Constructions Of Soil Mineral Nutrient Gradation Index

Published Date: 02 Nov 2017

Abstract: Recent years, reconciliation of contradiction between the contribution of fertilizer application to modern food production and the problems (e.g., cereal yield have no response to plus fertilizer rate, low profitable efficiency, negative effect on the environment and ecology, etc.) that caused by fertilizer excessive application, has been the focus of modern agronomy. 44 field experiments were conducted during the latest 3 years based on the concept that the contradiction may get moderated by construction of RFARS (Recommendation of Fertilizer Application Rate System) according to local farming conditions, in lowland rice farming region of Karst geography in Guangxi province, China. Every field experiment was consist of 3 factors [Nfer (N fertilizer), Pfer ( P fertilizer), and Kfer (K fertilizer), respectively], 4 fertilizer rate level [ 0 (zero rate), 1 (half of 2 level), 2 (local conventional fertilizer application rate), 3 (1.5 times of 2 level), respectively] and 14 subplots [treatments as ①N0P0K0 (0 level Nfer, 0 level Pfer, 0 level Kfer）, ②N0P2K2, ③N1P2K2, ④N2P0K2, ⑤N2P1K2, ⑥N2P2K2, ⑦N2P3K2, ⑧N2P2K0, ⑨N2P2K1, ⑩N2P2K3, ⑪N3P2K2, ⑫N1P1K2, ⑬N1P2K1, ⑭N2P1K1, respectively). SMNDI (Soil Mineral Nutrition Deficiency Index) was build up according to relationship between value of Nmin (mineral N), Pmin (mineral P) as well as Kmin (mineral K) from lab test of soil that sampling before transplantation of lowland rice, and relative yield that from the ratio of grain yield obtained from subplots ②N0P2K2 or ④N2P0K2 or⑧N2P2K0 [that zero Nfer or Pfer or Kfer application while other 2 factors receive 2 rate level], to the highest grain yield respectively from subplot (that 0 or 1 or 2 or 3 rate level of Nfer or Pfer or Kfer while other 2 factors receive 2 rate level). As a result, 6 grades of soil Nmin, Pmin and Kmin availability were defined according to 5 grain relative yields (65%, 75%, 85%, 90%, and 95%, respectively). Furthermore, the theoretical optimum Nfer, Pfer and Kfer rate for each field experiment was simulated by a quadratic equation with 3 variables combined with quadratic equation with 1 variable as well as the linear plus plateau model. Optimum Nfer, Pfer, Kfer rate recommendation equation was obtained by regression of optimum fertilizer rate recommendation of each experiment set against the soil test value of Nmin, Pmin, Kmin from the same site. Finally, advisory service of fertilizer rate recommendation is available for producer of lowland rice under local farming conditions based on soil test values of their farm.

Key words:"3414"field experiments; nutrient deficiency index; relative yield; fertilization recommendation; soil test

