Based on the indicator systems of previous studies (Saleth and Swaminathan, 1993; Singh and hiramath, 2010; Wang et al., 2016; You and Zhang, 2017) and the local situation of Poyang Lake Eco-Economic Zone, this study establishes an adaptive index system that is classified into four layers: Goal layer, sub-goal layer, criterion layer and basic indicator layer (Table 2).

From the perspective of ecology, the Ecological Security Index (ESI) is synthetically reflected by the indicators at two levels: ecological quality and ecological pressure. Among them, forest and waters, known as “the lung of the earth” and “the kidney of the earth”, are the basic indicators to reflect the ecological background. In addition, in view of the severe level of water quality deterioration in Poyang Lake and the positive significance of ecological red line demarcation for ecological protection, the environmental quality of surface water and the area ratio of the ecological protection red line are also taken as important ecological background indicators. Ecological pressure can be divided into two parts: land area and water area of Poyang Lake. The pressure on the water area mainly comes from fishing and breeding of fishery. The pressure on the land is mainly manifested in the ratio of rural population to farmland and the excessive application of chemical fertilizer. Therefore, the ecological pressure of each county (district) is comprehensively evaluated by fishery population, fishery breeding area, population-farmland ratio and chemical fertilizer application intensity.

From the perspective of economy, the Economic Efficiency Index (EEI) is comprehensively reflected by the indicators at two levels: production level and consumption level. Among them, the total output value of farming, forestry, animal husbandry and fishery and the total power of agricultural machinery represent the levels of rural productivity and mechanization, respectively. In addition, in view of the serious problem of wasted cultivation in rural areas, the multiple cropping index (MCI) of cultivated land is introduced to check the regional production security. Rural per capita disposable income, rural per capita electricity consumption and farmers' per capita living consumption expenditure are respectively used to evaluate the potential consumption capacity and the actual consumption level in rural areas.

From the perspective of society, the Social Equality Index (SEI) is comprehensively reflected by the indicators of rural-urban equilibrium and social security. The rural-urban gap is first shown in the income disparity between urban and rural residents. Urbanization and labor migration are important factors influencing the rural-urban gap. Therefore, rural-urban income disparity, urbanization rate and proportion of rural migrant laborer are selected to measure the rural-urban gap. Social security mainly embodies the provisions of education, health, and pensions, etc., so this study selects three indicators to quantify the pressure and level of social security, namely, the dependency ratio of the rural population, the number of beds per thousand people in hospitals, and the faculty-student ratio of basic education.

Entropy method is an objective evaluation method of index weighting based on possibility theory, which effectively avoids the influence of human subjectivity, and the evaluation results are more rigorous and accurate (Chu et al., 2015; Cui et al., 2018). The Analytic Hierarchy Process (AHP) is simple and easy to operate, and it is also widely used in subjective weighting of evaluation indexes (Aminbakhsh et al., 2013; Stefanidis and Stathis, 2013). Due to the multi- level nature of the SLS index system, this study uses the approach of combining AHP and entropy method to weight the indicators, that is, the sub-goal level and the criterion level are weighted by AHP, the index layer is weighted by the entropy method, and each index weight is finally obtained based on Weighted Processing. Let the influence weight of the sub-goal layer on the goal layer be w°, the influence weight of the criterion layer on the sub-goal layer be w”, and the influence weight of the index layer on the criterion layer be w‴, then the final weighted weight of each index is:

As the Analytic Hierarchy Process is relatively simple, only the entropy method is described here in detail. The measurement units of different indicators can indeed be different. In order to facilitate the comparison of different indicators, the entropy method requires the standardization of each index. In this study, we adopted the standardization of data ranges, and the expression is as follows:

where, ${{{x}'}_{ij}}$and ${{x}_{ij}}$ are the standardized and observed values of j sample of index i, respectively; $\underset{1\le j\le n}{\mathop{\text{m}ax}}\,{{x}_{ij}}$and $\underset{1\le j\le n}{\mathop{\text{m}in}}\,{{x}_{ij}}$are the maximum and minimum original values of index i, respectively; ${{I}^{+}}$and ${{I}^{}}$ are the index type, representing a positive index and negative index, respectively, and n is the number of observations.

