Michael James Friedel , Dave Stewart, Xiaofeng Lu, Helen Manley, Tom Dyer

Recent New Zealand legislation requires that regional councils set limits for water resource usage to manage the effects of abstractions in over-allocated catchments. Toward that end, an environmental modeling algorithm is proposed and demonstrated for applicability to sustainable stream management across the Otago Region of New Zealand. This four-layer algorithm includes a Data model, Base models, a Meta model, and a Sustainability model. The training and testing of Base models using limited natural catchment data (N=49) provided prediction accuracy equal to or better than very good (R2 > 0.9) when predicting naturalised Mean flow (Mean) and 7-day Mean Annual Low Flow (MALF). Bias-corrected Meta modeling provided naturalised empirical cumulative distribution functions (ECDFs) for predictions at each gauged location. Naturalised predictions are independently validated using statistical, basin transfer and water balance methods. Application of the Sustainability model to naturalised Mean and MALF predictions provided naturalised default minimum flows and naturalised default allocation rates that when combined with consented abstractions determined the probable naturalised allocation status of human-influenced catchments (N=317) across catchment scales (1st to 7th order streams). The ECDFs of naturalised hydrology provide flexibility in selecting the level of risk to manage water-resource sustainability for over-allocated catchments; for example, at the respective 10th, 20th, 30th, 40th, 50th, average, 60th, 70th, 80th, and 90th percentiles the number of over-allocated catchments is determined to be 72 (over conservative), 68, 62, 57, 54, 50 (most likely), 45, 37, 31, and 26 (under conservative). In addition to the Otago Region, the proposed algorithm can be applied to inform sustainable stream management in regional catchments across New Zealand.