0 Introduction

Along with utilization and wide spread of fertilizer, corn yield has increased significantly, and contribution a lot to resolution of contradiction between food and increasing population in China. However, serious problems [ e.g., [ 1 2 3 4 5 6 ]cereal yield has no response to plus fertilizer rate(Zhang F S, et al.,2007; Cao Z H,2003), out of balance of necessary nutrients for corn growth(Zhu Z L,2000; Wang Q B, et al.,19XX;), low profitable efficiency as well as negative effect on environment and ecology(Wang X R, et al., 2007; Wang M L, et al.,2010), etc.] that caused by fertilizer excessive application, will become obstacles to sustainability of food production with economy, efficiency, stability, and harmless to ecology. As a result, construction of RFARS(Sun Y X, et al.,2009;8;9;10; S. Dowdle et al., 1988;7 8 9 10 11) ( Recommendation of Fertilizer Application Rate System) according to soil test value of mineral N(Nmin), mineral P(Pmin) and mineral K(Kmin) is an important way to solve these problems. Studies focus on both the stabilization of soil fertility and the balance of the necessary nutrients for corn growth have been conducted by researchers all around the world since the last century. Even in 1940s，Bray R.H [12]from US had developed both concept ASN（Availability of Soil Nutrient）and Relative Yield，and his further research discovered that a significant relationship between soil mineral nutrients and corn yield( or nutrient uptake by corn) could be built up by quantitative mathematics models. Other researchers [13-24](e.g., Williams J D, et al.,2007; Olfs H W, et al., 2005;Scharf P C, et al., 2001; Chen Q, et al.,2005; Attanandana T, et al., 2007; Saleque M A, et al., 2004; Quaggio J A, et al.,1998; Liu X J, et al., 2005; Peltovuori T,1999; Anderson-Cook C M, et al., 1999; Heckman J R, et al,2006; Zarth B J, et al,2001) has a good contribution more or less to fertilization science too. Plenty field trial for fertilizer recommendation based on soil mineral nutrient test were conducted since 1980s in China. And many fertilizer recommendation techniques were developed by both the combination and complement among various methods or techniques. Chinese researchers have constructed different RFARS for many corns [e.g., lowland rice(25-26Li J, et al., 2010; Ji L, et al., 2011), winter wheat(27Li L, et al., 2010), oil-seed rape(28Yang L P, et al., 2011) and peanut(29Lou C R, et al., 2008), etc.] under various ecological regions and farming conditions. At the same time, many techniques and methodologies [e.g., 30-31 SPAD-502 leaf chlorophyll meter(Li Z H, et al., 2006), 32plant nitrate test(Jia L L, et al., 2007), 33Nitrogen Status diagnosis of rice by using a Digital Camera(Jia L L, et al., 2009),34 nitrate plant saps/petiole test(Wehrmann, J., et al., 1982) as well as 35Spectrum Technologies(Liu H B, et al., 2004), etc.] were introduced into China and spread widely. It is important to mention that all of the techniques mentioned above should combine with conventional fertilizer recommendation method to get improvement in fertilization.

1980s, as the scientific basis for both applications of fertilizer and spread of fertilization techniques, both RFARS and SMNDI (36 Zhou Q X, et al ., 1983)were once built up for lowland rice in China. However, 30 years have passed, both original RFARS and SMNDI could not provide sufficient direction any more to current corn production since many farming factors (e.g., soil fertility, yield level, plant variety, cultivation techniques, etc.) have changed significantly year to year. In 2005, "Testing Soil for Formulated Fertilization Project" was conduct by National Agriculture Department to provide new resolution to many problems (e.g., sufficient utilization of farmland potential, improvement of soil, protection of ecology and environment, efficiency of fertilizer, etc. ) and to get detail information (e.g., evaluation and gradations of farmland fertility, content of soil main mineral nutrient, index of soil nutrient sufficiency, etc.) which was necessary to adjust RFARS. To improve farm management around China, considering the significant variance of farming or environmental conditions among ecological regions, new RFARS need to be built up under local farming conditions. In this study, we extract data from 44 experiments that were conducted in the classic Karst topographic region of Guangxi province in order to build up RFARS for local farming conditions.

1 materials and design

1.1 materials

Our study was conducted in two counties: Debao County and Luocheng County, which have classic Karst geographic character and 5217Km2 as total area. There is several classic characters contribute to Karst geography：1) There are many mountains and rocks, but little farmlands, as a result, it is difficult to find experiment sites with a capacity of adequate area for repetitive plots. 2) With a thin soil layer and slope, farmlands that gestate from Karst geography have a weak ability in maintenance of nutrient and water. 3) There are 7 groups of soil, 20 sub-groups of soil and 71 soil types, and these types of soil intermingle in complex patterns. Paddy soil and lime soil take 80% of the County’s cultivated land, as a result, fertilizer recommendation could not only base on soil nutrient but also the Acidic and Basic capacity. 4) The fertility of cultivated land is relatively sterile and medium&low fertility farmland takes a large priority percentage of county farmland. 5) The vertical utilization discrepancy of the land resource is notable, and three-dimensional agriculture is prominent.