The proportion ${{P}_{ij}}$of index value of sample j of i index is as follows:

The information entropy e_i of index i is as follows:

Then the entropy weight (${w}'''$) of the index i is:

where, m is the number of indicators, n is the number of observations, $\text{i}\in [1,20]$, $j\in [1,34]$, $0\le {{{w}'''}_{i}}\le 1$, $\sum\limits_{i=1}^{m}{{{{{w}'''}}_{i}}}=1$. According to the AHP and expert scoring, the weights of each sub-goal layer to the total goal layer and the criterion layer to the sub-goal level are equal, that is, the weights of ecological security, economic efficiency and social equity to rural SLS are 0.3333, and the weight of each criterion layer is 0.5. The weight of each index layer to a different criterion layer is determined by the entropy values of the 34 counties (districts) (Table 2). Finally, the weighted processing of each index's comprehensive matrix is W= [0.0318, 0.0567, 0.0152, 0.0630, 0.0445, 0.0536, 0.0338, 0.0348, 0.0637, 0.0810, 0.0220, 0.0347, 0.1137, 0.0183, 0.0512, 0.0856, 0.0300, 0.0230, 0.0800, 0.0637].

Let the SLS have m (m = 20) evaluation indexes to form the index set $U=\{{{u}_{1}},{{u}_{2}},\cdots,{{u}_{m}}\}$; and let p (p = 5) evaluation levels: low level, relatively low level, medium level, relatively high level and high level, constitute the appraisal set $V=\{{{v}_{1}},{{v}_{2}},\cdots,{{v}_{p}}\}$. The index membership degree model adopts the low half echelon school function, then the positive index membership function expression is:

And the negative index membership function expression is:

In the formula, x⁺ and x^‒ represent the actual index values of positive and negative indicators, respectively; ${{u}_{i}}{{v}_{k}}$represent the membership degrees of the k^th level of the i index; and x_k represent the thresholds of corresponding grades, 2≤k≤4. Therefore, the fuzzy relation matrix can be expressed as follows:

The final SLS fuzzy comprehensive evaluation result, matrix A, can be expressed as:

In order to further reveal the spatial pattern and agglomeration characteristics of rural SLS in Poyang Lake Ecological Economic Zone, this study uses the Moran's I index of global spatial autocorrelation and local Getis-Ord $G_{i}^{*}$index to identify the overall spatial association and local cold-hot zones of SLS, and visualize them based on the ArcGIS platform. Among them, the formula for the spatial autocorrelation of Moran's I index is:

The Getis-Ord $G_{i}^{*}$index of hotspot analysis can be expressed as:

where, n is the number of evaluation units, w_ij is the spatial weight between evaluation units i and j, x_i and x_j are the SLS values of evaluation unit i and unit j respectively, and $\bar{x}$is the average value of SLS of all units. G is the significant statistic in essence. A high G value indicates similar high value spatial aggregation, namely a hot spot area, while a low G value indicates similar low value spatial aggregation, namely a cold spot zone.

Geographical detector analyzes the possible relationships between geographical variables by testing the spatial distribution consistency of the geographical variables (factor variables and outcome variables). It is often used to detect the dominant factors of the spatial differentiation of geographical elements (Wang et al., 2010; Liu and Li, 2017a). Through geographical detector, this study analyzes the impact of each index variable on the spatial differentiation of rural SLS, and determines the leading factors affecting the livelihood security in Poyang Lake Ecological Economic Zone. Based on the spatial differentiation theory and GIS spatial overlay technology, various kinds of geographical variables and outcome variables are classified according to the natural break method, and normalized under the same spatial scale. Then the power determinant value of each factor on the spatial differentiation of outcome variables is measured by q (Wang and Xu, 2017):

