GCAMnet: training and evaluation data for a GCAM emulator

Inputs: 12. Outputs: 44 quantities by 32 regions by 16 years. ~4000 GCAM runs for each of three sampling strategies: binary, hypercube and finite diff for dgsm. See https://doi.org/10.1016/j.envsoft.2026.106945 for details.输入维度：12项。输出为覆盖32个区域、16年时间跨度的44个量化指标。本数据集包含三种采样策略下的各约4000次GCAM (Global Change Assessment Model)运行结果，三种采样策略分别为用于导数全局灵敏度分析（DGSM, Derivative-based Global Sensitivity Measures）的二元采样、超立方采样与有限差分采样。详细信息请参阅：https://doi.org/10.1016/j.envsoft.2026.106945。