1.2 Design of experiments

This study consist of 44 "3414"experiment sets, all of which were consist of 3 factors (Nfer, Pfer, and Kfer, respectively),4 fertilizer rate level[0(zero rate),1(half of 2 level),2(local conventional fertilizer application rate) and 3(1.5 times of 2 level),respectively] and 14 subplots (configuration as table 1:①N0P0K0, ②N0P2K2, ③N1P2K2, ④N2P0K2, ⑤N2P1K2, ⑥N2P2K2, ⑦N2P3K2, ⑧N2P2K0, ⑨N2P2K1, ⑩N2P2K3, ⑪N3P2K2, ⑫N1P1K2, ⑬N1P2K1, ⑭N2P1K1, respectively) according to randomized complete block designs(37 Zhang F S,2006). 165 Kg•hm-2, 65 Kg•hm-2, 120 Kg•hm-2 were level 2 rate（local conventional fertilizer application rate） of Nfer, Pfer, and Kfer, respectively. 44 experiment sites which were selected from farmland that has both classic agrotype and classic fertility were proportioned well within the local area.

Just as mentioned before, it is difficult to find experiment sites with a capacity of adequate area for repetitive plots. As a result, total 44 experiment sets consist of 9 sites with 3 repeats and 8 sites with 2 repeats as well as 27 sites with no repeat. 14 subplots, each of which has the same volume of 20m2, were arranged randomly within one experiment set. Lowland rice( Oryza sativa L., local variety） was sowed in the third ten days of February and transplanted in first ten days of April by local conventional method. As illustrated in Table 1，40% Nfer combined with 50% Kfer and total Pfer was applied as a base fertilizer in each subplot of experiment set, furthermore，60% Nfer and 50% Kfer were applied as sidedress. Base fertilizer should be applied deeply into the plough layer of the subplot with shallow irrigation layer. And the fertilized subplot should not drain immediately after fertilization. Protection lines with 1m width were set around experimental sites, and the management during the growth period was done as same as conventional field scale production. Both of dry grain yield and fresh yield that harvests from subplot respectively in each experiment set were recorded by researchers.

Table 1: Subplot configuration of each experiment set and Nfer, Pfer, Kfer rate in each subplot.

1.3 Soil mineral nutrients tests

Soil samples were taken with an auger from different soil layer (0~40cm) before irrigation for transplantation of rice seeding, 6 soil samples were homogenized and 500g sub-sample was filled into a plastic bag. Samples were stored in a box with ice immediately after sampling and transported to the lab for analysis of pH value, Nmin, Pmin and Kmin as well as other factors. The methods(38 Zhang F S,2006) we toke to identify soil index were shown as follow: pH value (Glass-electrode method); Nmin (Kjeldahl determination); Pmin (NaHCO-Extraction and Ammonium Molybdate-Tartaric Emetic-Ascrbic Acid Colorimetry method); Kmin (Atomic Absorption Spectrophotometer and Ammonium Acetate Extraction method );slowly Kmin (Atomic Absorption Spectrophotometer and Nitric acid extraction method);available Cu, available Fe, available Mn, available Zn (Atomic Absorption Spectrophotometer and DTPA extraction method); organic matter (Azomethine-H colorimetry to acquire available B, Dichromate potassium Oxidation method).

2. Data processing and calculation

2.1 constructions of soil mineral nutrient gradation index

After harvest, relative yield was obtained from the ratio of grain yield of subplots（zero Nfer or Pfer or Kfer application rate while other 2 factors receive 2 level rate）, to the highest grain yield respectively of subplot (that's 0 or 1 or 2 or 3 level rates of Nfer or Pfer or Kfer while other 2 factors receive 2 rate levels). Equation of SNDI was built up according to the linear relationship between the value of mineral Nmin, Pmin, and Kmin from lab test of soil that sampling before transplantation of lowland rice and relative yield. And 6 grades of SNDI were determined by five soil nutrient values which were calculated by substitution of 5 grain relative yield values( 65%, 75%, 85%, 90% and 95%,respectively) into the above equation of SNDI. 6 grades of SNDI was defined as follow: very low (relative yield≤65%), low (65%<relative yield≤75%), medium (75%<relative yield≤85%), high (85%<relative yield≤90%), higher (90%<relative yield≤95%), highest (95%<relative yield).

2.2 Determination of optimum fertilizer rate in each experiment

QEV3 (quadratic equation with 3 variables) model combined with QEV1 (quadratic equation with 1 variable) model as well as LPP (linear plus plateau model) (39Ji L, et al., 2011)were used to simulate the relationship between fertilizer rates and expected grain yield for each experiment set. And the best - fitting model was selected based on scatter plot trends and coefficient of determination derived from each mode. Finally, optimum Nfer or Pfer or Kfer fertilizer rate were determined based on the optimum profile of each experiment site.