where, q is the power determinant value of each factor variable on the outcome variable, and the value is in [0, 1]. The greater the value, the greater the influence of the geographical factor on the spatial differentiation of the outcome variable. When q=1, it indicates that the spatial differentiation of the outcome variable is determined by the geographical factor; when q=0, it indicates that the outcome variable is randomly distributed, and the geographical factor has no influence on the outcome variable. Here, Y is the outcome variable, which refers to the SLS index and h is the index classification. In this paper, the geographical factor indicators are divided into L=5 categories by the natural break method (i.e., the factor variables are divided into five categories: low, relatively low, medium, relatively high and high). n and ${{n}_{h}}$ are the number of evaluation units in the whole study area and the number of evaluation units of type h, respectively. x ${{\delta }^{2}}$.and $\delta _{h}^{2}$are the global discrete variance of the outcome variable and the discrete variance of the h ^th type, respectively.

In order to explore the obstacle factors restricting the SLS in rural areas, this study introduces the obstacle degree model to makes a pathological diagnosis on the factors hindering SLS. The main indexes involved are “deviation degree” index and “obstacle degree” index, and their expressions are as follows:

where, ${{s}_{i}}$ represents the deviation degree of the i-th index, that is, the gap between the actual value and the target value of the index; ${{{x}'}_{i}}$ represents the standardized value of the i-th index; ${{y}_{i}}$ refers to the obstacle degree index; ${{w}_{i}}$ refers to the weight of the i-th index, and m refers to the number of evaluation indicators.

On the basis of analyzing the restriction level of each individual index evaluation factor, the barrier degree of each criterion layer and sub-goal to SLS is further determined, and their expressions are as follows:

where, ${{Y}_{j}}$represents the barrier degree of the j^th criterion layer or sub-goal layer; and ${{y}_{ij}}$ refers to the barrier degree of the i^th indicator of j^th criterion layer or sub-goal layer.

Using the Natural Break method, the indicators are divided into five levels: low level I, relatively low level II, medium level III, relatively high level IV and high level V. According to the comprehensive fuzzy calculation of index grading and corresponding weights, the evaluation results of rural SLS in Poyang Lake Ecological Economic Zone are shown in Fig. 2 and Fig. 3. Among the 34 counties (districts), 2 counties (Changjiang District and Fujiang County) have rural livelihood levels that are highly secure, accounting for 5.89%; 9 counties are relatively safe, accounting for 26.47%; 7 counties are moderately safe, accounting for 20.59%; 12 counties are relatively low safe, accounting for 35.29%; and 4 of them (Zhangshu City, Xingan County, Guixi City and Wannian County) have low safety, accounting for 11.76%. These data show that the overall SLS in Poyang Lake Ecological Economic Zone is low, so there is still room for improvement.

For the sub-goal layer the distribution of counties in terms of ecological security was: high (5 counties, 14.71%), relatively high (5, 14.71%), medium (6, 17.65%), relatively low (11, 32.35%) and low (7, 20.59%). The distribution of counties in terms of economic efficiency was: high (2 counties, 5.88%), relatively high (8, 23.53%), medium (11, 32.35%), relatively low (8, 23.53%) and low (5, 14.71%). The distribution of counties in terms of social equity was: high (4 counties, 11.76%), relatively high (7, 20.59%), medium (9, 26.47%), relatively low (9, 26.47%) and low (5, 14.71%). Therefore, the rural EEI and SEI of Poyang Lake Ecological Economic Zone are in the middle level overall, but the ESI is generally low.

