Land Use Suitability Assessment in Low-Slope Hilly Regions under the Impact of Urbanization in Yunnan , China

1Faculty of Resources and Environmental Science, Hubei University, Wuhan, Hubei 430062, China 2School of Public Administration, China University of Geosciences, Wuhan 430074, China 3State Key Laboratory of Water Environment Simulation, School of Environment, Beijing Normal University, Beijing 100875, China 4College of Geomatics, Shandong University of Science and Technology, No. 579 Qianwangang Road, Economic & Technical Development Zone, Qingdao, Shandong 266590, China 5Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Chaoyang District, Beijing 100101, China


Introduction
Due to the rapid urbanization in China [1], the demand for built-up land both in urban and rural areas has been dramatically increased with the shrinking of arable land [2][3][4].The conflict between land development and land conservation has become increasingly serious [5][6][7][8].Under the most stringent farmland protection system in China, the strategy of developing low-slope hilly regions is carried out to alleviate the conflict [9].The concept of low-slope hilly regions is put forward under the demand for land resources in mountain area with the social and economic development and urbanization construction.It refers to contiguously distributed land with slopes less than 25 degrees in the majority of hilly areas as well as unused low mountains and hills which can be used by town construction in the mountain area.It mainly includes a variety of reserved land types such as unused land, abandoned garden, and inefficient forest, which are mostly distributed in the regions with little plain and inadequate protection of cultivated land.Currently, there is a shortage of land available for the development of cities and towns in China [10,11].However, low-slope hilly regions account for a relatively large proportion of the available land.Current land use in China, such as a small amount of land per person, intense exploitation of land development, and shortage of reserved land, has limited the development of regional economy and society [12,13].Therefore, it is a good way to alleviate the conflict between land and people in the developed regions with high pressure of protecting arable land by optimally using low-slope hilly land [14].Compared to the plain areas, low-slope hilly land has the lower ecological carrying capacity and higher ecological sensitivity.Lack of good understanding of lowslope hilly land and objective evaluation of development activities will bring enormous environmental disturbance and destruction, resulting in serious or even irreversible consequences on the structure and function of ecosystems, biodiversity, and landscape in this area [15,16].Therefore, it is of great significance to do the suitability assessment in low-slope hilly land and classify development levels in order to ensure the orderly development of it [17][18][19].Since a framework for land evaluation reported by the Food and Agriculture Organization (FAO), the research of land use suitability assessment receives more and more attention.The theory, method, and specific scheme on land use suitability evaluation are also improved constantly.Currently, scholars usually carry out land suitability assessment in different regions [16,[20][21][22].However, the screening of evaluation factor and the determination method of factor weight are both significant on the influence of the cultivated land suitability assessment.On the one hand, the evaluation factor is different in different areas such as the plateau region, hilly area, and flood plain area, and the emphasis of the research is different.Accordingly, on the other hand, the calculation method of factor weight can be divided into two categories [12,23,24]: one is the mathematical logical reasoning based on knowledge and rules, containing the comprehensive index method and Fuzzy-AHP methods; this kind of methodology has higher dependency on knowledge and then makes the result more subjective.The other one is the data driven mechanism based on adaptive system such as neural network method.But its reasoning process is cumbersome and its algorithm is complex, which cannot effectively use existing knowledge and often leads to the fact that it is hard to explain the acquired rules.In view of the special natural and social economic conditions in Yunnan region, this paper determines the factor weights by the fuzzy weight of evidence model to avoid subjective influence, multiple collinear interference, and complex model algorithm.The paper provides a scientific basis for urbanization and optimal allocation of land through evaluating the development suitability levels of low-slope hilly resources (not including the ecological preservation areas) such as unused land, woodland, and grassland.