Firstly, QEV3model was illustrated as follows:

y=b0+b1x1+b2x12+b3x2+b4x22+b5x3+b6x32+b7x1x2+b8x1x3+b9x2x3 1)

In equation 1), y is the expect grain yield; x1, x2, x3 are Nfer, Pfer and Kfer rates, respectively; b0~b9 are constant. If model simulated the relationship successfully (quadratic term coefficients are negative, monomial coefficients are positive, and significant F-test), optimum Nfer, Pfer and Kfer rate will calculate out based on the axiom: Marginal single fertilizer cost= Marginal grain profile (dx•xp =dy•yp). Token x1, x2 and x3 as variable respectively and derivation, 3 equations were got as follows:

b1+2b2x1+b7x2+b8x3=x1p/y p 2)

b3+2b4x2+b7x1+b9x3=x2p/y p 3)

b5+2b6x3+b8x1+b9x2=x3p/y p 4)

In equation 2), 3) and 4), x1p, x2p, x3p and yp are Nfer, Pfer , Kfer price and grain price respectively, bi of 1) and prices were substituted into equation 2), 3), and 4), furthermore optimum Nfer, Pfer, Kfer rate in each experiment site will obtained by solving of an equation set.

Another, QEV1 model was illustrated as follows:

y=a+bx+cx2 5）

In equation (5), y is the expect grain yield. x is fertilizer (Nfer or Pfer or Kfer) application rate, a is intercept, b is monomial coefficient, c is quadratic coefficient. In each experiment set, the optimum Nfer recommendation rate was stimulated by data from subplots (②, ③, ⑥ and ⑪ , respectively ), at the same time, the optimum Pfer recommendation rate was stimulated by data from subplots (④, ⑤, ⑥ and ⑦, respectively), and Kfer recommendation rate was stimulated by data from subplots (⑥, ⑧, ⑨ and ⑩, respectively) . If model simulated successfully (quadratic term coefficient is negative, monomial coefficient is positive, and significant F-test), the optimum Nfer, Pfer, Kfer rate will calculate out based on the axiom: Marginal single fertilizer cost= Marginal grain profile (dx•xp =dy•yp). Taking x as variable and derivation both sides of equation 5), furthermore equation 6) was got as follows:

b+2cx=xp/yp 6）

In equation 6), xp and yp are fertilizer (Nfer or Pfer or Kfer) price and grain price, respectively, b of 5) and price were substituted into 6), and optimum Nfer, Pfer, Kfer rate in each experiment site will obtained by solving of equation 6).

The third, LPP model was shown as follows:

y=a + bx（x≤C） 7）

y=p（x>C） 8）

In equation 7) and 8), y is grain yield, x is fertilizer (Nfer or Pfer or Kmin) application rate, a is intercept, b is monomial coefficient, C is intersection of line and plateau (=optimum fertilizer rate), p is plateau yield (=highest grain yield).

Nfer, Pfer and Kfer were counted as N, P2O5 and K2O, respectively, and the current prices are 5.4￥•Kg-1,5.1￥•Kg-1 and 5.2￥•Kg-1 respectively, at the same time, grain current price is 1.7￥•Kg-1. It is important to mention that, the prices of fertilizers and grain are changeable, as a result, the optimum fertilizer recommendation will change a little (40 Hans-Werner Olfs,2005) along with market fluctuation.

2.3 Determination of the optimum fertilizer rate for Karst lowland rice regions

The optimum Nfer, Pmin, Kfer application rate for each experiment site were calculated out through 3 models above. Furthermore, logarithmic model was used to simulate the relationship between soil test nutrient values and optimum fertilizer rates in all of 44 experiment sets within Karst lowland rice regions. All of the equation was simulated and calculated by SAS software package.