For the criterion layer the distribution of counties in terms of ecological quality was: high (5 counties, 14.71%), relatively high (7, 20.59%), medium (8, 23.53%), relatively low (7, 20.59%) and low (7, 20.59%). The distribution of counties in terms of ecological pressure was: high (8 counties, 23.53%), relatively high (5, 14.71%), medium (6, 17.65%), relatively low (8, 23.53%) and low (7, 20.59%). The distribution of counties in terms of production level was: high (2 counties, 5.88%), relatively high (7, 20.59%), medium (10, 29.41%), relatively low (9, 26.47%) and low (6, 17.65%). The distribution of counties in terms of consumption level was: high (1 counties, 2.94%), relatively high (2, 5.88%), medium (7, 20.59%), relatively low (19, 55.88%) and low (5, 17.41%). The distribution of counties in terms of rural-urban equilibrium was: high (2 counties, 5.88%), relatively high (4, 11.76%), medium (5, 14.71%), relatively low (17, 50.00%) and low (6, 17.65%). The distribution of counties in terms of social security was: High (1 counties, 2.94%), relatively high (7, 20.59%), medium (14, 41.18%), relatively low (6, 17.65%) and low (6, 17.65%). It can be seen that the safety levels of the eight criteria layers are mostly middle or slightly low, while the consumption level and the rural-urban equilibrium in particular are very low, so they urgently need to be improved and optimized.

Through spatial autocorrelation analysis, the Moran's I index values of SLS and the related sub-indexes were calculated. The results are shown in Table 3. Except for the production level and SEI, the spatial Moran's I value of each index is positive, and they are significant at the 95% confidence level. These results also reflect that the rural SLS index and the sub target layer and criterion layer index of Poyang Lake Ecological Economic Zone are spatially auto-correlated positively, and the spatial aggregation features are quite significant.

Through hot spot analysis (Getis-Ord $G_{i}^{*}$), we find that the spatial distribution of SLS has a tendency of being higher in the north and lower in the south (see Fig. 2 and Fig. 3). The hot zones are centralized in Changjiang District, Fuliang County, Poyang County and Leping City in the northeast of Poyang Lake; while the cold zones are centralized in Fengcheng City, Zhangshu City, Yushui District, Xingan County which are in the southwest, and Yujiang County and Guixi City which are in the southeast. The spatial distribution of ESI has a more obvious tendency of being higher in the north and lower in the south. The hot zones are mainly centralized in Wuning County, Yongxiu County and Duchang County in the northwest of Poyang Lake, while the cold areas are centralized in Nanchang County, Jinxian County, Fengcheng City, Dongxiang County, Linchuan District and Zhangshu City in the southern Poyang Lake. The ESI hot zone mainly benefits from higher forest coverage and ecological red line protection area, while the cold zone is limited by low forest coverage, ecological red line protection area or wide fishery culture area and large fishery population. The EEI is in a circular distribution, in which it is distributed around the Nanchang metropolitan area from high to low. The hot zones are concentrated in Nanchang County, Jinxian County, Xinjian District, Donghu District, Fengcheng City and Yugan County, while the cold zones are centralized in Lushan city and Chaisang District. The difference of EEI between cold and hot zones mainly stems from economic location, paticularly as it is manifested in the regional differences of agricultural machinery gross power and rural per capita electricity consumption. The SEI distribution is just contrary to EEI. It is in a circular distribution, in which it is distributed around the Nanchang metropolitan area from low to high. The peripheral area of Nanchang city is relatively low, while the surrounding areas such as Lushan City, Chaisang District, Hukou County, Linchuan District, Fuliang County and Changjiang District are higher. The hottest zones are centralized in Fuliang County and Changjiang District, while the coldest area is in Yugan County. The indicator values show that the cold and hot areas of SEI are quite different in urbanization rates, rural-urban income disparities and rural population dependency ratios.