Study Area
Yunnan province (97 ∘ 31  ∼106 ∘ 11  W, 21 ∘ 8  ∼29 ∘ 15  N) is a lowlatitude inland, spreading about 394,000 square kilometers which accounts for 4.11 percent of total area in China (Figure 1) .It comprises mainly mountains, plateaus, hills, and small basins.Mountains and plateaus occupy around 94% of the province.Much of the province lies within the subtropical highland.There is little variance in annual temperature but a large diurnal temperature range.It has distinct wet and dry seasons.The temperature changes greatly with the terrain.Due to the interaction of climate, biology, geology, topography, and so on, various types of soil are formed.It is characterized by the vertical distribution of soil.The area of red soil accounts for 50% of the province.
Rainfall is unevenly distributed in terms of seasons and regions [25].Yunnan province, with strong intensity in urban land use, is one of the provinces where mountain towns are widespread in China.Therefore, it is an inevitable choice for urbanization development to moderately develop the lowslope hilly regions in Yunnan province.ArcGIS10.0 is used to process Kriging interpolation on the rainfall distribution and average temperature distributions with meteorological data, which results in creating an annual rainfall distribution map and an annual temperature distribution map.Topographic data are used to generate digital elevation model (DEM) (whose spatial resolution is 1 km × 1 km) and a slope map.Buffer zones of water sources and state roads are obtained later.Therefore, based on the optimum distance, a map of distance to water and a map of distance to state roads are obtained, respectively.Interpolation analysis is done through socio-economic data, which generates a map of GDP per capita and a population distribution map in Yunnan in the grid with cells of 1 × 1 square kilometers [26][27][28].The proportion of plain areas is obtained through land use data in Yunnan province [29].With the help of geoscience data analysis system, the existing built-up land parcels in Yunnan are extracted as training samples suitable for construction and are used to make a distribution map of the sample.It should be noted that GeoDAS4.2 was developed by the Geomatics Research Group at York University, and it is a GIS software that synthesizes the fuzzy weight of evidence model, fractal model, spatial statistics, artificial intelligence, and other modern data processing techniques (http://www.yorku.ca/yul/gazette/past/archive/2002/030602/current.htm).A sampling interval of 1 km × 1 km pixels is determined as the unit of area according to the minimum shape and size of the parcel of evaluation objects.Each map layer is saved in the form of GRID.

Research Methods.
Fuzzy weight of evidence is evolved from weight of evidence which is firstly used in medical diagnosis without considering the space.Since 1980s, Agterberg et al. [30,31] had modified this method and applied it Zhaotong on mineral forecast.Since this method can be applicable to integrate multiple information and spatial decision support systems, it has been applied in the evaluation on various mineral resources, geological disaster risk assessment, and environmental evaluation in recent years.

The Principle of the Method.
Weight of evidence method is a log-linear model under a Bayesian probability criterion.A priori probabilities are firstly calculated.Then conditional probabilities are calculated under a certain geological evidence model.Weight of evidence method includes posterior logit model, the general weight of evidence model, fuzzy weight of evidence model, and weighted weight of evidence model.The principle of the method is as follows [32][33][34].
(1) Calculate A Priori Probabilities.Suppose total area of the study area  is () which is divided into cells of a fixed area . is the event to be predicted.It follows that there are () = ()/ cells totally in the study area where ( ) represents the area and ( ) is the number of cells.Therefore, the frequency of events  in the study area  is ().The probability of the event is where () is the a priori probability, and the occurrence can be expressed as () = ()/(1 − ()).
(2) Calculate Weight of Evidence.Any weight of evidence (the map layer)  is binary.Its weight is defined as where  + and  − represent the weight for presence of  and weight for absence of , respectively.(/) = ( ∩ )/() denotes the probability of selected weight of evidence  in any unit cell of , where ( ∩ ) stands for the occurrence of weight in  and () is the total number of occurrence in the study area.Other equations can be explained in the same way.
(3) Calculate Posterior Probability.If there is  number of weights which are independent of each other on the occurrence, the probability of the occurrence of any cell in the study area can be expressed by the posterior probability logarithm which is expressed as follows: where  = (1, 2, 3, . . .), and  indicates the presence and absence of the weight, namely, where  +  is the weight for presence of weight . −  is the weight for absence of weight .0 is the weight for missing of weight .Accordingly, the posterior probability can be expressed as The The map layer 2 (weight of evidence) The map layer n (weight of evidence) The map layer 1 (weight of evidence) (4) Membership Function.Weight of evidence approach will result in loss of data while doing binary process for map layers which will affect the accuracy of assessment.Cheng and Agterberg [35] have developed a fuzzy weight of evidence approach which automatically copes with the missing data on discrete layers.Fuzzy weight of evidence uses ambiguity membership of layers which is determined by membership function to deal with multiclassification layers.Multivalued fuzzy membership function 0 ≤ () ≤ 1 is used to fit training samples.The weight of evidence is calculated in the end.suitable for urban development.The priori probability can be known by integrating (1) that the total number of unit cells of training samples is 2332 (Figure 3).There are 220,084 unit cells in the whole areas.The priori probability is 0.010596.