3 results

3.1 Gradation of soil nutrient sufficiency index

3.1.1 Soil Nmin sufficiency index grades

There is one of 44 experiment sets was defined as unsuccessful after inspection of the data that with the extreme low grain yield, and 3 experiment sets have a soil Nmin of under 1g•Kg-1, but corresponding relative yield reach sticking 85%, as well as 2 experiment sets have a soil Nmin of above 2.5g•Kg-1, but corresponding relative yield under 70%. As a result, removing relative data from above 6 experiment sets, reminder 38 experiment sets were used to simulate the relationship between relative yields and samples Nmin, results showed in Figure 1a, the regression equation is y = 0.1442x1+0.4565 (correlation coefficient R2=0.439, indicates significance at P<0.01). Relative yields( 65%, 75%, 85%, 90% and 95%, respectively) were substituted into the equation, and 5 soil mineral N indexes (1.34g•Kg-1、2.03g•Kg-1、2.73g•Kg-1、3.07g•Kg-1and 3.45g•Kg-1, respectively) were calculated out by the equation. Finally, soil Nmin sufficiency index was graded as shown in Table 2.

**Indicates significance at P＜0.01.The same as the followings

Figure 1a: The relationship between soil Nmin contents and relative yields of crop (n=38)

Figure 1b: The relationship between soil Pmin contents and relative yields of crop (n=39)

Table 2: Gradation of soil mineral nutrient index

3.1.2 Soil Pmin sufficiency index grades

There is one of 44 experiment sets was defined as failed after inspection of the data that with the extreme low grain yield, and 2 experiment sets have a soil Pmin of under 5mg•Kg-1, but corresponding relative yield reach sticking 85%, as well as 2 experiment sets have a soil Pmin of above 25mg•Kg-1, but corresponding relative yield under 70%. As a result, removing relative data from 5 above experiment sets, reminder 39 experiment sets were used to simulate the relationship between relative yields and samples Pmin, results shown in Figure 1c, the regression equation is y = 0.01003x2+0.6515 (correlation coefficient R2=0.545, indicates significance at P<0.01). Relative yields (65%, 75%, 85%, 90% and 95%, respectively) were substituted into the equation, and 5 soil Pmin indexes (-0.0015mg•Kg-1, 9.8mg•Kg-1, 20mg•Kg-1, 24.8mg•Kg-1 and 29.7mg•Kg-1, respectively) were calculated out by the equation. Because soil Pmin=-0.0015mg •Kg-1 was a calculated value and nonsense. We deleted this index. Finally, soil Pmin sufficiency index was graded as illustrated in Table 2.

3.1.3 Soil Kmin sufficiency index grades

There is one of 44 experiment sets was defined as failed after inspection of the data that with the extreme low grain yield, and 1 experiment set has a soil Kmin of under 20mg•Kg-1, but corresponding relative yield reach sticking 85%, as well as 1 experiment set has a soil Kmin of above 100mg•Kg-1, but corresponding relative yield under 75%. As a result, removing relative data from 4 above experiment sets, reminder 41 experiment sets were used to simulate the relationship between relative yields and samples Kmin (results shown in Figure 1c). The regression equation is y = 0.00336x2+0.5844 (correlation coefficient R2=0.677, indicates significance at P<0.01). Relative yields（65%, 75%, 85%, 90% and 95%, respectively）were substituted into the equation, and 5 soil Kmin indexes( 19.5mg•Kg-1, 49.5mg•Kg-1,79.3mg•Kg-1,94mg•Kg-1 and 109mg•Kg-1, respectively) were calculated out by the equation. Finally, soil Kmin sufficiency index was graded in Table 2.

Figure 1c: The relationship between soil Kmin contents and relative yields of the crop (n=41)

Figure 2a: Relationship between soil Nmin contents and fertilizer recommendation rates (n=38)

3.2 Construction of recommendation for Nfer, Pfer and Kfer application rate system

The optimum fertilizer recommendation rate of 20% of"3414"experiment sets was determined successively by QEV3 model (41-44 Peltovuori T,1999; Zarth B J, et al., 2001; Li J, et al., 2010; Li L, et al., 2010). However, optimum Nfer, Pfer and Kfer recommendation rate was determined successively with a surprising percentage of 68%, 86% and 84%, respectively by QEV1 model. In this study, experiment sets that were failed to determine the optimum fertilizer rate by QEV3 model were simulated furthermore by both QEV1 model and LPP model. And the model with the highest correlation coefficient was selected to calculate optimum fertilizer application rate in each site. At the same time, the optimum fertilizer application rate was defined as "0" if all models are failed to determine the optimum fertilizer application rat.

