SSSAJ Grow Your Career with SSSA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 6 May 2005
Published in Soil Sci Soc Am J 69:842-855 (2005)
DOI: 10.2136/sssaj2004.0150
© 2005 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
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 Google Scholar
Google Scholar
Right arrow Articles by Fan, T.
Right arrow Articles by Gao, Y.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Fan, T.
Right arrow Articles by Gao, Y.
Agricola
Right arrow Articles by Fan, T.
Right arrow Articles by Gao, Y.
Related Collections
Right arrow Soil Fertility and Productivity
Right arrow Other Soil Management
Right arrow Plant Nutrition
Right arrow Root Growth

Soil Fertility & Plant Nutrition

Long-Term Fertilizer and Water Availability Effects on Cereal Yield and Soil Chemical Properties in Northwest China

Tinglu Fana,*, B. A. Stewartb, William A. Paynec, Wang Yonga, Junjie Luoa and Yufeng Gaoa

a Dryland Agricultural Institute, Gansu Academy of Agricultural Sciences, Lanzhou 730070, Gansu, P.R. China
b Dryland Agriculture Institute, West Texas A&M Univ., Canyon, TX 79016
c Texas Agriculture Experimental Station, Texas A&M Univ., Bushland, TX 79012

* Corresponding author (Fantl{at}hotmail.com)


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Wheat (Triticum aestivum L.) and corn (Zea mays L.) rotation is important for the region's food security in Northwest China. Grain yield and water-use efficiency (WUE: grain yield/estimated evapotranspiration [ET]) trends, and changes in soil properties during a 24-yr rainfed fertilization experiment in Pingliang, Gansu, China, were recorded. Mean wheat yields for the 16 yr ranged from 1.29 Mg ha–1 for the unfertilized plots (CK) to 4.71 Mg ha–1 for the plots that received manure (M) annually with inorganic N and P fertilizers (MNP). Corn yields for the 6 yr averaged 2.29 and 5.61 Mg ha–1 in the same treatments. Yields and WUEs declined with years except for the CK and MNP treatments for wheat. Wheat yields for the N and M treatments declined about 80 kg ha–1 yr–1, compared with about 60 kg ha–1 yr–1 for the NP treatment and the N plus straw treatment receiving P every second year (SNP). Likewise, the corn yields and WUEs declined significantly for all treatments. Grain yield-ET relationships were linear with slopes ranging from 0.51 to 1.27 kg ha–1 m–3 for wheat and 1.15 to 2.03 kg ha–1m–3 for corn. Soil organic C (SOC), total N (TN), and total P (TP) gradually increased with time except the CK, in which TN and TP remained unchanged but SOC and available P (AP) decreased. Soil AP decreased in the N treatment. Soil available K declined rapidly without straw or manure additions. The greatest SOC increases of about 160 mg kg–1 yr–1 occurred in SNP and MNP treated soils, suggesting that long-term additions of organic materials could increase water-holding capacity that, in return, improves water availability to plants and arrests grain yield declines, and sustains productivity.

Abbreviations: AK, available potassium • AP, available phosphorus • CK, unfertilized plots • CWSI, crop water stress index • DI, drought index • ET, estimated evapotranspiration • FE, fallow efficiency • GSP, growing season precipitation • M, farmyard manure • MNP, manure annually with inorganic nitrogen and phosphorus • PET, potential evapotranspiration • SNP, straw and nitrogen annually and phosphorus every second year • SOC, soil organic C • SOM, soil organic matter • TN, total nitrogen • TP, total phosphorus • WUE, water use efficiency


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
NORTHWEST CHINA is a vast, semiarid area with average annual precipitation ranging from 300 to 600 mm. More than 90% of the cropland in this area receives no irrigation. The main crops are wheat and corn, which are periodically rotated. There are about 1.3 million ha of wheat and corn rotations in the Loess Plateau region of Northwest China. This cropping system produces about 40% of the local food needs (Fan and Song, 2002), and has emerged as the most important one for food security in this dryland region (Xing et al., 2001). Generally, three or more years of continuous wheat are followed by two or more years of continuous corn. In a typical system where winter wheat follows corn, wheat is seeded immediately following corn harvest. In all other combinations, land is fallowed between crops to store water in the soil for the subsequent crop. The fallow periods between crops range from null for wheat following corn to 9 mo for corn following wheat, and from about 3 mo for wheat following wheat to 6 mo for corn following corn. Shangguan et al. (2002) reported that fallow efficiency (FE), expressed as soil water accumulation divided by precipitation received during fallow periods, for the area was about 35 to 40%. The importance of storing soil water during fallow periods for increasing grain yields of subsequent crops has been supported by many dryland studies including those in the U.S. Southern Great Plains by Johnson (1964), Musick et al. (1994), and in the China Loess Plateau by Shangguan et al. (2002).

Of all farming practices, rational fertilization and management of soil fertility are among the most important measures to improve grain yield and WUE (grain yield per unit of seasonal ET in kg ha–1 m–3) toward a sustainable crop production that will be required to meet the food demand of the region's growing population. The importance of soil fertility to optimizing WUE has long been recognized (Power et al., 1961; Viets, 1962; Stewart, 1989; Cai et al., 2002). Therefore, maintenance of soil fertility will be essential to improve and sustain grain yields. Soil organic matter (SOM) will be critical to this goal because it directly and indirectly affects various chemical, physical, and biological soil properties that are related to plant behavior. Soil organic matter is recognized as a cornerstone for successful farming in most areas (Vanlauwe et al., 2001; Merckx et al., 2001; Zhang and He, 2004). Challenges for dryland farming in Northwest China are low WUE resulting from low SOM and low soil fertility (Zhang et al., 1997; Zhu, 1984). Organic C in the surface layer of this region is generally <5.8 g kg–1 (Xing et al., 2001) because low quality manure is applied at rate of 30 Mg ha–1 and crop residue is usually taken out from field for feed or fuel. Farmyard manure (not pure manure but a mixtures with soil) and inorganic fertilizers are widely applied as a crucial approach for both improved soil properties and efficient water use for crop production (He and Lin, 1992).

