The Climatic Temporal Feature Space: Continuous and Discrete

Climatic zones, representing seasonal variations in temperature (T) and precipitation (P), are generally mapped geographically using discrete classifications with distinct boundaries. However, it is well known that global T P vary continuously space time steep gradients occurring infrequently. The objective of this analysis to use complementary forms dimensionality reduction quantify the spatiotemporal climate system produce a continuous representation based on temporal feature historical alone. We characterize principal components (PCs) identify parsimonious set endmember patterns bounding all observed patterns. These provide basis for linear mixture model can represent decadal any geographic location as fractions Inverting each T+P series gives an estimate fractional contribution series. resulting fraction maps Euclidean proximity observations at every climates space. implied by variance partition + 67,420 land-based suggests effectively 3D, accounting 92% total variance. From topology space, we 4 endmembers upon which base model. Inversion normalized yields estimates misfit distribution 99% < 0.21. For comparison, also render spaces from ensembles 2D manifolds within derived suites t-distributed Stochastic Neighbor Embeddings (t-SNE) discontinuities Comparison spatial PC(t-SNE) across hyperparameter settings reveals consistent structure little sensitivity rendered t-SNE. Combining physically interpretable resolved PC finer scale manifold t-SNE provides alternative cannot character its