Tracing spatiotemporal changes in agricultural and non-agricultural trade networks of India

The evolving international economic instability and international trade relationship demand a nation to move towards a self-reliant integrated system at a sub-national scale to address the growing human needs. Given India’s role in the global trade network, it is critical to explore the underlying extensive complex trade network at the domestic scale. The potential advantages of complex interaction among the different commodities remain unexplored despite the known importance of trade networks in maintaining food security and industrial sustainability. Here we perform a comprehensive analysis of agricultural flows in contrast with non-agricultural commodities across Indian states. The spatio-temporal evolution of the networks from 2010–2018 was studied by evaluating topological network characteristics of consistent spatially disaggregated trade data. Our results show an increase in average annual trade value by 23.3% and 15.4% for agriculture and non-agriculture commodities, respectively, with no significant increase in connectivity observed in both networks. However, they depict contrasting behavior concerning the spatio-temporal changes, with non-agriculture trade becoming more dependent on production hubs and the agriculture trade progressing toward self-reliance, which signifies the evolution of the diversification in the existing agrarian trade network. Our findings could serve as an important element in deepening the knowledge of practical applications like resilience and recovery by devising design appropriate policy interventions for sustainable development.

commodity transfer.We use the consumer price index as an indicator for calculating the adjusted value of the corresponding year.The mathematical representation is given in (Equation 1).

AdjustedV alue j =
V alue i × CP I i CP I j (1) Where j represents the current year, and i depicts the original year, CPI is the consumer price index for different commodities.We obtain the CPI data for the temporal window of 2010 to 2018 from the Reserve Bank of India (RBI) [2].We replicate a similar analysis of trade transfer for inflation-adjusted monetary value.The

Statistical relationship between trade transfer, Population and Gross state domestic product
We explore the statistical relationship of the Domestic Interstate Trade Network (DITN) with non-topological parameters, including states' population and Gross State Domestic Product (GSDP).The latest demographic data is available from the 2011 census, which may not precisely depict the present population and the population's variation in the study period.To address this issue, We obtain the population data of each state for the time frame of 2010 to 2018 from the Worldpop data repository at a spatial resolution of 1 km [3].We then aggregate the population data for each state through zonal statistics in ArcGIS.We analyse the statistical relationship between time-averaged population with time-averaged trade import and export for each state in agricultural and non-agricultural trade networks In the case of GSDP, we collect the individual state's GSDP from RBI [2] for the period of 2010 to 2018.To compare the actual growth of the GSDP, we convert the GSDP value to GSDP at Purchasing Power Parity (PPP), which indicates the purchasing power of an individual state.To calculate the GSDP at PPP, we use the annual exchange rate of INR (|) to USD ($) and the PPP conversion factor obtained from the world bank [4].The calculation of GSDP at PPP is shown in (Equation 2).
GSDP @P P P y,s = GSDP y,s × P y E y Where y depicts the year, s represents the state, P is the PPP conversion factor, and E is the exchange rate.
We follow a similar approach to identify the statistical relationship for the time-averaged value of GSDP at PPP against the trade import and export for both agriculture and non-agriculture trade as we did in the case of the population (Fig 12).We also quantify the statistical relationship of the import and export trends with GSDP at PPP trend (Fig 13).

Trade import
Trade import Fig 8 to Fig 11 shows the result for inflation-adjusted trade transfer.
Fig 12.The statistical relationship of import and export trend with population trend is also analysed for individual states and shown in Fig 13.

Fig 1 .
Fig 1. Interstate trade network based on physical values (weights).The chord diagrams show the average trade flows between different states in India.The links' width showcases the trade volume in quintals, and the links' colours correspond to the exporting regions (a and b), depicting the trade network for Agricultural and Non-Agricultural commodities.(c and d) shows average exports and imports for Agriculture and (e)and (f) for Non-agriculture commodities in the study area.

Fig 2 .Fig 3 .
Fig 2. Temporal changes of total trade volume and network characteristics.a) Total traded value(in 10 8 quintals), (b) Average degree, (c) Network density, and (d) Average Clustering Coefficient for 2010-2018.here,β indicates the slope of a linear trend line fitted to scattered points (values with * represent significant trend at p-value < 0.05).The results are presented for the Agriculture (green) and Non-Agriculture (purple) interstate trade networks.

Fig 4 .Fig 5 .
Fig 4. Temporal variation of exports and imports of leading exporters The temporal variation of exports and imports of leading exporters of agriculture (Punjab and Haryana) and non-agriculture (Odisha and Chhattisgarh) commodities over the period 2010-2018.Also shown are the respective linear regression slope (β).

Fig 6 .
Fig 6.Statistical relationship between indegree and outdegree (A) Scatter plots of average indegree and outdegree over nine years (2010-2018) for Indian states.β indicates the slope of the fitted linear regression line to scattered points, and r indicates the Pearson correlation coefficient.(B) Scatter plots of the year-wise slope of indegree and outdegree.The size and numerical above the marker represent the respective correlation values.In all panels, shown are the fitted regression line (dash line) and corresponding 95 % confidence band.The results in sub-panels (a) and (b) are for agriculture trade and non-agriculture trade, respectively.

Fig 7 .MayFig 8 .
Fig 7. Spatio-temporal network centralization metrics.(a and b) shows each state's betweenness and degree centrality over the temporal window of 2010 to 2018 for agricultural trade and non-agricultural trade, respectively.The solid line represents the correlation coefficient r > 0.6, the dashed line depicts the r between 0.5 and 0.6, whereas the dotted line represents r< 0.5

Fig 9 .Fig 10 .Fig 11 .
Fig 9. Temporal changes in total trade value and topological characteristics of the network based on inflation-adjusted values.(a)Total traded value (in inflation-adjusted 100 billion |), (b) Average network degree, (c) Network density, (d) Average clustering coefficient, (e) Average betweenness centrality, and (f) modularity for 2010-2018.here, β indicates the slope of a linear trend line fitted to scattered points (values with * represent a significant trend at p − value < 0.05).The results are presented for the Agriculture (green) and Non-Agriculture (purple) interstate trade networks.

Fig 12 .
Fig 12. Statistical relationship between trade export/import and population/GSDP at PPP.(a) and (c) shows the statistical relationship between the time-averaged population vs export and import of each state for agriculture, whereas similarly (b) and (d) show the same for non-agriculture DITN, respectively.(e) and (g) depict the statistical relationship between time-averaged GSDP at PPP and the export/import of each state for agriculture.In comparison, (f) and (h) show the same for non-agriculture DITN.

Table 1 .
The monetary value of each commodity in the window of 2010-2018