Integrating LCA and network model to understand food-energy-water nexus in the Ningxia, China

. It is necessary to link and feedback the FEW-Nexus in order to find a sustainable method of food, energy, and water (FEW) supply. We better understand nexus by integrating LCA and network models based on previous work in this paper. The results show that (1) 1.16 billion kg of grain were traded in Ningxia in 2020. The grain export area is primarily in Ningxia's central and southern regions, and grain trade relieves the pressure on grain demand in the north. (2) Grain trade has resulted in the movement and reconfiguration of water footprints and carbon footprints, alleviating some of the pressure on local water resources. However, some counties have low agricultural resource utilization efficiency. For example, Xiji failed to use water effectively in water-rich areas to meet the needs of water-scarce areas. (3) The spatial association network of the FEW-Nexus is unbalanced, and the associated channels should be enriched. Ningxia can improve the utilization of energy and water resources by strengthening key nodes of food flow.


Introduction
Food production is a complex system that includes water and energy consumption, greenhouse gas (GHG) emissions, and other resource and environmental issues [1] .The term food-energy-water (FEW) nexus has gotten a lot of attention because it refers to the direct or indirect link between the food, energy, and water sectors.It aims to look for integrated solutions to resource shortages while avoiding the pitfalls of silo approaches [2] .A number of framework papers demonstrate the complexities of quantitative FEW-Nexus relationship research [3][4][5] .In 2014, the FAO examined how the FEW-Nexus approach can be used to ensure food security and sustainable agricultural development.At the same time, the quantitative research of FEW-Nexus has made some achievements, as well as some innovations and breakthroughs in research methodology [6] .
Previous research has used a variety of systems analysis methods and modeling to identify interactions between FEW systems, frequently focusing on a single or two factors [7] .Vora presents a food trade network model that estimates the embodied energy and greenhouse gas (GHG) emissions in food virtual water trade in the United States [8] .Based on the network model, Mahjabin built the US food and energy flow network and quantified the implicit virtual water in food flow and energy transfer [9] .In the past few decades, a Life Cycle Assessment (LCA) has become the main methodological framework for assessing the environmental impact of consumer goods.Harun et al. conducted an assessment for the environmental sustainability of rice production in Malaysia based on LCA [10] .Li focus on the effects of direct and indirect resource consumption with a life-cycle approach [7] .Most LCA-related studies, however, simply focus on food and rarely take into account the perspective of food trade integration networks.Food trade networks encompass a wide range of resource (energy, water, etc.) and emission streams.Quantifying the sources and destinations of food flows, as well as the associated resource-specific flows, is critical for determining the FEW-sustainability Nexus's and resilience.
Ningxia is located in arid and semi-arid regions, agriculture has always been a major water user, and the large differences in water endowment between north and south are significant, as are food production efficiency differences.To provide a quantifiable understanding of the FEW-Nexus in food production and trade, this paper employs the life-cycle approach (LCA) and network model, which will help clarify sustainability challenges and contribute to the development of policies that promote conservation and efficiency in the FEW-Nexus.

Data sources
The data used in this paper include meteorological data, agricultural data, population data, and energy data.meteorological data (precipitation, temperature, wind speed, relative humidity, and sunshine hours, etc.) were from China Meteorological information center (https://data.cma.cn/).Agricultural data, including grain

Water footprint (WF) estimation
The water footprint of food production refers to the amount of water used to produce a unit of food in a region over a specific time period (generally one year), including blue water, green water, and gray water footprints, which can be calculated as follows (the gray water footprint is not considered): blue green Where, c ET reference evapotranspiration of water per unit area of crop (mm), Peff is the effective rainfall (mm), A is the Area under crops (ha).

Carbon footprint (CF) estimation
The carbon footprint is used as a measure of the greenhouse gas emissions directly or indirectly caused by a product according to LCA, which can be calculated as follow: Where, i I represents the amount of an agricultural resource input (including fertilizers, pesticide, agricultural plastic film, electrical power, etc.), i EF is the emission factor [11] .

Network model construction
This paper assumes that all food production can be consumed in a single year and that export crops are associated with crop surplus areas while imported crops are associated with crop shortage areas.As a result, the net transfer amount of crop is calculated as follows: Where is the net transfer value of crop i in area j, a negative value means import while a positive value means grain export, ton; , T are the crop i production of area j and the province, respectively, ton; j P and C P are the population of are j and the province population, mean value.
Then, we convert the food transfer data into a network by creating an adjacency matrix.In order to facilitate network analysis, the spatial transfer matrix needs to be binarized.We also calculated the corresponding network indicators with the Gephi program [12] .The network statistics we obtained include node degree and average weighted degree, as well as higher-order attributes such as hierarchy, network density, and efficiency, etc.Therefore, in this study, we chose relevant indicators to understand the structure and characteristics of interregional food and energy networks.
Besides, community detection is a technique used to reveal the aggregation behavior of networks.Community detection is essentially a network clustering method, where each social network consists of several "communities", and by calculating the modularity, the network can be divided into several communities with similar internal properties, and for two nodes that are closely connected, the larger the weight of the edge, the more easy to divide into the same communities.The topology of the association network can be obtained by Gephi program using OpenOrd layout for visualization.