Long-term experiments are invaluable for assessment of cropping system effects on soil properties, grain yield and WUE, and risk management (Regmi et al., 2002; Dawe et al., 2000; Camara et al., 2003). Many long-term experiments have been used to test effects of fertilization on grain yield and soil properties (Jenkinson, 1991; Mitchell et al., 1991; Sandor and Eash, 1991; Brown, 1991; Bhandari et al., 2002; Zhu, 1997; Wang et al., 2002), but few continuous long-term studies are available from Northwest China. Therefore, preliminary fertilizer recommendations, which need further calibration through multi-year field experiments, were provided based on studies within a short period of time. The long-term experiment reported here began in 1979, and is the longest running annual cropping system experiment in the Loess Plateau. The study was based on assumptions on crop water use to aim to (i) examine grain yield and WUE trends for wheat and corn in annual cropping systems under long-term organic and inorganic fertilization; and (ii) monitor long-term effects of fertilization on soil properties.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Experimental Site
A long-term permanent plot experiment has been conducted since April 1979 at the Gaoping Agronomy Farm, Pingliang, Gansu, China. The site is in the central part of the Shizi highland plateau (35°16' N lat., 107°30' E long, 1254 m altitude). Its dark loess soils are classified as Calcarid Regosols (FAO/UNESCO, 1988) with an average SOM of 9.1 g kg–1, corresponding to about 5.3 g kg–1 of SOC. Soil texture of the 0- to 20-cm depth interval is silt loam (sand 231 g kg–1, silt 433 g kg–1, and clay 336 g kg–1) with a bulk density of 1.30 Mg m–3. Analysis of soil samples taken from the experimental area in October 1978 indicated that the top 15 cm of soil had a pH of 8.2, SOC content of 6.2 g kg–1; TN of 0.95 g kg–1 (Black, 1965); TP content of 5.7 g kg–1 (Murphy and Riley, 1962); AP of 7.2 mg kg–1 (Bray and Kurtz, 1945); and available K (AK) of 165 mg kg–1 (Shi, 1976). Soil organic matter was analyzed by the Walkley–Black (WB) procedure (Allison, 1965; Walkley and Black, 1934) and the value divided by 1.724 to estimate SOC in g kg–1 that was converted to Mg ha–1 of SOC for given depth assuming a bulk density of 1.30 Mg m–3. Groundwater level remained at a depth of about 80 m below soil surface for the duration of the study.

Under average climatic conditions, the area has an aridity index (P/PET: precipitation/potential evapotranspiration) of 0.39 and receives 540 mm of precipitation, about 60% of which occurs from July through September. May through June is the driest period for crop growth and light precipitation is common during December and January. The mean monthly maximum and minimum temperatures for the wheat-growing period (October–June) are 21.1 and –12.9°C, and 24.2 and 8.1°C for the corn-growing season (April–September). The study area is representative of a typical farming region completely dependent on precipitation.

Experimental Design and Treatments
The experiment began in 1979 with a corn crop on land that had been cropped to corn the previous year. There was one crop each year. Six fertilizer treatments arranged in a randomized complete block design with three replications. Corn was grown in 1979 and 1980, wheat from 1981 through 1984, corn in 1985 and 1986, wheat from 1987 through 1990, corn in 1991 and 1992, wheat from 1993 through 1998, soybean [Glycine max (L.) Merr.] in 1999, sorghum [Sorghum bicolor (L.) Moench] in 2000, and wheat in 2001 and 2002. Data for the 16 yr of wheat and 6 yr of corn are presented in this paper (Table 1).


View this table:
[in this window]
[in a new window]
  		Table 1. Growing season precipitation (GSP), potential evapotranspiration (PET), estimated seasonal evapotranspiration (ET) and crop water stress index (CWSI = 1 – ET/PET), drought index (DI), and year's type classified by DI values in a long-term (1979–2002) rainfed fertilization experiment in Pingliang, Gansu, China.

 
The experimental layout was located in a 0.44-ha area. Each plot was 16.7 by 13.3 m with a buffer zone of 1.0 m between each plot. The six treatments were (i) CK, unfertilized; (ii) N, nitrogen fertilizer annually; (iii) NP, N and P fertilizers annually; (iv) SNP, straw (S) plus N added annually and P fertilizer added every second year; (v) M, farmyard manure added annually; and (vi) MNP, farmyard manure plus N and P fertilizers added annually. Urea and superphosphate were the N and P sources, and were broadcast applied at rates of 90 kg N ha–1 and of 30 kg P ha–1, respectively. Manure was added at rate of 75 Mg ha–1 (wet weight). Deep plowing of approximately 23 cm was performed in July after wheat harvest or in October after corn harvest except for the years in which wheat followed corn. In those years, shallow disk tillage was done after corn harvest and wheat was seeded immediately.

Generally, the farmyard manure was a mixture of about 1:5 ratio of manure to loess soils so its nutrient content was quite variable from year to year. The N, P, and K contents of the manure mixture taken in 1979 were 1.7, 6.8, and 28 g kg–1 in dry weight, indicating that manure is very low in N, and high in P and K. Although the specific amounts of nutrients added with manure each year were not determined, an application of approximately 75 Mg ha–1 (wet weight) supplied roughly 40 kg N ha–1, 200 kg P ha–1, and 840 kg K ha–1 in manure annually to crops. For SNP treatment, 3.75 Mg ha–1 of wheat straw approximately 10 cm in length was returned to the soil before plowing, and P fertilizer was added with the straw every second year. There was very little wheat straw or corn residue on the other treatments because all crops were harvested at the ground level and removed from the plots before thrashing the grain. The SNP treatment was the only one that had residue returned to the plots. The C content of the straw was 42.9% so there was approximately 1.6 Mg C ha–1 added each year to the SNP treatment.

