The flowchart in

Figure 2 displays a series of basic steps for implementing our tool. Application of this approach, based on a raster, is mainly in three stages: land use spatial allocation, opportunity costs calculation, and tradeoff analysis. The agricultural land use optimal allocation (ALUOA) system [

20] was applied in the first stage. The spatial opportunity costs calculation is the intermediate stage that integrates the tradeoff analysis with the ALUOA system. Then it determines the final land use spatial reallocation based on the payment policy. We designated the improved tool as the “ALUOA-TOA” system.

**Figure 2.**
Flowchart of the integration of trade analysis and the land use optimal allocation system.

Firstly, our method uses the ALUOA system to generate an origin land use allocation map, which aims to maximize the economic income and would be realistic for farmers’ adoption without any payments. Secondly, it simulates opportunity costs of different land uses under the payment policy by calculating the potential crop yields and net returns. Different payment prices result in different opportunity costs for farmers’ crop selection and change. Finally, our method simulates the tradeoff of famers and the government. With the economic incentive from payments, farmers make the tradeoff on economic incomes of different crops and decide whether to adopt the payment policy. Meanwhile, the government would evaluate a reasonable payment price, which is a tradeoff on the environmental compensation and ecosystem service benefit. Then areas of crop change were identified and the original land uses were reallocated in a spatial display.

The details of our framework are as follows, with a case study. All data were processed and converted to pixels with 25-m resolution in ArcGIS 9.3 (ESRI Inc., Redlands, CA, USA). The 25-m pixel is seen as the analysis unit reflecting the farmers’ behavior.

#### 2.2.1. Land Use Spatial Allocation

The ALUOA system was developed by our study team and applied in the newly reclaimed region. It consists of three steps: land suitability evaluation (LSE), land use area optimization, and land use spatial allocation [

20]. Interested readers can find content details in paper [

20]. The LSE assesses the degree of satisfaction of the crop requirement by using a weighted linear method for summing up different evaluation criteria, which include soil qualities, terrain factors, water supply conditions, climatic data, and socially locational factors (

Table 1). In this study, wheat, corn, rice, cotton, sugar beets, oil plant, bast fiber plants, clover, vegetables, and fruit are crop types for planning purposes, according to the agricultural crop structure in the newly reclaimed region.

**Table 1.**
Land suitability evaluation criteria, data source, and processing.

**Table 1.**
Land suitability evaluation criteria, data source, and processing.
Criterion | Input dataset | Data source | Format | Processing |
---|

Soil texture | 1:100,000 Soil type maps of Yili region | Institute of Geographical Sciences and Natural Resources Research | polygon | Format transformation |

Soil depth | Soil sampling points | Fieldwork by the research team | point | Kriging interpolation |

Soil organic matter | Soil sampling points | Fieldwork by the research team | point | Kriging interpolation |

Sand dune waviness | 1:50,000 Topographic maps; Land use map 2008 | Institute of Geographical Sciences and Natural Resources Research;
Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences | raster | Selecting the sand distribution from the land use map and calculating the relative height from the DEM digitized from the topographic maps |

Soil erosion | 1:50,000 Topographic maps | Institute of Geographical Sciences and Natural Resources Research | raster | Calculating the gully density from the above DEM |

Water supply and drainage | 1:100,000 Land resource map of China | Institute of Geographical Sciences and Natural Resources Research | raster | Resampling |

Salinity | 1:100000 Soil type maps of Yili region; 1:1000000 Land resource map | Institute of Geographical Sciences and Natural Resources Research | raster | Format transformation |

> 10 °C accumulated temperature | 1:1000000 Accumulated temperature map of Yili region | Institute of Geographical Sciences and Natural Resources Research | raster | Format transformation |

Nearest distance to towns | 1:100000 Spatial map of towns in Yili region | Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences | raster | Distance calculation |

Nearest distance to roads | 1:100000 Spatial map of roads in Yili region | Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences | raster | Distance calculation |

