seven sites.

 


View larger version (49K):
[in this window]
[in a new window]
  		Fig. 4. Relationships between grain yield in the farmers' fertilizer practice (FFP) and grain yield or nutrient uptake measured in N (0-N), P (0-P), and K (0-K) omission plots. Values shown are field average of four successive rice crops grown from 1997 to 1999.

 
Across the whole range, grain yield in the farmers' practice was higher than that in 0-N plots, suggesting a universal yield response to fertilizer N. However, the slope of this relationship (0.82, Fig. 4) indicates that yield increases due to fertilizer N decreased with increasing INS or grain yield in the 0-N plot. A negative correlation was found between grain yield in 0-N and the agronomic N use efficiency (r = -0.37, P < 0.001), suggesting that farmers did not adjust N rates and timing of N according to variation in INS. Similar observations were made in earlier studies (Olk et al., 1999). Grain yields in FFP were only higher than those in 0-P or 0-K plots if the latter were below about 4 Mg ha-1. In some cases, yields in 0-P or 0-K plots exceeded those obtained in the FFP (Fig. 4) because of higher N rates (120–180 kg N ha-1) and better timing of N applications in the 0-P and 0-K plots, indicating suboptimal N management in the FFP. Relationships between plant nutrient uptake in omission plots and grain yield in the FFP were generally more scattered than those obtained for grain yield measurements (Fig. 4). Scatter in these relationships is attributed to factors that affect yield but not nutrient uptake, but it also includes greater measurement error associated with measuring nutrient uptake. Of particular importance are climatic conditions and pest infestations during anthesis and grain filling of rice because they tend to affect yield after most of the total plant nutrient accumulation has been completed.

Fields endowed with better soil or owned by farmers who have always had better crop and soil management are likely to have higher IS of most nutrients than other farmers. To explore this, a fuzzy-k-means cluster analysis was performed using farm-specific average grain yields in 0-N, 0-P, and 0-K plots as variables. Four fuzzy classes provided the best classification result. Class means were then calculated for a range of soil and other characteristics (Table 6). Between-classes variation accounted for 60 to 62% of the total variation in grain yields in 0-N, 0-P, and 0-K plots and 44% of the variation in farmers' yield. However, variation in soil fertility properties (4–24%) or fertilizer use by farmers (2–17%) was less explained by this classification, indicating lack of consistent relationships with the yields measured in the omission plots (Table 6). For example, the crop-based classification accounted for only 4% of the variation in extractable soil P and K, demonstrating weak relationships between these soil tests and the yield classes established.


View this table:
[in this window]
[in a new window]
  		Table 6. Characteristics of classes of farms established by fuzzy-k-means clustering of grain yield measured in N (0-N), P (0-P), and K (0-K) omission plots. Values shown are class means. Classes 1 to 4 represent increasingly higher yields and/or soil fertility.

 
The first class represents 45 farms with low yields in omission plots of all three nutrients. It mainly includes direct-seeded rice farms at the OM site but also farms at SU in Indonesia and several other sites. Typically, soil pH was in the acid range, and extractable P was lower than at other sites, whereas extractable soil K and plant K uptake varied from low to medium levels. Farmers' use of P and K fertilizer was at the lower end, and average yield was only 4.4 Mg ha-1. The second (27 farms) and third (35 farms) classes represent medium levels of INS, IPS, and IKS. Class 3 mainly included farms with high clay content and less soil organic matter at AD and MA sites. However, soil pH, grain yields in FFP, and fertilizer N were higher in Class 3 than in Class 2. The fourth class included the 48 farms in which grain yields in 0-N, 0-P, and 0-K were all at medium to high levels relative to the whole sample. Most farms at JI in China belonged to this class, which reflects both traditionally high input use and the adoption of hybrid rice with a high yield potential. However, with the exception of OM, membership in Class 4 was well distributed among all sites. This class may represent the farms with the best soil but also those with the best crop management at each site. Measured average values of INS, IPS, and IKS in this class were significantly higher then in all other classes, and soils also had slightly higher amounts of extractable nutrients and organic matter.

The fuzzy-k-means cluster analysis (Table 6) was a first exploratory attempt to group farms according to differences in IS that may be related to the quality of crop and soil management over time. Management expresses itself in higher crop yields but also factors such as more balanced fertilizer use or high adoption of rice hybrids (JI). Long-term on-farm monitoring will be required to establish true cause–effect relationships and how they may affect soil fertility, crop productivity, and the overall sustainability of the system.