Winter wheat (cultivars Qingxuan 8271, Longyuan 935, and Ping 93-2) was seeded in rows 14.7 cm apart at rates of 165 kg ha–1 on about 20 September each year when wheat followed wheat, and in early October when wheat followed corn. Corn hybrid Zhongdan 2 was seeded about 20 April each year that corn was grown by hand in hills every 33 cm in rows 66.5 cm apart. About 3 wk after seeding, corn plants were thinned to one plant per hill. Later, if tillers developed, they were removed to avoid competition with the main stem. Hand weeding was done to manage the weeds and plant protection measures were applied when needed. Crops were harvested manually close to the ground using sickles and all harvested biomass was removed from the plots. Grain yields were determined by harvesting 20 and 40 m2 areas at the center of each plot for wheat and corn, respectively. Grain samples were air-dried on concrete, threshed, and oven-dried at 70°C to a uniform moisture level, and then weighed.

Soil Sampling and Analysis
Eighteen soil samples of three replicates of six treatments were collected annually during 1979 through 1991 and 1996 through 1998 at 15 d after harvest. Each sample was a composite of three random 2-cm diam. cores per plot. A 5-cm i.d. auger was used to sample the 0- to 15-cm soil depth to determine the effect of fertilizations on soil nutrient contents by the methods listed earlier. The entire volume of soil was weighed and mixed thoroughly and a subsample was taken to determine dry weight. The fresh soil was mixed thoroughly, air-dried for 7 d, sieved through a 2.0-mm sieve at field moisture content, mixed, and stored in sealed plastic jars for analysis. Representative subsamples were drawn to determine TN, TP, AP, AK, and SOM by the methods listed earlier. Total P, AP, and AK were not determined in the samples taken in 1996.

Seasonal Evapotranspiration and Crop Water Stress Index
Monthly precipitation was measured using a rain gauge at a weather station close to the experimental farm, and PET was derived simultaneously from 255-200 Class ‘A’ Evaporation Pan located on the station (FAO, 1998, p. 78–84.). Generally, seasonal ET values are calculated by summing seasonal soil water depletion amounts with the growing season precipitation (GSP) amounts. For this study, soil water amounts at seeding and harvest were not measured so ET values could not be determined for each plot. However ET amounts were estimated by assuming FE values based on reported studies and authors experience (USDA, 1974; Dimes et al., 1996; Shangguan et al., 2002; Xing et al., 2001; Nielsen and Anderson, 1993). Therefore, estimated seasonal ET amounts were calculated by: ET = (FE x fallow season precipitation) + GSP. Growing season precipitation for wheat was October through June, and April through September for corn. Surface runoff was not considered because individual plots were surrounded with border dikes. Drainage was assumed negligible because of semiarid conditions. In this assumption, mean FE values used were 35% for the 3-mo fallow period when wheat followed wheat, 30% for the 6-mo fallow period when corn followed corn, and 25% for the 9-mo period when corn followed wheat. A null FE value was assumed when wheat followed corn because there was no fallow period and the corn generally had used most or all of the plant available soil water. Although these FE values are somewhat arbitrary, we think they provide reasonable estimates of water used from stored soil water and allow estimates to be made of seasonal ET values. For the 24-yr study, there were 16 wheat crops and 6 corn crops. Three of the wheat crops followed corn and one followed sorghum (calculations for sorghum were same as for corn), and 12 wheat crops followed wheat. Of the six corn crops, three followed corn and three followed wheat.

The methodology used resulted in all fertilizer treatments having the same amounts of seasonal ET. Although in reality there were probably some differences, the amounts are believed relatively small because water was generally limiting so all treatments extracted most or all of the plant available water from the soil profile. Viets (1962), in his classic review on fertilizers and the efficient use of water, stated that under arid and semiarid conditions fertilizers often had no effect on ET. He concluded whether fertilizers increase consumptive use not at all or only slightly, all evidence indicates that WUE can be greatly enhanced if fertilizers increase grain yield. Huang et al. (2003) showed little or no differences in seasonal ET values for various fertilizer treatments during a 15-yr study in Shanxi Province relatively with similar soil and climatic conditions to those in our study.

The crop water stress index (CWSI) is defined as being equal to 1 – (ET/PET) (Jackson et al., 1988; Olufayo et al., 1996) and was used in this study to assess the relationship between it and crop grain yield. Mean seasonal CWSI values for wheat and corn each year were calculated using estimated seasonal ET and derived PET from the weather station. A CWSI value of 1 would indicate full water stress and a value of 0 would indicate no stress. In this study, WUE was expressed as grain yield divided by estimated ET. At the same time, assuming that the 24 yr of record closely represent the climate of the region, probability values, expressed by the relative frequency from the 24 yr (1979–2002), of 75, 50, and 25% for CWSI and ET in the above four annual cropping systems, were calculated based on monthly PET records and estimated ETs, respectively. Therefore, grain yields corresponding to these probabilities can be predicted by using the functions of ET and CWSI related to grain yield as presented in this study.

Data Analyses
Analysis of variance for the randomized complete block design was done to determine main and interactive effects of treatment and year on wheat and corn grain yield and WUE using PROC GLM (SAS Institute, 1991), and the fertilization treatment by year mean square was used as the error term to test for treatment and year effects in the 16-yr wheat and the 6-yr corn. There were significant interactions between treatments and years so means separation tests for 16-yr wheat and 6-yr corn were not conducted. One-way ANOVA was therefore made for individual years, and mean separation tests among fertilizer treatments were conducted using the least significant difference (LSD) procedure at the 0.05 probability level only when F was significant. Linear regression analyses were done using each plot data to determine trends (slopes) of grain yield and WUE and using composite soil data from three replicate plots to assess trends of various soil parameters over the years. Linear regression analyses were also used to identify the impact of seasonal ET, and CWSI on both wheat and corn grain yield. The P values (Pr > t) of the slopes were used to test whether the observed changes were significantly different from 1.