Three scenarios were developed based on the previous study [

20]. The basic scenario was consistent with the agricultural structure of cereal, cash, and forage crops in the Yili region. Considering the common crop choices of four surrounding counties (Huocheng, Cabuchaer, Gongliu, and Xinyuan), cereal, cash, and forage crops account for 55.48%, 43.36%, and 1.15%, respectively, which represents a typical “cereal-cash” dualistic structure. Then we set the lower limit of percentages for cereal, cash, and forage crops as 50%, 30%, and 5%, respectively. Furthermore, wheat is a basic cereal food component in Yili; we set its proportion at not less than 20%. Based on crop farming trends and practices from the past 20 years, areas of rice and cotton will be expanded in the future. To curb the detrimental aspects of its development, we set the highest proportion of planting area for rice at not larger than 15%. With the above situations, we set these constraint conditions for the linear programming. The linear programming determined the optimal area of each land use type based on Equation (1) (constraint conditions) and Equation (2) (goal function) as follows:

where

I means the total land resource economic income (unit: yuan);

x_{1},

x_{2},

x_{3}, …,

x_{9} are the area proportions of wheat, corn, rice, cotton, sugar beets, oil plants, bast fiber plants, vegetables, fruits, and clover for the allocated area; and A is the area.

With the spatial distribution of LSE and the optimal area for each crop, the module allocates each land use type by a hierarchical optimal allocation method. Final land use spatial allocation maps are then generated. The basic scenario map was used in this study for the fundamental analysis (

Figure 3). Two other land use scenarios are intended to adjust the agricultural structure and increase ecosystem service values to realize agricultural sustainability. This suggests that the reasonable agricultural structure in the Yili region should be adjusted at a higher proportion of livestock farming [

21]. They were also designed with higher area proportions of clover than in the basic scenario.

Clover is a promising forage crop in the Yili region; it has higher economic value than traditional forage crops and an especially high ecosystem service in helping prevent land degradation [

21]. Compared to other crops, which maintain the vegetable cover only in the growing season, clover maintains the vegetable cover in the whole year as a perennial crop. Continuous vegetable cover would effectively protect the soil, which could otherwise be easily eroded in the spring. In addition, it has a high ecosystem service value in sand fixation and soil conservation in the study area.

However, the economic benefit of clover is lower than many other crops’. The above two scenarios are intended to protect the environment and prevent land degradation—the policy goals of mandatory planning set by the government and experts—but are not easily adopted by farmers. To effectively implement the sustainable land use allocation planning, the government should provide farmers with payments to turn to clover, even though it has low economic benefits. Therefore, clover was chosen as the target crop of the payment policy for the tradeoff analysis in our study.

**Figure 3.**
Spatial allocation map of the basic land use scenario.

**Figure 3.**
Spatial allocation map of the basic land use scenario.

#### 2.2.2. Conceptual Framework of Tradeoff Analysis

We integrated the tradeoff analysis with the ALUOA system to provide more information on land use planning. With payments for farmers provided by the government, the tradeoff analysis can evaluate how many lands would turn to clover at a specific payment price. Furthermore, our approach identified areas adopting the payment policy on the spatial display.

The tradeoff analysis in the ALUOA-TOA system was based on the TOA-MD model, with its basic assumptions [

14,

15,

16,

17,

18] but some different settings. It assumes that farmers make land use and management decisions to maximize their perceived economic wellbeing. A farmer’s choice between two competing land uses (

a and

b) is determined by the opportunity cost [

18] as follows:

where

ω is the opportunity cost;

v is the excepted value on land uses, which is defined as the net return (yuan/ha) in our study;

r is a vector of input and output prices for land uses;

s indexes the site (and in this study

s is a pixel); and

a and

b mean the land use at the pixel.

Alternately, it is assumed that practice

b produces more ecosystem services than practice

a, since the analysis here is based on the difference between the two practices. The parameter

e(s) is interpreted as the expected ecosystem service values obtained from changing land use, not as the realized supply of ecosystem services. With the

p_{e} as the ecosystem price (or payment price) for practice

b, the income of practice

b is

v(

r,s,b) +

p_{e}·

e(s), and the choice of farmers is assumed to be determined by Equation (4), as follows:

If ω(r,s,p_{e}) ≥ 0, farmers adopt practice a; if ω(r,s,p_{e}) < 0, farmers adopt practice b. In particular, for the pixels where ω(r,s) < 0, farmers adopt practice b without any payments.

