An NSGA-III-based predict-optimize-evaluate framework for low carbon land use optimization: A case study of the...

Reconciling carbon neutrality with competing demands for economic development, food security, and ecological conservation is a critical land governance challenge in functionally heterogeneous mega-regions. This study develops an integrated predict-optimize-evaluate framework and applies it to the Yangtze River Delta (YRD), a world-class urban agglomeration. The framework predicts six objectives-net carbon sink, economic development, food production, land use suitability, spatial compatibility, and transition probability-for 2030 and 2060. It then employs a Multi-Objective Spatial Genetic Algorithm (MOSGA), which couples NSGA-III reference-point evolution with patch-based spatial genetic operations. Under the SSP2-4.5 scenario, MOSGA simultaneously optimizes land use structure and spatial layout. Finally, it evaluates the carbon-economy-ecology nexus through carbon emission intensity, land-use-induced carbon source-sink transfers, and spatial clustering analysis. The results are as follows. (1) Optimized land use configurations reduce regional carbon emissions from 1987.07 Mt in 2020 to 1324.72 Mt by 2060, while carbon emission intensity declines from 0.79 to 0.09 t C/10⁴ CNY (an 89% reduction) and regional GDP grows from 25,201.42 to 154,871.94 billion CNY, demonstrating structural decoupling of economic growth from carbon output; however, regional carbon sink capacity reaches only 116.86 Mt by 2060, leaving a residual carbon deficit of 1207.86 Mt (2) Zone-differentiated optimization proves effective: Agricultural Production Zones expand cropland to 92,283 km² and raise food production to 109.81 Mt by 2060; Ecological Function Zones maintain forest coverage exceeding 51,000 km²; and Optimized and Key Development Zones accommodate urban expansion consistent with their economic roles, with KDZ GDP rising to 87,294 billion CNY by 2060. (3) Rural residential consolidation emerges as the pivotal spatial lever, releasing approximately 4500 km² of settlement land and forming the largest source-to-sink pathway (8.84 Mt during 2020-2060), while spatial clustering identifies priority emission reduction zones in urban-industrial cores and priority carbon sink enhancement zones in peripheral ecological counties. Collectively, this study provides a transferable methodological framework for carbon-neutrality-oriented multi-objective land use optimization, applicable to functionally heterogeneous mega-regions worldwide that operate under comparable zoning-based governance systems.