Overview of different grain trade
In this paper, the three grain products (wheat, maize, and rice) with the highest trade volume in the main grain trading counties of Ningxia in 2020 were chosen as the subject of study (Figure 1), and the total trade volume of these three grain products accounted for 85.74% of the total trade volume of all grain products selected, according to statistical data.The volume of grain imported and exported by each country is closely related to climatic conditions, population, and trade barriers.
The main rice-producing areas are concentrated in Ningxia's northern plains.The most important rice export areas are Pingluo and Helan.Pingluo was one of them, exporting 10.02×10 4 tons of rice to other counties.Crop patterns differ between regions due to varying climate, precipitation, and soil; rice and wheat are the most important crops in Northern Ningxia, while maize is the main crop in Southern Ningxia.Ningxia's most productive crop is maize.The main output areas are Zhongping, Tongxin, and Pingluo, while the main input area is Yinchuan.Maize input will total 55.92 × 10 4 tons, accounting for 67.32% of total maize input.

Water and carbon footprint flow network
We found that total 7.53×10 8 m 3 of water footprint was transferred through grain trade network in 2020(Figure 2).Yinchuan, as the largest water input footprint area, has 3.59×10 8 m 3 of virtual water input, accounting for nearly half of the total virtual water input.The main virtual water export areas are Xiji, Pingluo and Penyang, which are also the main food export areas.
Similarly, we analyze the grain carbon footprint flow network.The results show that there are 90 links between 19 counties and districts in the network, with a total of 2.154 7 million tons of CE (excluding the internal flow of counties) (Figure 3).Pingluo and Xiji are the two counties with the highest carbon footprint output.Through grain flow, these two counties exported 54.11×10 4 tCO 2 -eq and 32.23 × 10 4 tCO 2 -eq carbon footprints, accounting for 25.11% and 14.96% of total grain carbon footprint output, respectively.The areas with the highest grain input, Yinchuan and Shizuishan, also have the highest grain carbon footprint input.Through grain flow, they input 13.58×10 5 tCO 2 -eq and 25.09×10 4 tCO 2 -eq grain carbon footprints.
The flow of virtual water has alleviated the pressure on local water resources to some extent.However, there are some counties where virtual water flows from waterscarce areas to water-rich areas, such as Xiji, failing to effectively use water resources from water-rich areas to meet the needs of water-scarce areas, and the flow pattern of virtual water lacks rationality.

Network features and analysis of the FEW-Nexus
As shown in Fig. 4, it presents the network for grain trade.In these networks, the size of the nodes represents the degree of weighting of the transport values direction of the flow is clockwise.Different colors represent different communities.In terms of community detection, the differentiation of grain trade network communities and the geographical location of the county nodes are closely linked, and the grain trade network has a high degree of differentiation and better independence of communities.The network currently differentiates into three major communities.The first community contains 9 counties or districts, including Yinchuan, Helan, Shizuishan, Pingluo, Haiyuan, Hongsipu, Tongxin, Yongning and Zhongning, mainly in the northern plains of Ningxia, and the grain trade value accounts for 69.44% of the overall network flux, of which Yinchuan is more representative and is the core of the entire network.The second community includes three counties, Litong, Lingwu and Qingtongxia, concentrated in the central and northern part of Ningxia, contributing 10.26% of the grain trade, which plays a bridging role in the whole network to promote the grain trade in the northern and southern part of Ningxia.The third community includes 7 counties in Shapotou, Yanchi, Yuanzhou, Jingyuan, Longde, Pengyang and Xiji, mainly located in southern Ningxia, accounting for 20.29% of the network flux correspondingly.With the help of Gephi program, the overall network characteristics of FEW-Nexus spatial correlation network are calculated (Table 1).In this paper, 19 counties are used as network nodes, and the theoretical maximum number of relationships is 342.However, in fact, the correlation number and correlation density of agricultural water and carbon footprint networks are low, and the degree of correlation needs to be improved.There is still a large space for collaboration to jointly achieve resource collaborative optimization and carbon emission reduction.The network grade degree is 1, and the network efficiency is at a high level, indicating that the spatial transfer distribution is uneven, and there is a large grade difference.The footprint flow with large flow exists in a few core provinces, and the stability of the network structure is not strong.The flow is more concentrated on a few flow paths, with less dispersion.

Conclusions
Based on the agricultural data and meteorological data of three grain crops (rice, wheat and maize) in Ningxia, this paper estimates the grain trade situation, quantifies the virtual water and carbon footprint flow driven by grain flow, and uses Gephi to construct the water footprint and carbon footprint network model, and the FEW-Nexus was systematically analyzed.The results are shown as follows: (1) There are great differences in grain transportation, which is attributed to the differences in natural conditions, social economy, climate, farming methods and planting structure in different regions.(2) Food trade has resulted in the flow and reconfiguration of virtual water and carbon footprint, which has relieved some of the pressure on local water resources.However, some counties have low agricultural resource utilization efficiency.For example, Xiji failed to use water resources effectively in water-rich areas to meet the needs of water-deficient areas.
(3) The FEW-Nexus spatial association network is unbalanced, and the associated channels need to be enriched.The network described in this paper contributes to the understanding of the production and transfer of grain in Ningxia.In order to further this work in the future, a coupled water and nitrogen footprint framework has the potential to optimize water availability through water footprint and water quality through nitrogen footprint.
Overall, this research and other work similar to it allows us to understand the resources involved, the transfer of those resources, and how we can make the production of those resources more sustainable.The FEW-Nexus framework for conducting research and building understanding around competing resources will continue to play a role in sustainability research.Overall, this research and similar work helps us understand the resources involved, the transfer of those resources, and how we can make resource production more sustainable.The development of the FEW-Nexus framework and research on competing resources will continue to play a role in sustainability research.

Table 1 .
Network properties of FEW-Nexus