To analyze further treatment effects on grain yield in response to years with different rainfall, drought index (DI), defined as /{sigma}, where CWSIi and are crop water stress indexes for individual and average year for wheat and corn, and {sigma} is standard deviation for CWSI, was calculated. Xing et al. (2001) used DI to distinguish among wet (–2 < DI < –0.5), normal (–0.5 ≤ DI ≤ 0.5), and dry (0.5 < DI < 2) years.


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Crop Water Stress Index
There were large differences in the amounts and distribution patterns of GSP, fallow period intervals, and PET and CWSI values for the various annual cropping systems (Table 1). For the 16 yr of wheat, the average GSP, estimated seasonal ET, PET, and CWSI values were 289 mm, 384 mm, 928 mm, and 0.57, respectively. The CV values were 24.9, 24.5, 12.0, and 24.6% in the same respective order. The highest CWSI value was 0.83 in 1981 when the seasonal ET was only 191 mm and the PET was 1096 mm. This was the year with the lowest amount of seasonal precipitation during the wheat growing season, and also a year when wheat was seeded immediately following corn. The lowest CWSI was 0.31 when wheat followed wheat in the wet year of 1983. During that year, seasonal ET was 553 mm and the PET was 853 mm, respectively. The calculated DI values for the 16 yr of wheat ranged from –1.85 to 1.83 (Table 1). Four years, 1981, 1987, 1995, and 2001, had DI values ranging from 0.74 to 1.83, and were classified as dry years, with an average CWSI of 0.75. Four years, 1983, 1984, 1989, and 1990, were wet with DI values ranging from –0.92 to –1.85 with an average CWSI of 0.39. The remaining 8 yr of wheat, 1982, 1988, 1993, 1994, 1996, 1997, 1998, and 2002, were classified as normal years with DI values ranging from –0.40 to 0.50 and an average CWSI of 0.57.

For the 6 yr of corn, the GSP, ET, PET, and CWSI values were 400 mm, 461 mm, 1015 mm, and 0.54, respectively. The CVs were 23.0, 20.6, 6.1, and 18.8% in the same respective order. The highest and lowest CWSI values for corn, 0.70 and 0.43, occurred in 1986 and 1980 (and 1985), respectively, when corn followed corn. The estimated ET for 1986, the dry year, was 333 mm compared with 569 mm for 1980, the wet year (Table 1). The difference in PET values, however, differed by only 65 to 935 mm compared with 1000 mm. The DI values for the 6 yr of corn ranged from –1.09 in 1980 to 1.57 in 1986. Two years, 1986 and 1992, were classified as dry. Two years, 1979 and 1991, were classified as normal and the remaining 2 yr, 1980 and 1985, were classified as wet. The mean CWSI values were 0.65, 0.55, and 0.43 for the dry, normal and wet years, respectively. The DI values were 0.60 and 1.57 for the dry years, –0.00 and 0.21 for the normal years, and –1.09 and –1.08 for the wet years.

From data in Table 1, it is noteworthy that if an exceptional dry year of 1981 was not included, the estimated ET for 15 yr of wheat had declined with amounts of 6.8 mm per year at the 0.04 probability level. For 6 yr of corn, the estimated ET declined at amounts of 10.8 mm per year at the 0.1 probability level. This indicated that water stress gradually increased for both wheat and corn crops during the respective growing season.

Grain Yield
Statistical analyses of the 16 yr of wheat grain yields and 6 yr of corn grain yields showed that fertilization treatments impacted grain yields significantly, but grain yields were still highly influenced by precipitation and its interaction with fertilization, which represented by high significant effect of years in the ANOVA (Table 2). For each year, there were significant effects due to treatments using LSD (data not shown) for both wheat and corn. Grain yields in fertilized plots were generally higher than those in the unfertilized control (CK) plots. Grain yields for wheat and corn were consistently highest in the MNP treatment and lowest in the CK treatment, and that was always lower in the N treatment than in the M treatment and all others that received P fertilizer along with N fertilizer (Fig. 1).


View this table:
[in this window]
[in a new window]
  		Table 2. Grain yield and water use efficiency (WUE) changes, significance of yield and WUE change (P > t) over years, linear regression coefficient (R2), and analysis of variation (ANOVA) in a long-term (1979–2002) rainfed fertilization experiment in Pingliang, Gansu, China.

 


View larger version (79K):
[in this window]
[in a new window]
  		Fig. 1. Changes of grain yield and water use efficiency (WUE) in a long-term (1979–2002) rainfed fertilization experiment in Pingliang, Gansu, China. CWSI is crop water stress index (1 – ET/PET), where ET is estimated seasonal evapotranspiration for wheat and corn, and PET is potential evapotranspiration.