In previous studies, this can provide a summary proportion of farmers adopting the payment policy but cannot provide the spatial variability of farmers’ adoption [

14,

15,

16,

17,

18]. It is unclear where practice

b changes from practice

a, a situation that will hamper decision making. Our strategy is to simulate the spatial distribution of opportunity costs for the tradeoff analysis. Then we can reallocate the basic land use scenario, changing practice

a to practice

b based on payments for farmers. From the spatial distribution of LSE scores, we can simulate the spatial distribution opportunity costs at each payment price in the tradeoff curve by applying series of crop yield simulations, input and output calculations, and opportunity costs calculations (

Figure 2).

#### 2.2.3. Crop Yield Simulation

LSE is the key in linking tradeoff analysis to the ALUOA system, which is a model for predicting potential land production [

22,

23]. The LSE is a prerequisite in achieving optimum utilization of the available land resources, while preserving highly suitable lands with high yields [

24,

25]. It collects and processes in mathematical equations climatic and other physical parameters that affect crop yields [

26]. Thus it can also be used to predict the crop yield based on land suitability [

23,

27]. There have been numerous attempts to predict crop yields from data on land qualities using fuzzy S-membership functions, which are appropriate and robust for both quantitative and linguistic variables [

28,

29,

30]. The membership function expresses the degree of an observed yield belonging to a certain LSE [

28]. The S-membership function assume that a high LSE—a highly suitable condition for a certain crop growth—would determine a high crop yield but with a limit, and vice versa.

The input and management of the same crop were assumed to be the same in each pixel, which was accepted for an ex ante evaluation. Then potential crop yields are determined by the LSE. We used S-membership functions to estimate potential crop yields based on the available field crop yield data. The S-membership functions connect the crop yield to the specific land suitability score (

Figure 4) [

28,

29,

30], with the equation as follows:

where

S means the membership value of crop yield to the land suitability score,

x is the crop yield, α and γ are the lower and upper limits of crop yields, and β is (α + γ)/2. The α and γ are the ideal limits of crop yields at local conditions, which characterized soil qualities, terrain factors, water supply conditions, climatic data, and socially locational factors with the assumption of the same input and management. They belong to the worst and best LSE values, respectively, and the membership values are 0 and 1, respectively.

S should be calculated from the standardized land suitability scores using Equation (6):

where

l is the land suitability score of a given crop at the pixel, and

l_{min} and

l_{max} are the lower and upper limits of land suitability scores in the whole study area.

The critical values (α and γ) of S-membership functions are usually difficult to determine and are always selected according to expert judgment and experience [

31]. From the household survey and statistical data, we can collect the highest, mean, and lowest crop yields (defined as

x_{h}, x_{m}, and

x_{l}, where the

x_{m} = (x_{h} + x_{l})/2

= β). The actual yields are determined by various factors based on the ideal LSE condition; therefore,

x_{l} and

x_{h} would not be up to the lower and upper limits but are close to α and γ. We proposed a parameter

p, which is the membership value of

x_{l} and close to 0 (for example, the default setting of

p is 0.1 in this case study). Then (1 –

p) is the membership for

x_{h}, which is close to 1. We can establish the following equation set:

Therefore,

α and

γ can be calculated from Equation (5). The crop yield can be calculated as follows:

**Figure 4.**
S-membership function to connect land use evaluation to crop yields.

**Figure 4.**
S-membership function to connect land use evaluation to crop yields.

#### 2.2.4. Ecosystem Service Value Estimation

The

e(

s) is interpreted as the expected ecosystem service values obtained from changing land use. However, ecosystem services include a considerable number of types [

32] and we do not need to estimate all of them in our case study. Compared to other crops, clover has larger ecosystem service values in soil conservation, namely preventing land degradation in the Yili region. Therefore, we estimated the decrease of soil erosion as the supply of ecosystem service values from changing other crops to clover. The Universal Soil Loss Equation (USLE) is the most popular method used for soil erosion modeling and assessment [

33] and was applied to quantify the amount of annual soil loss in the two situations described above. The equation of USLE is as follows:

where

A is the amount of average soil loss (ton·ha

^{−1}·a

^{−1});

R is the rainfall erosivity factor (MJ·mm·ha

^{−1}·h

^{−1}·a

^{−1});

K is the soil erodibility factor (ton·ha·h·ha

^{−1}·MJ

^{−1}·mm

^{−1});

L is the slope length factor;

S is the slope factor;

C is the vegetation cover factor; and

P is the erosion control practice factor. Factors

C and

P are dimensionless.