After optimum fertilizer recommendation rate in each experiment was calculated by above models, logarithmic model was used to simulate the relationship between soil mineral nutrient values and optimum fertilizer application rates in all of 44 experiment sets.

3.2.1 Optimum Nfer rates for each experiment site

As shown in Figure 2a, relationship between soil Nmin test value and Nfer recommendation in all 43 successful experiment sets, has been described by the equation: y = -99.1·ln (x) + 257.6 (correlation coefficient R2=0.677, indicates significance at P<0.01). 6 soil Nmin indexes [1.21g•Kg-1 (lowest test value of soil Nmin), 1.34g•Kg-1, 2.03g•Kg-1, 2.73g•Kg-1, 3.07g•Kg-1 and 3.45g•Kg-1, respectively] were substituted into the equation, and 6 Nfer rates (238kg•ha-1, 228kg•hm-2, 186kg•hm-2, 157kg•hm-2, 145kg•hm-2 and 135kg•hm-2, respectively) were calculated out by the equation. Finally, Optimum Nfer application rate recommendation system was built up as shown in Table 3.

3.2.2 Optimum Pfer rates for each experiment site

As shown in Figure 2b, relationship between soil Pmin test value and Pfer recommendation in all 43 successful experiment sets, has been described by the equation: y = -35.2·ln(x) + 156.1(correlation coefficient R2=0.567, indicates significance at P<0.01). 5 soil Kmin indexes [3.5mg•Kg-1 (lowest test value of soil Kmin), 9.8g•Kg-1, 20g•Kg-1, 24.8mg•Kg-1 and 39.7mg•Kg-1, respectively] were substituted into the equation, and 5 P fertilizer rates (112kg•hm-2, 75kg•hm-2, 50kg•hm-2, 43kg•hm-2, 145kg•hm-2 and 36kg•hm-2, respectively) were calculated out by the equation. Finally, Optimum Pfer application rate recommendation system was built up as shown in Table 3.

3.2.3 Optimum Kfer rates for each experiment site

As shown in Figure 2c, relationship between soil Kmin test value and Kfer recommendation in all 43 successful experiment sets, has been described by the equation: y = -43.21·ln(x) + 274.4 (correlation coefficient R2=0.567, indicates significance at P<0.01). 6 soil Kmin indexes[ 6.99 mg•Kg-1(lowest test value of Kmin), 19.5 mg•Kg-1, 49.5g•Kg-1, 79.3 mg•Kg-1, 94 mg•Kg-1 and 109 mg•Kg-1, respectively] were substituted into the equation, and 6 Kfer rates( 190kg•hm-2, 146kg• hm-2, 105kg• hm-2, 85kg• hm-2, 78kg• hm-2 and 71kg• hm-2, respectively) were calculated out by the equation. Finally, Optimum Kfer application rate recommendation system was built up as shown in Table 3.

Figure 2b: Relationship between soil Pmin contents and fertilizer recommendation rates (n=39)

Figure 2c: Relationship between soil Kmin contents and fertilizer recommendation rates (n=41)

Table 3: Optimum fertilizer application rate system based on soil mineral nutrient deficiency index grades

4 Discussions

4.1 soil mineral nutrient sufficiency indexes

The relative yield of the subplot with specific nutrient deficiencies in each experiment set is an important parameter for construction of SMNDI which is the basis of TSFFS (Testing Soil for Formulated Fertilization System). Resulting from complex soil parent materials and complicated topography, the variance of fertility and soil nutrient content in field to field is significant. As a result, it will cause a waste of fertilizer if the soil mineral nutrient sufficiency index system was built up too roughly(45 46 Fageria N K,et al., 1997; Sun Y X, et al., 2009). Limited by Law of Diminishing Marginal Returns (47 Ji L, et al., 2011), the curve that describes the relationship between fertilization rate and relative yield began to flatten when the relative yield reaches 90% and above, furthermore resulting in fertilizer efficiency decline. So we had refined the soil mineral nutrient grades [very low (relative yield<65%), low(65%≤relative yield<75%), medium (75%≤relative yield<85%), high (85%≤relative yield<90%), higher (90%≤relative yield<95%) and highest (95%≤relative yield), respectively] to save fertilizer and to protect the environment as well as to improve efficiency.