 
The average grain yields of wheat for the 16 growing seasons were 1.29, 2.36, 3.54, 4.15, 3.87, and 4.71 Mg ha–1 for the CK, N, M, SNP, NP, and MNP treatments, respectively (Table 2). During the study period, the wheat yield declined with time except CK and MNP treatments (Table 2), indicating grain yield when manure was added with the chemical fertilizers could arrest the grain yield decline, and low yield without any nutrients added fluctuated highly at coefficient of variation (CV) of 40% among years. In the NP and SNP treatments, the grain yield declines of 57 and 61 kg ha–1 yr–1 were significant at the 0.05 probability level, but mean yield for the SNP treatment was 7% higher than the NP treatment. In contrast, the N and M treatments showed grain yield reductions of 77 and 81 kg ha–1 yr–1 that were significant at 0.01 probability level and accounted for 16 to 20% of the yield variability. Although these differences in grain yield reduction were similar, the N only treatment had a CV of 55% that was the highest among all treatments. The M treatment CV was only 35% that compares closely to the CV of 28% found in the NPM treatment that showed the greatest stability of all treatments. These results indicated that depletion of other nutrients limited wheat yield. In addition, grain yield differences between treatments were also influenced by how dry or wet the growing season was (Fig. 1). For the 4 yr classified as dry by the DI, the average CWSI was 0.75. The CK had an average grain yield of only 0.60 Mg ha–1 compared with 1.75 Mg ha–1 for the 4 yr classified as wet having an average CWSI of 0.37. In comparison, an average yield for the N only treatment was 0.86 Mg ha–1 for the dry years and 4.22 Mg ha–1 for the wet years. The yields of the N treatment were still low in the normal and wet years when compared with treatments receiving fertilizer P in addition to N or when organic materials were added, but they were much greater than the CK treatment. The MNP treatment produced the greatest yields and averaged 3.1 Mg ha–1 for the dry years and 6.2 Mg ha–1 for the wet years. These results show the importance of adequate soil fertility even in the dry years because the average yield of wheat from the MNP was about five times greater than the yield of the CK for the same years (Fig. 1).

Like wheat grain yields, corn grain yields were also significantly influenced by treatments, and the mean yields for the 6 yr were 2.29, 3.02, 4.39, 4.75, 4.75, and 5.61 Mg ha–1 for the CK, N, M, SNP, NP, and MNP treatments, respectively. However, unlike wheat, the corn yield declines over that period were highly significant for all treatments, and the declined amounts ranged from 160 to 250 kg ha–1 yr–1 that were much higher than in wheat. These declines explained 40 to 66% of the yield variability that were also substantially greater than those for wheat (Table 2). The yield declines for the CK, N, and M treatments did not differ, but yield declines increased by 32% for the N treatment and 92% for the M treatment compared with the CK. The yield decline for NP was similar to that for MNP. These obvious yield declines for wheat and corn were likely due to both negative changes in soil characteristics and precipitation variation. As suggested earlier, the water stress gradually increased for both crops because the estimated ET was showing a downward trend that resulted in an upward trend of CWSI. The years of 1984 for wheat and 1985 for corn were wet, and the highest yields were on the MNP plots. Grain yields were 7.0 Mg ha–1 for wheat and 7.9 Mg ha–1 for corn. These comparable yields suggest that N likely became limiting for corn growth. The yield changes for dry, normal, and wet years were similar to those of wheat, but the year effect was greater for corn (Fig. 1).

More importantly, the average wheat yield for the 16 yr in the SNP treatment was 0.28 Mg ha–1 higher than that in the NP treatment. There was no difference, however, for corn yield between these two treatments, and the yield decrease in the SNP was only 160 kg ha–1 yr–1 that was smallest in all treatments. Both wheat and corn yields from the M treatment were consistently greater than the N treatment. These results clearly showed a positive impact of annual application of organic materials such as straw and manure on these dryland crops. However, it is not clear whether the impact was due to improved water relationships resulting from increased SOM or improved fertility, particularly K, as will be discussed later.

Water Use Efficiency
The linear regression lines between grain yield and estimated ET were statistically significant for both wheat (16 yr) and corn (6 yr) for all treatments (Table 3). The linear regression coefficients (slopes) ranged from 1.15 to 1.27 kg ha–1 m–3 for wheat and from 1.34 to 2.03 kg ha–1 m–3 for corn across treatments except for the CK, in which the slopes were 0.51 and 1.15 kg ha–1 m–3, respectively. The slopes for the CK were low presumably because plant nutrients were often more limiting than water for crop production. The slopes for the fertilized wheat plots in this study were similar to the value of 1.22 kg ha–1 m–3 reported by Musick et al. (1994) for wheat grown in the semiarid U.S. Southern Great Plains. The slope values for the fertilized corn plots were close to the values of 2.05 kg ha–1 m–3 reported by Musick and Dusek (1980) and 1.53 kg ha–1 m–3 by Tolk et al. (1998) in the U.S. High Plains.


View this table:
[in this window]
[in a new window]
  		Table 3. Fitted linear slopes and y-intercepts for the relationship between grain yield and estimated seasonal ET in a long-term (1979–2002) rainfed fertilizer experiment in Pingliang, Gansu, China.

 
Water-use efficiencies varied greatly between crop years and treatments and ranged from 0.12 kg ha–1 m–3 for unfertilized wheat in the dry year of 1995 to 1.42 kg ha–1 m–3 for the plot fertilized with NPM in the normal year of 1982 (Fig. 1). Similarly, values ranged from 0.31 kg ha–1 m–3 for unfertilized corn in the dry year of 1986 to 1.44 kg ha–1 m–3 for the NPM treatment in the wet year of 1985. Similar to the ANOVA results for grain yields, WUE values of both crops were significantly affected by treatments, years, and their interactions (Table 2). For each year, effects of the fertilized treatments were statistically significant as assessed by LSD (data not showed). Average WUE values for the 16 yr of wheat were 0.32, 0.57, 0.91, 0.99, 1.08, and 1.20 kg ha–1 m–3 for the CK, N, M, NP, SNP, and MNP treatments, respectively. For the 6 yr of corn, WUE values averaged 0.47, 0.63, 0.94, 1.0, 1.02, and 1.19 kg ha–1 m–3 for the respective treatments. In all years, the MNP had the highest WUE value, the CK had the lowest value, and the N only treatment value was lower than all treatments other than the control. Compared with yield data, CVs for WUE values were consistently low within for all treatments, particularly for the M, NP, SNP, and MNP treatments where the CVs (Table 2) were about half of those for yield, suggesting that WUE values were relatively stable from year to year. For wheat, when both M and NP were used simultaneously, it is particularly noteworthy that lowered grain yields during dry years did not concomitant to substantially lowered WUE values (Fig. 1). Moreover, the increased yields during normal and wet years did not find attributed to increased WUE values much compared with the dry years. Similar findings occurred for corn in the dry and wet years.

