new error-controlled lossy compression algorithm for large-scale scientific data, based on pointwise error bounds as opposed to absolute error bounds. Significantly improves the prediction hitting rate (or prediction accuracy) for each data point while greatly reducing data size.