Fuzzy Map Layers and Weight of Evidence.
The selection of evaluation factors is under the principle of dissimilarity, stability, and reality.On the basis of previous research on land use suitability assessment in low-slope hilly regions and built-up land, the evaluation index system of land use suitability assessment in low-slope hilly regions mainly refers to climate, topography, geography, and socioeconomic conditions.Impact factors include temperature, rainfall, elevation, slope, proportion of plain area, distance to water, distance to state roads, population distribution, and GDP per capita (Figure 4).
Each fuzzy weight of evidence is calculated by Geo-DAS4.2.The influence of individual factors on the incident can be known which leads to the objective decision on choosing evaluation factors.Final map layers and parameters of fuzzy weights of evidence can be seen in Table 1 through (2) and multivalued fuzzy membership function.

The Posterior Probability and Its
Modification.These map layers are synthesized and the posterior probability map in Yunnan is calculated by (5).Then the posterior probability map of land use suitability in low-slope hilly regions in Yunnan is made by taking reserved land parcels in low-slope hilly regions as the crop box (Figure 5).Posterior probabilities are affected by the setting of the unit cell.However, they do not affect the distribution of posterior probabilities.Therefore, posterior probabilities do not represent the probability of occurrences but the distribution after occurrences.The map layer of fuzzy weight of evidence model is required to meet the needs of conditional independence.However, in reality, it is very difficult to achieve this which inevitably results in the deviation of posterior probabilities.The modified posterior probability can overcome the deficit of less accurate estimation on appropriate points.
Data tests of the linear function correction, logarithmic function correction, and exponential function correction are carried out with the help of GeoDAS4.2'sposterior probability correction module.Ultimately, exponential function model  = 0.12 0.96 with optimal fitting degree is determined.After the modification of the posterior probability, the

Classification of Land Use Suitability in Low-Slope Hilly
Regions.The posterior probability map is modified according to an exponential function model, which is used to create the modified posterior probability map.On the basis of this, the distribution frequency of modified posterior probabilities is analyzed.Distribution frequency curve is used.Finally, land use suitability in low-slope hilly regions in Yunnan province is divided into four levels (Table 2, Figure 8) according to obvious inflection point of the frequency curve and parameters of the weight of evidence of those evaluation factors.
There are four levels in terms of land use suitability in low-slope hilly regions in the study area: highly suitable, moderately suitable, marginally suitable, and unsuitable.
(1) 9.33% of the land is highly suitable for development, mostly distributed in Kunming, Lijiang, Dali Bai autonomous prefecture, Qujing, and so on.Spatially, highly suitable areas are mainly located around cities where its terrain is relatively flat with slopes below 5 degrees and altitude mostly below 1500 m.The plain area accounts for more than 15% of the total area.Usually, the distance to water and state roads is no more than 15 km.GDP per capita is higher than 20,000 yuan.It is densely populated with 1,500 people/km  In addition, comparing the results of this paper with the achievement of the outcome of overall land use plan, it can be known that, as far as quantity structure is concerned, it is quite scientific and reasonable that the area of highly suitable resources of low-slope hilly regions is much larger than the built-up land index of the planning phase (2006-2020); as for the spatial distribution, the layout of the built-up land in the planning phase is basically consistent with the highly suitable low-slope hilly regions or is distributed in its perimeter zone.