Similar to grain yield declines discussed earlier, wheat WUEs over years also declined with lapse of time except the CK and MNP treatments. For corn, WUE declined linearly and the change was significant in all treatments (Table 2). The relative declines in WUE were greater for corn than for wheat.

For the SNP and MNP, mean WUEs for four dry wheat years were 1.02 and 1.17 kg ha–1 m–3, and for two dry corn years were 0.95 and 1.01 kg ha–1 m–3, respectively. The WUE values for these two treatments were consistently higher than those for other fertilized treatments, indicating that the combination of NP and organic materials resulted in the most efficient use of water.

Crop Water Stress Index and Grain Yield
Grain yields were also strongly correlated with CWSI values for both wheat and corn (Fig. 2 and 3). The declining linear regression slopes ranged from 3.2 to 8.7 Mg ha–1 per unit increase in CWSI for wheat, and from 9.7 to 18.2 Mg ha–1 for corn across all treatments. During the 24-yr study, CWSI values ranged from 0.31 to 0.83. The downward slope magnitude was almost double for corn as compared with wheat, suggesting that corn is more sensitive and pronounced showing that the same amount of water stress will lower corn yields more than wheat yields.



View larger version (26K):
[in this window]
[in a new window]
  		Fig. 2. Relationship between wheat grain yield and crop water stress (CWSI) in a long-term rainfed fertilization experiment in Pingliang, Gansu, China. A CWSI value of 1 or 0 would indicate full stress or no stress. ***Linear regression coefficient (R2) significant at 0.001 probability level.

 


View larger version (26K):
[in this window]
[in a new window]
  		Fig. 3. Relationship between corn grain yield and crop water stress (CWSI) in a long-term rainfed fertilization experiment in Pingliang, Gansu, China. A CWSI value of 1 or 0 would indicate full stress or no stress. ***Linear regression coefficient (R2) significant at 0.001 probability level.

 
Grain yields of both wheat and corn increased as CWSI values decreased, but this differed with treatments. The greatest yield increases occurred with the MNP and NP treatments, followed closely by the SNP and M and N. The yield increase for the CK was constrained by lack of nutrients so increased available water seemed not efficiently utilized. Based on the data presented in Fig. 2 and 3, grain yields of both wheat and corn were <1 Mg ha–1 when CWSI value was >0.68 for the CK, and > 0.72 for the N. For similar CWSI values, wheat yields were >3 Mg ha–1 for the MNP and SNP, and corn yields were about 2.5 Mg ha–1. This clearly illustrates the importance of adequate soil fertility even when water is limiting. In the driest year in which wheat was grown, even when the CWSI values approached 0.83, yields were still about 2 Mg ha–1 for the fertilized treatments (MNP and SNP), but only 0.4, and 0.7 Mg ha–1 for the CK, and N treatments, respectively.

Grain Yield Predication Based on Probabilities of Evapotranspiration and Crop Water Stress Index
Dryland farming is greatly impacted by seasonal precipitation variations both in amount and distribution, making risk assessment a necessary and important farm decision tool. The ET and CWSI probabilities of producing various amounts of wheat and corn grain for different cropping sequences and different fertility treatments are shown in Fig. 4 and 5, respectively. The results suggest that, for example, when wheat follows wheat, there is a 50% probability that the CWSI during the growing season will be 0.59 or less and the estimated grain yield for a control area will be about 1.2 Mg ha–1 or more (Fig. 4). In contrast, for an area of high fertility like the NPM plot, the grain yield would be expected to be about 4.5 Mg ha–1 or more. For the same treatments, there is a 75% probability that the CWSI will be 0.52 or less and the yield of the control would be about 1.4 Mg ha–1 or more, but a high fertility area would likely yield about 5.1 Mg ha–1 or more. The lowest wheat yields would be expected when wheat follows corn because there is essentially no fallow period for storing soil water. In that case, there is a 50% probability of CWSI (0.70) that the grain yield of wheat will not be more than about 1 Mg ha–1, but even in this case, a high fertility area would be expected to yield about 3.6 Mg ha–1. This reinforces the need for adequate soil fertility management for reducing risk and maintaining production in dry years. Figure 5 shows similar trends when estimated seasonal ET values are considered rather than CWSI values.



View larger version (41K):
[in this window]
[in a new window]
  		Fig. 4. Estimated grain yield for different cropping systems based on probabilities of crop water stress index (CWSI) and grain yield–CWSI relationships. Probabilities, expressed by the relative frequency, of CWSI values are calculated based on 24 yr of estimated seasonal Evapotranspiration (ET) and potential evapotranspiration (PET) records. Regression functions in Fig. 2 and 3 were used to estimate grain yield related to probabilities of CWSI. Data points ({uparrow}) are for minimum, 75% probability, 50% probability, 25% probability, and maximum CWSI values in 24 yr of records, respectively.

 


View larger version (36K):
[in this window]
[in a new window]
  		Fig. 5. Estimated grain yield for different cropping systems based on probabilities of estimated seasonal evapotranspiration (ET) and regression function of grain yield to estimated ET. Probabilities, expressed by the relative frequency, of ET values are calculated based on 24 yr of crop seasonal precipitation plus rainfall amount estimated by assuming fallow efficiency in fallow periods of the various cropping system. Regression functions in Table 3 were used to estimate grain yield related to probabilities of ET. Data points ({uparrow}) are for minimum, 75% probability, 50% probability, 25% probability, and maximum ET values in 24 yr of records, respectively.