The ecological quality is better in northern mountainous regions (like Lushan City and Yongxiu County) and counties around the Poyang Lake (such as Duchang County, Hukou County and Lianxi District) which are characterized by higher forest coverage, and larger water area and ecological red line area. However, the ecological quality is relatively poor in the southern area of Ganfu Plain, including Zhangshu City, Gao'an City, Yushui District and Xingan County. The areas with high ecological pressure are mainly concentrated in suburban areas with high populations, such as Nanchang County, Jinxian County, Yugan County, or areas with large fishery populations and wide culture areas, like Dongxiang County and Pengze County. The production level and consumption level are both circularly distributed around the Nanchang metropolitan area from high to low. The hot zone of the production level is concentrated in Fengcheng, Jinxian, Yugan and Dongxiang counties around Nanchang city, while the cold zone is in Chaisang District. The hot zone of the consumption level is mainly concentrated in related districts of Nanchang City, while Yujiang County is the significant cold area. Most counties in the Poyang Lake Ecological Economic Zone are relatively low in the rural consumption aspect. The development gap between urban and rural areas appears to be large, with a low rural-urban equilibrium index. Only Donghu District, Fuliang County and Changjiang District have developed mutually between their rural and urban areas. As to social security, there exists an unexpected phenomenon that Nanchang metropolitan area, as the cold zone, is significantly lower than the surrounding areas. This is because urbanization has caused the rural population to migrate to the suburban areas of Nanchang, which has brought about high population density and reduced the average access to social security, such as educational resources and medical care.

In order to further verify the dominant factors affecting SLS spatial differentiation, this study calculates the power determinant value of each index on SLS through the geographic detector model. As shown in Fig. 4, rural per capita electricity consumption (0.50), fishery breeding area (0.46), rural population dependency ratio (0.44), urbanization rate (0.41), fishery population (0.35), population-farmland ratio (0.29), and rural-urban income disparity (0.26) are the top seven factors influencing rural SLS.

Rural per capita electricity consumption is a key indicator reflecting farmers' consumption level and rural economic efficiency, and it plays a decisive role in SLS spatial differentiation. As shown in Fig. 5a, the spatial distribution of rural per capita electricity consumption is highly consistent with SLS, as it also shows the distribution of high in the north and low in the south. The values of rural per capita electricity consumption are higher in Fuliang County, Changjiang District, Poyang County, Leping City, Duchang County, Pengze County, Lushan City and Jing'an County, while lower in Xingan County, Guixi City, Yugan County, Dongxiang County and Linchuan District. The significant hot zone of rural per capita power consumption is concentrated in Qingshanhu District, Nanchang County and Wanli District due to their developed economic and infrastructure conditions.

Fishery breeding area and fishery population also play important roles in shaping the spatial pattern of SLS, with q values of 0.46 and 0.35, respectively. This is due to the fact that fishery activities, such as overfishing and enclosure breeding, have resulted in serious water pollution and great pressure on the fish ecosystem in Poyang Lake. As fishery breeding area and fishery population are both negative indexes, their spatial patterns are opposite to that of SLS. It can be seen from Fig. 5b and Fig. 5e that the cold zone of SLS just corresponds to the hot zones of fishery breeding area and fishery population.

As a negative index, the rural population dependency ratio has a decisiveness of 0.44 to SLS. The population dependency ratio reflects the pressure of the aging population and supporting offspring, and a higher dependency ratio often requires more investments in social security. It can be seen from Fig. 5c that the high and low value areas of rural population dependency ratio correspond to the low and high value areas of SLS.

The spatial pattern of urbanization rate is quite similar to that of SLS (Fig. 5d), nearly sharing the same districts of high and low values. While the spatial patterns of population-farmland ratio and rural-urban income disparity are opposite to that of SLS, this phenomenon fully proves that non-agricultural transition and urbanization are effective ways to reduce the rural-urban gap and improve rural SLS. It is also noteworthy that in the developed and urbanized areas, such as Yuehu, Wanli, Qingshanhu and Donghu, the values of SLS are quite higher despite the high population-farmland ratio and rural-urban income disparity. This is because the farmers in these districts are more fully employed in non-agricultural sectors and rely much less on farmland. Although the rural-urban income disparity is greater, farmers in these urban areas tend to have relatively abundant income and higher SLS compared with farmers in other areas.