Table 4: Variations of nutrient factor in cultivated land soils of this evaluation with the second survey

Organic matter content/

(g·kg-1)

Area/hm2

Percentage∕%

Available P content/

(mg·kg-1)

Area/hm2

Percentage∕%

Available K content/

(mg·kg-1)

Area/hm2

Percentage∕%

Second survey

This evaluation

Second survey

This evaluation

Second survey

This evaluation

Second survey

This evaluation

Second survey

This evaluation

Second survey

This evaluation

≥40

66

2 492

0.3

13.21

≥30

191

2 615

0.87

13.86

≥150

234

1.07

≥30～40

4 796

7 504

21.85

39.78

≥20~30

300

3 674

1.37

19.48

≥100~150

1 178

316

5.36

1.68

≥20～30

11 378

8 325

51.81

44.13

≥10~20

1 868

8 068

8.51

42.77

≥50~100

5 747

5 123

26.17

27.15

≥10～20

3 896

539

17.74

2.86

≥5~10

5 014

3 934

22.83

20.85

≥30~50

4 291

6 982

19.54

37.01

＜10

1 823

5

8.3

0.02

<5

14 586

574

66.42

3.04

<30

10 509

6 444

47.86

34.16

4.2 Compared with historical data

Local SMNDI was built up once during the second soil survey in 1980s as follows: 1) Pmin was divided into 4 grades (Pmin<3mg•Kg-1, 3mg•Kg-1≤Pmin<5mg•Kg-1, 5mg•Kg-1≤Pmin<10mg•Kg-1 and 10mg•Kg-1≤Pmin, respectively); and 2) Kmin was divided into 4 grades (Kmin<30mg•Kg-1, 30mg•Kg-1≤Kmin<50mg•Kg-1, 50mg•Kg-1≤Kmin<100mg•Kg-1 and 100mg•Kg-1≤Kmin, respectively). Although this SMNDI had a great contribution to local agriculture ever before, it can’t offer available information any more for local farming since many farming factors (e.g., soil fertility, yield level, plant varieties, cultivation techniques, etc.) have changed significantly within past 30s years. Take Debao county as an example, shown in Table 4, data on soil mineral nutrient which were tested in the Evaluation of Cultivated Land Fertility and Gradation which was accomplished in 2012, show that the overall farmland fertility has increased significantly in comparison with the second cultivated land survey which was conducted in 1980s (48Zhao L, et al., 2013).

Along with the difference of soil mineral nutrient, SMNSI has changed greatly too: 1) Pmin and Kmin were divided roughly into 4 grades in 1980s, but Nmin and Kmin were divided into 6 grades and Pmin was divided into 5 grades this time. 2)along with Pmin has increased significantly, Pmin index has a higher level in this study (Pmin<9.8mg•Kg-1, 9.8mg•Kg-1≤Pmin<20mg•Kg-1, 20mg•Kg-1≤Pmin<24.8mg•Kg-1, 24.8mg•Kg-1≤Pmin<29.7mg•Kg-1 and 29.7mg•Kg-1≤Pmin, respectively) than that of 1980s(Pmin<3mg•Kg-1, 3mg•Kg-1≤Pmin<5mg•Kg-1, 5mg•Kg-1≤Pmin<10mg•Kg-1 and 10mg•Kg-1≤Pmin, respectively), in the meanwhile, Kmin indexes do not change too much in this study (Kmin<19.5mg•Kg-1, 19.5mg•Kg-1≤Kmin<49.5mg•Kg-1, 49.5mg•Kg-1≤Kmin<79.3mg•Kg-1, 79.3mg•Kg-1≤Kmin<94mg•Kg-1, 94mg•Kg-1≤Kmin<109mg•Kg-1 and 109mg•Kg-1≤Kmin, respectively) than that of 1980s (Kmin<30mg•Kg-1, 30mg•Kg-1≤Kmin<50mg•Kg-1, 50mg•Kg-1≤Kmin<100mg•Kg-1 and 100mg•Kg-1≤Kmin, respectively). This comparison indicated that: along with many farming factors (e.g., soil fertility, yield level, plant varieties, cultivation techniques, etc.) have changed significantly, SMNSI and RFARS should change according to farming conditions.