With different crops at the same site,

R,

K,

L, and

S are the same and

P is assumed to be the same for land use planning. Then

e(

s) (ton·ha

^{−1}·a

^{−1}) is represented as the change of

C factor from changing crops to clover (ton·ha

^{−1}·a

^{−1}). It can be calculated by the following equation:

where

C_{b} and

C_{a} represent the vegetation cover factor of clover and other crops, respectively.

As we are lacking field data for

C_{b} and

C_{a}, we use the remote sensing estimation, which has been widely used in China, to calculate them [

34]. The equations are as follows:

As mentioned above, the ecosystem service value of clover, which maintains vegetation through the whole year, is much larger than that of other land use types. Also, other crops have a similar growing season range of vegetation cover. We assumed that the

C_{a} of other crops is the same except for the

C_{a} of rice, which has a relative lower

C value. To calculate

C_{b} and

C_{a}, we chose some typical areas in the Yili region, are already planted with clover, rice, and other crops, as calculation samples. The NDVI values are the monthly data, which can reflect the temporal vegetation cover for different crops. We used the data on average NDVI values from 2000 to 2010, provided by Geospatial Data Cloud [

35]. Then the

C_{b} is 0.081, the

C_{a} of the rice is 0.180, and the

C_{a} of other crops is 0.405.

The spatial map of soil erosion (A

_{0}) without the implementation of land use planning in the study area was used as the current situation, which was calculated by Equation (10) and is provided by the Data Center [

36]. The

C is 0.5 and we defined it as

C_{0} with the grassland in our case study. Finally, the

e(s) can be calculated by the following equation:

#### 2.2.5. Procedure of Spatial Tradeoff Analysis

The net return of crops per pixel can be calculated by Equation (14):

where

r is the net return;

p is the product price (yuan/kg);

x is the crop yield (kg/ha) by the method in

Section 2.2.3; and

c is the cost of crop (yuan/ha).

The key characteristics of crops mentioned in the Yili region are shown in

Table 2, based on statistical data provided by the Yili Municipal Bureau of Statistics (2007) [

37] and consultation with local experts.

**Table 2.**
Key characteristics of crops in the Yili region.

**Table 2.**
Key characteristics of crops in the Yili region.
| Crop price (yuan/kg) | Crop cost (yuan/ha) | Mean crop yield (kg/ha) | High crop yield (kg/ha) | Low crop yield (kg/ha) |
---|

Wheat | 1.80 | 5209.8 | 5250 | 7500 | 3000 |

Corn | 1.30 | 5621.55 | 10,500 | 15,000 | 6000 |

Rice | 1.85 | 5545.35 | 8250 | 11,250 | 5250 |

Cotton | 12.00 | 6151.05 | 1500 | 2025 | 975 |

Sugar beet | 0.28 | 9925.05 | 54,000 | 75,000 | 33,000 |

Oil plant | 4.80 | 4640.55 | 2250 | 3000 | 1500 |

Bast fiber plant | 2.20 | 6083.1 | 4875 | 6750 | 3000 |

Vegetables and fruit | 0.77 | 25,041 | 67,500 | 75,000 | 60,000 |

Clover | 1.00 | 4500 | 7500 | 10,500 | 4500 |

With

p_{e} (yuan/ton) for

e(

s) (ton·ha

^{−1}·a

^{−1}) calculated by

Section 2.2.4, the opportunity costs between clover and other crop types can be calculated by Equation (4). To coordinate the issue of dimensions, we set a parameter

D = 1 a, then Equation (4) was rewritten as follows:

At each payment price, areas for the opportunity costs ω(r,s,p_{e}) from positive to negative can be identified in the spatial distribution. Opportunity costs from positive to negative mean that changing crops would lead to a larger income and a farmer would adopt the payment policy. The adoption rate, which means the proportion of farmers adopting the payment policy, ranges from 0 to 1, and can then be calculated to create the tradeoff curve. Next, these pixels are reallocated for clover, while other pixels remain unchanged based on the basic land use scenario. Finally, a land use reallocation map can be generated according to the specific payment price.