Conclusion and Discussion
Due to global climate change and urbanization in China, the conflict between land development and land conservation has become increasingly serious [36].

Figure 1 :
Figure 1: Location and DEM of Yunnan province.
Estimate the number of suitable pointsSet up the evaluation parameters

Figure 2 :
Figure 2: Flowchart of fuzzy weight of evidence.

.
The procedures of fuzzy weight of evidence (Figure2) are as follows.(1) Identify research objectives, such as suitability of the development of the low-slope hilly regions; select the samples and calculate the a priori probabilities.(2) Determine the spatial layer related to the target; screen indicators; and establish an index system.(3) Extract map layers related to the target; use fuzzy membership functions to represent the credibility of the map layer.(4) Calculate the weight of fuzzy map layer; determine the value of the map layer; and screen indicators.(5) Integrate multiple fuzzy map layers and calculate a priori probabilities in order to obtain a suitability map in low-slope hilly regions.(6) Modify posterior probabilities; test the model; and make an interpretation for suitability distribution.

Figure 7 :Figure 8 :
Figure 7: Modified posterior probabilities of land use suitability and numbers of corresponding points.

Table 1 :
Parameters of fuzzy weights of evidence for each map layer.

Table 2 :
Classification of land use suitability in low-slope hilly regions in Yunnan.51% of land is unsuitable for development.These areas are more dispersed with slopes above 15 degrees in general and altitude mostly above 2200 m.The proportion of plain area is less than 5% of the total area.It is far away from water resources and state roads.GDP per capita is low and population density is relatively small.It is far away from cities and towns.It is not suitable for living due to the relatively unpleasant weather.At the same time, it is prone to geological disasters and ecological risks, inappropriate for development and construction.In summary, the four levels of the suitability assessment in low-slope hilly regions in Yunnan province can provide a basis for the priority of development in these regions.Under the consideration of social-economic development, urbanization, and ecological safety, the paper offers strategies to develop low-slope hilly regions, which focus on the exploration of unused land and give preferential protection to woodland and grassland.
The development of lowslope hilly regions is an important measure to alleviate the conflict.This paper takes Yunnan province as a typical case study, integrating natural geographical features and elements of human geography in Yunnan province to analyze land use suitability of low-slope hilly regions in Yunnan province and avoid potential geological disasters and ecological risks in those regions.Land use suitability of low-slope hilly regions in Yunnan province is evaluated by analyzing some factors referring to climate, topography, geography, society, and economy.Nine map layers regarding temperature, rainfall, elevation, slope, proportion of plain areas, distance to water, distance to state roads, population distribution, and GDP per capita are selected.The results show that (1) 9.33% of lowslope hilly regions in Yunnan are highly suitable for development.26.18% of land is moderately suitable.45.98% of land is generally suitable.The remaining 18.51% of land is unsuitable for development.This outcome of spatial distribution of lowslope hilly land in Yunnan province can finally optimize the size and layout of low-slope hilly land and promote the rational use of it.However, it is worth mentioning that not only the suitability of natural quality should be considered, but also some factors such as social and economic suitability, ecological suitability, and land policy should be taken into consideration.The related research will provide references for the plan of "town of mountain" in Yunnan and for the development of low-slope hilly land in Yunnan province.As for the model used in this study, it should be pointed out that training samples chosen in the study must be highly recognized appropriate points to ensure scientific results.Therefore, future research should pay more attention to how to ensure diversification of the given sources and improve the accuracy of given information.