4.3 Fertilization recommendations in different soil parent materials

Combined with data from Evaluation of Cultivated Land Fertility and Gradation in 2012, both 48.01% of soil that originate from Quaternary Red Clay and 46.24% of soil that originated from Sandy Shale have a Nmin content of Nmin≥2.5g•Kg-1 or above (relative yield≥80%). As a result, to avoid waste of N fertilizer, we suggest resampling, if soils that originate from Quaternary Red Clay or Sandy Shale give a low Nmin content test, in the practical fertilizer rate recommendation. 100% of soils that originate from Silica Shale have a low content of Nmin (1.01~1.5g•Kg-1, relative yield<67%), we recommend applying superfluous Nfer to cultivate the soil in such type of farmland.

89.00% of the soils that originate from Limestone have a Pmin content of Pmin≥10mg•Kg-1( relative yield≥75%）. 100% of soils that originate from Silica Shale has a low content of Pmin<10mg•Kg-1, we recommend applying superfluous Pfer to cultivate the soil in such type of farmland.

81.00% of soils that originate from Granite Moorstone have a Kmin content of Kmin≥100mg•Kg-1（relative yield≥92%）. Those soils (100% of soils originate from Siliceous Shale, 24.5% of soil originate from Quaternary Red Clay, 25.95% of soil originated form River Alluvium, 30.95% of soil originated from Sandy Shale) have a low Kmin content of Kmin<50mg•Kg-1（relative yield<75%）, we recommend applying superfluous Kmin to cultivate the soil in such type of farmland.

5 conclusions

1) Such characters (fewer subplots treatment, higher efficiency, suitable for QEV3 model, suitable for QEV1 model, and suitable for LPP model) were advantages of "3414"experiment that was conducted according to square optimum regression design. RFARS could be built up by "3414" experiment sets well.

2) Along with the fertilizer application year to year, soil mineral nutrients have changed significantly. In this study, grain yield in some experiments did not increase along with increasing of fertilizer factors, and this may cause from variances between conventional fertilization and plant actual demand. As a result, the optimum fertilizer recommendation for each site was simulated by QEV3 model with a low success rate (18%). And some QEV1 model was failing to simulate the relationship too. Except grain yield in some experiments did not increase along with increasing of fertilizer factors, other reasons (uneven management in all subplots, no repeat subplots, and experimental design defect) should be considered as well. Considering higher Nmin, Pmin, Kmin contents in many subplots, and combined with experiences, we could give a recommendation of "0" fertilizer rate or just apply a little "starter" for subplots that did not increase grain yield along with increasing of fertilizer,

3) Studies on 44 "3414"experiments in the Karst area of Guangxi province indicated that: Nfer rates are 238~228, 186~228, 157~186, 145~157, 135~145, and 0~135Kg•hm-2 when soil Nmin tests were classified by: very low, low, medium, high, higher, and the highest( Nmin<1.46g•Kg-1, 1.34g•Kg-1≤Nmin<2.03g•Kg-1, 2.03g•Kg-1≤Nmin<2.73g•Kg-1, 2.73g•Kg-1≤Nmin<3.07g•Kg-1, 3.07g•Kg-1≤Nmin<3.45g•Kg-1 and 3.45g•Kg-1≤Nmin), respectively. Pfer rates are 75~112, 50~75, 43~50, 36~43 and 0~36Kg•hm-2 when the soil Pmin test was classified by low, medium, high, higher, and the highest (Pmin<9.8mg•Kg-1, 9.8mg•Kg-1≤Pmin<20mg•Kg-1, 20mg•Kg-1≤Pmin<24.8mg•Kg-1, 24.8mg•Kg-1≤Pmin<29.7mg•Kg-1 and 29.7mg•Kg-1≤Pmin), respectively. As well as Kfer rates are 190~145, 105~146, 85~105, 78～85, 71~78, and 0~71 Kg•hm-2 when the soil Kmin test was classified by very low, low, medium, high, higher, and the highest (Kmin<19.5mg•Kg-1, 19.5mg•Kg-1≤Kmin<49.5mg•Kg-1, 49.5mg•Kg-1≤Kmin<79.3mg•Kg-1, 79.3mg•Kg-1≤Kmin<94mg•Kg-1, 94mg•Kg-1≤Kmin<109mg•Kg-1, 109mg•Kg-1≤Kmin), respectively.