 
The CWSI values and seasonal ET amounts shown in Fig. 4 and 5 were calculated using the 24 yr of weather records and the fallow efficiency assumptions discussed in the Materials and Methods section. If actual plant available soil water amounts at seeding time are known, they should be used along with probabilities of seasonal precipitation to possibly improve the risk assessment.

These results clearly show that wheat following wheat, corn following corn, and corn following wheat are all well adapted systems for the Loess Plateau of China. The risk of low yields when wheat follows corn is considerably greater because of limited stored soil water, but even this sequence is fairly dependable when soil fertility and SOM are maintained at a high level.

Soil Chemical Properties
Soil Organic Carbon and Total Nitrogen
Levels of SOC and soil TN were greatly affected by the various treatments during the study period. The SOC at the beginning of the study in 1979 was 6.1 g C kg–1 in the 0- to 15-cm depth. From 1979 to 1998, SOC concentrations increased for all treatments except the CK, but the greatest increases occurred for the three plots of MNP, SNP and M that received organic materials (Fig. 6). Slopes values in Table 4 indicated that 165.0, 157.3, and 118.7 mg C kg–1 yr–1 were added each year for the MNP, SNP, and M treatments, respectively. The NP and N treatments also increased SOC levels but at much lower rates of 44.3 and 20.9 mg C kg–1 yr–1, respectively. Soil organic C decreased in the CK plots at an indicated rate of 18.3 mg kg–1 yr–1. If soil bulk density is 1.30 Mg m–3 in the 15-cm depth, the annual rates of C increase in the three treatments receiving annual organic matter additions compare favorably to values of 0.2 to 0.3 Mg C ha–1 yr–1 for fertilized maize reported by Lal (2000) for western Nigeria, and the 0.31 Mg C ha–1 yr–1 value reported by Jenkinson (1991) for the 157-yr study at Rothamsted. In the SNP treatment, this would suggest that about 15% of the total C added was converted into SOC each year. Based on a literature review in semiarid regions, Rasmussen and Collins (1991) concluded that SOC levels typically increase at a rate of 10 to 25% of the amount of added C.



View larger version (33K):
[in this window]
[in a new window]
  		Fig. 6. Trend changes of soil fertility in long-term fertilizers experiment in Pingliang, Gansu, China. Data showed here were continuous from 1979 through 1991 and 1996 through 1998; soil samples for 1992–1995 were not analyzed (Dashed lines), and total P, available P, and available K for 1996 were not determined except total N and soil organic matter.

 

View this table:
[in this window]
[in a new window]
  		Table 4. Slopes of the linear regression vs. time for each soil property, significance of soil properties change (P > t values) in a long-term rainfed fertilization (1979–2002) experiment in Pingliang, Gansu, China.

 
These clear findings shown that the combination of organic and inorganic fertilization enhanced the accumulation of SOC and maintained the highest productivity and WUE, consistent with many other studies in the world. It was deduced that SOC amounts increased at a rate of about 0.30 Mg ha–1 yr–1 when the manure or straw was added to the soil in combination with NP fertilization. At this rate, the SOM amounts of the surface 15 cm of soil can be increased about 0.4% in 25 yr.

Similarly, TN concentrations increased over the life of the study except the CK, in which it remained fairly stable. Slope magnitudes (Table 4) ranged from 3.6 mg kg–1 yr–1 for the N only treatment to 15.6 mg kg–1 yr–1 in the MNP treatment. As evidenced for SOC accumulations, N accumulation was higher when manure or straw was added along with NP fertilizers. This may have been partially due to a slow release of N from manure and straw, resulting in smaller losses of N as suggested by Bhandari et al. (2002). In addition, the MNP and SNP treatments produced higher amounts of crop biomass and therefore likely had more extensive root systems that may have contributed to increased N levels. The consistent SOC and TN trends illustrate the importance of long-term additions of organic materials to soil for maintaining SOC and sustaining land productivity.

Wheat yields for the NP, MNP, and SNP treatments for different years were analyzed in relation to estimated seasonal ET (mm) and SOM (Mg ha–1 to 15-cm depth based on soil bulk density of 1.30 Mg m–3) to gain some insights regarding the effect of SOM. All of the selected treatments received the same amount of NP fertilizer but the SOM was considerably higher in the MNP and SNP plots than in the NP plots (Fig. 6). Grain yield (Mg ha–1) was related to estimated ET and SOM as follows:

where Yw = grain yield of wheat, ET = mm seasonal evapotranspiration, and SOM = %SOC x 1.724. The ET coefficient 1.24 kg m–3 is similar to values found by Musick et al. (1994). The SOM contribution of 42.9 kg ha–1 is, however, larger than the 15.6 kg ha–1 that Bauer and Black (1994) found for Typic Agriborlls in the USA, but similar to the 40.7 kg ha–1 reported by Martin et al. (1999) in the semiarid Argentine Pampas. The increased grain yield with increasing SOM in this study is perhaps a combination of increased soil fertility and of increased soil physical properties that improved WUE. This would imply that a decline in SOM will result in decreased yields as a consequence of both loss of fertility and decreased soil water holding capacity that, in turn, reduce water availability to plants.

It was hypothesized that perhaps the grain yields and WUEs of the MNP treatments would increase with time as SOM increased, but this could not be detected from the yield and WUE values. Although the causes of yield decline are not separated from weather effect and SOM increase, gradual increase of water stress might be a major reason, and low input of N and declining soil fertility should be of much concern. As discussed earlier, about 130 kg N ha–1 that included only 90 kg N ha–1 with chemical fertilizer and the remaining 40 kg from manure was usually adequate for wheat but may have been deficient for corn, particularly in years receiving normal or higher precipitation. Currently recommended N levels are 145 kg N ha–1 for wheat and 180 kg N ha–1 for corn in the study region.