   DISCUSSION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
CONCLUSIONS
 REFERENCES
 
For more slowly changing soil properties, the means of our sample (Table 2) were nearly identical to those reported for a sample of 410 paddy soils surveyed in tropical Asia from 1963 to the early 1970s (Kawaguchi and Kyuma, 1977). These authors concluded that the nature of parent materials determines the base status and inherent soil fertility potential of paddy soils in Asia, whereas soil properties associated with organic matter or available P appeared to be less related to it. More recent surveys conducted in the Philippines, Thailand, and Cambodia indicated that spatial variation in soil fertility properties, particularly available soil nutrients, was little related to more stable soil and landscape features in rainfed and irrigated rice areas (Oberthür et al., 1996; Dobermann and Oberthür, 1997; Oberthür et al., 2000). This limits the use of conventional soil classification systems or soil maps for fine-tuning fertilizer recommendations in rice-growing areas. Although some improvements are possible through developing agronomically oriented soil classification systems (Lin, 1985; White et al., 2000), our farm classification results (Table 6) indicate little association between crop-measured IS and many soil properties measured. This raises questions about the general validity of soil measurements for fine-tuning fertilizer recommendations to the future needs in these environments. Crop-based indices of IS are therefore likely to be a key component of SSNM at high yield levels of rice.

Average yields in nutrient omission plots (Table 3) confirm that N deficiency is a general feature of irrigated rice systems in Asia, whereas, on average, P and K supply are less limiting factors at present yield levels. However, soils, climate, and crop management cause variability of IS among irrigated rice domains, years, climatic seasons, and fields within a domain. Most of the variability in IS occurred among farms or due to interactions of year, cropping season, and farms. Implications for sampling are discussed elsewhere (Dobermann et al., 2003). Strong correlations between grain yield measurements in different treatments (Fig. 4) reflect inherent variation in IS and climate but probably also differences among farms that are due to crop management.

Initial or more gradual long-term declines in INS have been observed in some long-term experiments with irrigated rice (Cassman et al., 1995). However, our on-farm data and those from other experiments (Ladha et al., 2000) suggest that there is little change in INS for periods of at least 4 to 6 yr in continuous irrigated rice systems that have been intensively cultivated for decades. For time periods that are relevant for fertilizer management decisions, such systems appear to have reached a near steady-state situation in terms of N inputs and outputs. Nitrogen inputs from rainfall and irrigation do not represent a significant net gain of N in most irrigated rice systems (Cassman et al., 1998). Likewise, although 15N balance studies suggest that between 10 and 50% of applied fertilizer N can be recovered from the soil (Vlek and Byrnes, 1986; De Datta et al., 1988), most of this N is incorporated into the organic pool by final harvest, and pool substitution occurs (Craswell et al., 1985; Schnier, 1994). Little mineral N is left in the soil after harvest of rice (Dobermann et al., 1994), and continuous fertilizer N application to rice rarely results in a significant net N gain that could lead to an increase in INS over relatively short time periods. A quasi steady-state INS level for a period of several years mainly represents the consistent (slow) N supply from soil organic matter and biological N2 fixation. A measured average INS value is therefore likely to be valid for periods of at least 5 yr, unless major disturbances such as a shift in crop rotation and residue management (Witt et al., 2000) or changes in length of fallow periods (Dobermann et al., 2000) occur. Similarly, changes in IPS and IKS are likely to be small over periods of few years and well within the measurement error range. Further research is required to develop models for predicting changes in IPS and IKS as a function of nutrient input–output balances (Dobermann et al., 1998) to then re-evaluate fertilizer management strategies that are based on a combination of crop nutrient requirements and input–output balances (Witt et al., 2002) and to identify optimal time intervals for resampling within a SSNM framework.

The large seasonal and spatial variability in INS (Fig. 2) confirms earlier observations made in long-term experiments or at a limited number of on-farm research sites (Cassman et al., 1996a, 1996b). At some sites such as OM, seasonal differences between HYS and LYS crops were rather consistent (Fig. 2), whereas at others such as MA, differences between HYS and LYS crops were less predictable. Variability in INS is an ubiquitous feature of irrigated rice environments in tropical and subtropical Asia and probably one of the major reasons for the large seasonal fluctuations in optimal fertilizer N rates observed (Dawe and Moya, 1999). Improved location-specific guidelines for N use can be developed from knowledge of the average INS, yield potential (climatic season), genotype, crop establishment method, and water management (Dobermann and Fairhurst, 2000). However, due to the tight relationship between crop growth and N uptake, more significant gains in yield and N use efficiency appear to require field-specific, real-time N management to improve the congruence between N supply and crop N demand (Peng et al., 1996).

Approximate knowledge of the average near steady-state INS level combined with diagnosis of plant N status using simple tools such as a leaf color chart may offer the greatest opportunities to manage N (Witt et al., 2002; Balasubramanian et al., 1999). Although variability of IPS and IKS is large too, both nutrients are less limiting at present yield levels and less affected by within-season dynamics than INS. They can also be manipulated easier through management of the overall P or K input–output balance because losses of P and K from lowland rice soils tend to be small (Dobermann et al., 1998). Slight errors in estimating IPS or IKS and in the overall P or K management are unlikely to significantly affect yields and profitability at most sites. Measuring IPS and IKS in HYS crops and utilizing such crop-based estimates for fertilizer recommendations in both HYS and LYS is probably sufficiently accurate, but more research needs to be conducted to confirm this. One exception, for example, would be sites with high amounts of K-rich irrigation water applied to dry-season crops, as opposed to wet-season crops grown on the same site because the latter receive little supplementary irrigation. In general, however, both P and K can be managed with simpler means than N, probably by estimating average IPS and IKS for larger areas as the basis for domain-specific fertilizer recommendations rather than by using more elaborate field-specific approaches.


