Skip to content

Nowak (2010) KNN Temporal Disaggregation
Type Nonparametric
Resolution Monthly to Daily
Sites Univariate / Multisite
Class NowakDisaggregator

Overview

The Nowak disaggregator converts synthetic monthly flows to daily flows by borrowing within-month daily patterns from the closest historical analogs. For each synthetic month, it identifies the K nearest historic months by total flow, randomly selects one using Lall-Sharma kernel weights, and applies its daily proportions to the synthetic total.

Algorithm

Preprocessing

  1. Validate and store daily observed flows.
  2. Build historic monthly totals at the index gauge (sum across all sites for multi-site disaggregation).

Fitting

For each calendar month m, fit a sklearn.NearestNeighbors model on the scalar historic monthly totals. Only historic months within +/- max_month_shift calendar days of month m's center are included in the pool.

Disaggregation

For each synthetic monthly flow Q_syn_m:

  1. Find neighbors - query the KNN model for the K nearest historic months by Euclidean distance on total flow.
  2. Select one neighbor - draw with probability proportional to Lall-Sharma kernel weights (Lall and Sharma 1996). Neighbors are ranked by distance (rank i = 1 is closest), then selected with probability: 
    K(i) = (1/i) / sum_{j=1}^{K} (1/j)
    This harmonic weighting favors closer analogs while maintaining stochastic diversity.
  3. Disaggregate - apply the selected month's daily proportions: 
    q_d = Q_syn_m * (q*_d / sum(q*_d'))
    For multi-site data, each site is disaggregated independently using the same selected analog month.
  4. Smooth month boundaries (if blend_days > 0) - apply weighted blending across adjacent months to reduce discontinuities at month transitions, then rescale each month to preserve the original monthly total exactly.

Parameters

Parameter Type Default Description
Q_obs pd.Series or pd.DataFrame - Observed daily streamflow with DatetimeIndex
n_neighbors int 5 K - number of candidate neighbors
max_month_shift int 7 Days of calendar flexibility around each month
blend_days int 2 Days of overlap smoothing at month boundaries to reduce discontinuities (0 = no blending)
name Optional[str] None Optional name identifier for this disaggregator instance
debug bool False Enable debug logging

Properties Preserved

  • Daily flow patterns within each month (borrowed from historical record)
  • Monthly totals (exact, by construction - preserved even after boundary blending)
  • Multi-site consistency (same analog month applied to all sites)

Not preserved: - Month-to-month daily transitions (partially addressed by blend_days boundary smoothing) - Daily patterns outside historical range

Limitations

  • Cannot produce daily patterns not seen in the historical record
  • Leap year handling requires proportional adjustment when analog and target months differ
  • Quality depends on having a sufficiently long historical record to find good analogs

References

Primary: Nowak, K., Prairie, J., Rajagopalan, B., and Lall, U. (2010). A nonparametric stochastic approach for multisite disaggregation of annual to daily streamflow. Water Resources Research, 46(8). https://doi.org/10.1029/2009WR008530

See also: - Lall, U., and Sharma, A. (1996). A nearest neighbor bootstrap for resampling hydrologic time series. Water Resources Research, 32(3), 679-693.

Implementation: src/synhydro/methods/disaggregation/temporal/nowak.py Tests: tests/test_nowak_disaggregator.py