Total Phosphorus and Available Phosphorous
Soil TP increased significantly with lapse of time for all treatments except the CK, in which trend lines showed no change (Fig. 6 and Table 4). There were large differences, however, among the treatments. Great gains for TP occurred ranging from 7.5 to 11.6 mg kg–1 for the treatments receiving manure or combinations of NP and organic materials, but little increase of 1.9 mg kg–1 for the N treatment (Table 4). Slopes of the trend lines for soil AP (Table 4) indicated declines of about 0.20 mg kg–1 yr–1 for the CK and N only treatments, but increases of 0.19, 0.25, 0.38, and 0.67 for the M, NP, SNP, and MNP treatments, respectively. In 1998, the 20th year of the study, the TP in the NPM treatment was 114% of the TP in the M, and the AP was 189% of that in the same treatment. This shows that inputs of P with manure combined with inorganic fertilizer exceeded plant needs and resulted in a substantial build-up of both soil TP and AP. As estimated earlier, about 200 kg P ha–1 was applied annually as part of the manure.

Available Potassium
In contrast to soil AP, soil AK showed significant yearly declines from a beginning level of 160 mg kg–1 for treatments that did not receive manure or straw (Fig. 6). Trend lines for AK (Table 4) showed yearly decline rates of 2.06, 2.44, and 2.95 mg kg–1 yr–1 for the CK, N, and NP treatments, respectively. This is contrary to the general belief that most soils of the Loess Plateau in China and of the alluvial floodplain in Asia are high in K and that K is a rare limiting factor (Su, 2001; Bajwa, 1994). However, Liu et al. (2000), and Liu and Yao (2003) also showed similar declines in northwest China, and Kraus (2001) reported AK declines in arid/semiarid climatic conditions in West Asia and North Africa. In contrast, AK levels increased with time at rates ranging from 0.98 to 1.18 mg kg–1 yr–1 for plots receiving manure or straw. This demonstrates that inputs of K with organic materials resulted in a build-up of soil AK because manure or straw generally contains high amounts of K as mentioned early. This study clearly showed that high levels of production resulting from the use of inorganic N and P fertilizers and removal of aboveground biomass greatly reduced the level of AK. The extent of K deficiency in this study, if any, is not clear although grain yields of the SNP treatment were significantly higher than the NP treatment, particularly in the latter years when the AK levels were greatly reduced. Potassium could possibly have become limited in the NP treatment, but there is no data available to support this conjecture.

Soil available K decline without straw or manure in this study should be also paid great attention to intensive annual cropping system of wheat and corn rotation in Loess Plateau of Northwest China. For long, application of K by farmers in the region is typically rare, and they are not aware of soil K deficiency when high amounts of NP fertilizer are added to improve grain yield and think the soil is usually high in K.


   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
The objectives of this study were to follow grain yield and WUE trends for wheat and corn in annual cropping systems under a 24-yr dryland fertilization experiment, and to monitor long-term effects of fertilization on soil properties. Results showed that the addition of organic materials and inorganic fertilizers significantly enhanced grain yields, water use, and soil chemical properties if compared with no additives or addition of only inorganic N and P. Overall, the grain yields and WUEs for both wheat and corn crops showed downward trends over the study period, with the exception of MNP and CK treatments for wheat. Comparatively, the decline amounts in corn were much higher than in wheat. Unlike grain yield, soil nutrients showed a gradual build-up of SOC, TN, and TP. But, there was a gradual depletion of soil nutrients for the CK plots and for the plots receiving only inorganic fertilizers with lapse of time. A gradual decline of available nutrients would certainly explain a decline in grain yields with increasing years, but this is inconsistent with the build up of SOM and the macronutrients N and P in the M, NP, SNP, and MNP. This might be attributed soil chemical properties changing both negatively and positively based on the treatments, and was most likely linked to gradual dry weather as well as their interactions. Nonetheless, results indicate that adding only N or NP fertilizers may result in a deficiency of other nutrients and a decline in soil chemical properties, and that addition of organic materials along with inorganic fertilizers is necessary for sustainable production.

The study concludes the importance of returning straw to the soil, or adding manure in cases where the straw is removed. Particularly, returning straws to the soil should be recommended to farmers of this dryland area. This will help them toward better land use, minimize of production input costs, and improve the agricultural economy. To meet the food demands of a rising population in this area, considerable chemical fertilizers are required for grain production in this dryland region. But results, from this long-term fertilization experiment under cereals crop rotation system of wheat and corn, demonstrate nutrient depletion and deterioration of soil fertility as well as grain yield declines will be inevitable when only chemical fertilizers added or without nutrients or low amount of N. Therefore, the farmers of this area should be encouraged to manage the nutrients and soil fertility based on a balanced approach and organic materials should be combined with inorganic fertilizers to increase crop productivity and agricultural sustainability, as it is difficult to build up soil N, P, and K once they are depleted.

More importantly, considering this long-term experiment, researchers and farmers and government should come together in research, extension, evaluation, and dissemination of technologies to fully meet the objectives. Thorough activities like regular meetings, lectures, and field days, a feed and feedback information system between scientists and farmers should be recommended to let farmers have to be given access to the most appropriate and cost-effective technologies for their particular circumstances in nutrient and soil fertility management.


   ACKNOWLEDGMENTS
 
This study was financially supported by Chinese Ministry of Science and Technology under Key Technologies R&D Programme 2001BA508B11, a special support for China Scholarship Council. These assistances are gratefully acknowledged. We also thank Shiming Gao for his assistance, and appreciate our team work colleagues who provide us with unpublished data.

Received for publication April 28, 2004.


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





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 Google Scholar
Google Scholar
Right arrow Articles by Fan, T.
Right arrow Articles by Gao, Y.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Fan, T.
Right arrow Articles by Gao, Y.
Agricola
Right arrow Articles by Fan, T.
Right arrow Articles by Gao, Y.