   CONCLUSIONS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
REFERENCES
 
Whereas N deficiency is a general feature of irrigated rice systems in south and southeast Asia, average IPS and IKS are less limiting at present average yields of 5.2 to 5.3 Mg ha-1. However, large geographical differences occur, and unless fertilizer rates are increased, current levels of IPS and IKS may be insufficient for supporting the forecasted future yield increases needed to sustain rice supply, particularly at sites such as OM, MA, SU, and TH. Long-term on-farm monitoring needs to be improved to quantify historical trends in soil fertility, particularly more subtle changes associated with intensive rice cropping.

Spatial and temporal variation in soil fertility and crop-based estimates of IS among and within irrigated rice domains was significant. In addition to soil and climatic differences among geographical regions, the large variability in IS among domains and within them is probably more recent and reflects differences in cropping intensity, crop establishment methods, residue management, input use, and yields. Although spatial variability within rice fields accounted for about 20% of the total IS variance in the rice domains sampled, managing this component appears difficult.

Nutrient management for rice should focus on developing fertilizer recommendations for spatial domains with relatively uniform agroecological characteristics, cropping practices, and socioeconomic conditions. Domain-specific IS estimates and fertilizer recommendations are probably sufficient for managing less dynamic nutrients such as P and K. In addition, where feasible, field-specific N management using crop diagnostics should be used to manage the seasonal and spatial variability in INS.


   ACKNOWLEDGMENTS
 
We acknowledge contributions made by many researchers and support staff participating in the Project on Reversing Trends of Declining Productivity in Intensive Irrigated Rice Systems. We particularly wish to thank G.O. Redondo, S. Serrano, A.P. Estigoy, M. Elliot, C. Espana, R. Dawang, E.M. Punzalan, R.T. Cruz, and J. Bajita (Maligaya); Z. Susanti, Pahim, A. Djatiharti, D. Subarja, Atim, T. Rustiati, Y. Ariyani, U. Sutaryo Suhana, and I. Juliardi (Sukamandi); S. Selvam, R. Sakunthala, V.Gnanabharathi, G. Sasikumar, S. Sridevi, S. Sujatha, and S. Antony Samy (Thanjavur); M. Sivanantham, D. Kabilar, M. Selvakumar, S. Natarajan, R. Jayaseelan, S. Arumugam, and S. Selvaganabathy (Aduthurai); He Yunfeng, Huang Xueping, Wu Jiangxiang, and Ding Xianghai (Jinhua); and Tran Quang Tuyen, Tran Thi Ngoc Huan, Trinh Quang Khuong, Nguyen Thanh Hoai, Le Ngoc Diep, Ho Tri Dung, and Nguyen Xuan Lai (Omon). We thank Kenneth G. Cassman and Daniel C. Olk for leading the project from 1994 to 1996 and Violeta Bartolome and Graham McLaren (IRRI) for help with the statistical analysis. The International Rice Research Institute (IRRI), the Swiss Agency for Development and Cooperation (SDC), the International Fertilizer Industry Association (IFA), the Potash and Phosphate Institute/Potash and Phosphate Institute Canada (PPI/PPIC), and the International Potash Institute (IPI) provided funding for this research.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
{dagger} A. Dobermann, current address: Dep. of Agron. and Hortic., Univ. of Nebraska, P.O. Box 830915, Lincoln, NE 68583-0915. Back


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 




This article has been cited by other articles:


Home page
 Agron. J.Home page
Z. Segda, S. M. Haefele, M. C. S. Wopereis, M. P. Sedogo, and S. Guinko
Combining Field and Simulation Studies to Improve Fertilizer Recommendations for Irrigated Rice in Burkina Faso
Agron. J., September 19, 2005; 97(5): 1429 - 1437.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
M. M. Alam, J. K. Ladha, S. R. Khan, Foyjunnessa, Harun-ur-Rashid, A. H. Khan, and R. J. Buresh
Leaf Color Chart for Managing Nitrogen Fertilizer in Lowland Rice in Bangladesh
Agron. J., May 13, 2005; 97(3): 949 - 959.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
A. K. Shukla, J. K. Ladha, V. K. Singh, B. S. Dwivedi, V. Balasubramanian, R. K. Gupta, S. K. Sharma, Y. Singh, H. Pathak, P. S. Pandey, et al.
Calibrating the Leaf Color Chart for Nitrogen Management in Different Genotypes of Rice and Wheat in a Systems Perspective
Agron. J., November 1, 2004; 96(6): 1606 - 1621.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science