Through the analysis by the obstacle degree model, the obstacle degree is calculated for each subsystem, criterion and indicator of SLS. As shown in Table 4, from the indicator level, the top seven obstacles (and their mean obstacle degree values) are: rural per capita electricity consumption (0.16), number of beds per thousand people in hospitals (0.11), urbanization rate (0.10), the total power of agricultural machinery (0.09), the total output value of farming, forestry, animal husbandry and fishery (0.09), the area ratio of ecological protection red line (0.07), and the rural-urban income disparity (0.06). Among the total of 34 districts, more than 30 districts have values of these above seven indicators which are all greater than 0.03. This means that these seven indicators are the main limitations restricting rural SLS, and have become the common problems in Poyang Lake Ecological Economic Zone.

From the perspective of the criterion level, the obstacle degrees of SLS criteria are successively 0.21 (consumption level), 0.20 (production level), 0.18 (rural-urban equilibrium), 0.18 (social security degree), 0.16 (ecological quality) and 0.08 (ecological pressure). From the sub-goal level, the obstacle degree of SLS in Poyang Lake eco-economic zone is mainly reflected in economic efficiency (0.41), social equity (0.36), and ecological security (0.23).

Therefore, in the Poyang Lake Eco-Economic Zone, countermeasures should focus firstly on activating rural economic efficiency, especially increasing the investment in infrastructure such as rural power, popularizing rural mechanization, and increasing the total output value of farming, forestry, animal husbandry and fishery. Secondly, we should pay attention to rural social security such as medical care and education, accelerate urbanization and non-agricultural transition of farmers, and narrow the income gap between urban and rural residents. In addition, the implementation of the ecological protection red line is of great significance for improving the regional ecological security and ensuring the safety of rural sustainable livelihoods.

Through the case study, the LSA framework is shown to be effective, so it can not only provide efficient methods and adaptive indicators, but also guide the research procedures, and help with problem analysis. This study proposes five strengths of the LSA framework. 1) The LSA framework has constructed a four-layer indicator system which will be comprehensive and flexible for assessing SLS in the case analysis. It innovates the SLS index by incorporating three sub-goal indexes (i.e., ecological security, economic efficiency and social equity), eight criterion-level indexes (i.e., ecological quality, ecological stress, production level, consumption level, rural-urban equilibrium and social security), and 20 new indicators. With the structure of a fixed sub-goal layer and criterion layer, and a flexible indicator layer, this semi-fixed indicator system turns out to be efficient and easy to operationalize. 2) The LSA framework provides efficient and applicable methods for analyzing, visualizing, and interpreting the results. The AHP-Entropy combined weighting method is more objective and accurate than each individual method, and the fuzzy comprehensive evaluation method is widely accepted for its simplicity and efficiency in index evaluation. The spatial auto-correlation model based on a GIS platform helps to visualize the spatial patterns of rural SLS and the local Getis-Ord $G_{i}^{*}$index identifies both the hot spots and the cold spots. 3) This framework has the benefits of identifying key factors, main constraints, and specific pathways for building livelihood security. The geographic detector model is employed to detect the dominant factors driving the spatial differentiation of SLS, while the obstacle degree model is constructed for clarifying the main obstacles of SLS. Thus, we can determine corresponding measures to improve livelihood security. 4) This framework focuses on the district level, which is in favour of macro-control and regional balance in rural SLS. 5) Lastly, the LSA framework is scientifically designed and tightly connected with the National Rural Revitalization and the Fishing Ban, which make it usable and reasonable in the implementation of the current policy programme. The indicators in this study have been selected carefully to embody the content of the National Rural Revitalization and Fishing Ban, and have to be adapted to the specific situation of the Poyang Lake Eco-economic Zone. As a result, the information on rural recession is mainly reflected by the economic efficiency and social equality indicators, such as the multiple cropping index, the rural per capita electricity consumption, the rural-urban income disparity, the rural population dependency ratio, etc., while the Fishing Ban content is mainly contained in the ecological security indicators, such as the fishery population, the fishery breeding area, the ecological protection red line area ratio, etc. Overall, the LSA framework has the potential to help the government identify specific interventions to improve livelihood security and rural sustainability. This framework can be also applied in other similar areas by the merit of its simplicity and flexibility.

Although using the LSA framework helps the in-depth analysis of livelihood security, there are also some drawbacks and critiques of this framework. First, the LSA framework is mainly applicable at the district level and useful for macro analysis on a community scale, county scale and even larger scales. However, it cannot identify the livelihood characteristics of individual farmers, and it may be difficult to incorporate household-scale factors into the security measurement. Second, it only provides a “snapshot” of livelihood security, and not necessarily a dynamic measure of how livelihood security is changing. If the goal was to look at how livelihood security has changed over time, a dataset focused on the indicators of security at two different points in time would be needed.

Under the background of the National Rural Revitalization and Fishing Ban in the Yangtze River Basin of China, rural SLS analysis can provide a valuable perspective for the realization of the goal of rural revitalization and environmental protection of Yangtze River. The SLS in rural areas requires the coordination of the three dimensions of ecological security, economic efficiency and social equity, which are different in approach but equally important for achieving the five dimensions of rural revitalization: “Industrial prosperity, livable ecology, rural civilization, effective governance, and affluent life” (Liao and Chen, 2017; Ma et al., 2018). The smooth implementation of the Fishing Ban project also requires the premise of guaranteeing the sustainable livelihood security in rural areas and compatible incentives for fishermen (Pang and Jin, 2020). The case study of Poyang Lake eco-economic zone shows that rural per capita electricity consumption, urbanization rate and rural-urban income disparity respectively embody rural economic efficiency and social equity and are all in the top seven indicators in terms of power of determinant and obstacle degree. They are not only the dominant factors affecting the spatial differentiation of rural SLS, but also the main obstacles restricting the construction of rural SLS. This implies that National Rural Revitalization and Fishing Ban should focus on the rural-urban integration rather than merely on rural areas (Wang, 2016; Liu, 2018), and pay more attention to improving rural sustainable livelihood security and promoting regional balanced development (Tian et al., 2016a, 2016b; Liu and Liu, 2016; Wang, 2018).

Specifically, local governments should take actions from three aspects. 1) In the ecological subsystem, the government should make a reasonable and equilibrated adjustments for the ecological red line area to avoid a red line ratio that is too high or too low, and it especially needs to highlight the joint efforts with national nature reserves and national forest parks (Kong et al., 2013). At the same time, abiding by the Fishing Ban makes it necessary to take various measures to guide the transformation of fishermen's livelihoods, gradually reduce or even eliminate overfishing and aquaculture in lakes, and restrict the area of fishery breeding. More attention should be paid to those areas with large fishery populations and breeding areas, such as Nanchang, Jinxian, and Yujiang. 2) In the economic subsystem, it is urgent to activate the rural economy by reinforcing rural power infrastructure, promoting agricultural mechanization, and increasing the total output value of farming, forestry, animal husbandry and fishery. Specifically, agricultural mechanization services in most of Jiujiang City, including Chaisang, Lianxi, Xunyang, Hukou, Lushan, and Gongqingcheng City, DeAn County, cannot meet the needs of the large-scale agriculture development. Rural power infrastructure in Linchuan, Dongxiang, Yugan, Chaisang, Guixi and Xingan has also lagged behind and should be reinforced immediately. 3) In the social subsystem, the aging problem is prominent and social security in rural areas, such as medical treatment and pensions, needs to be improved. Meanwhile, the policy on urbanization should be adjusted and strengthened to accelerate farmers' non-agricultural transition and narrow the rural-urban income gap. The areas with the lowest SEI value, such as Xinjian, Wanli, Nanchang, Guixi and Xingan, have become key areas of social governance.