Summary Tables¶

Annual Results (`table_results`)¶

Produces an annual summary table by aggregating daily results. Years with more than 10% missing ConcDay values are excluded.

annual = w.table_results()
print(annual)

Column	Description
`DecYear`	Decimal year (mid-point of the period)
`Q`	Mean discharge (m^3^/s)
`Conc`	Mean daily concentration (mg/L)
`Flux`	Mean daily flux (kg/day)
`FNConc`	Flow-normalised concentration
`FNFlux`	Flow-normalised flux
`GenConc`	Generalized concentration (if `kalman()` was run)
`GenFlux`	Generalized flux (if `kalman()` was run)

The period of analysis can be customised:

# Calendar year
annual = w.table_results(pa_start=1, pa_long=12)

# Water year (default)
annual = w.table_results(pa_start=10, pa_long=12)

Computes changes in flow-normalised values between specified years:

changes = w.table_change(year_points=[1985, 1995, 2005, 2010])
print(changes)

Returns one row per consecutive pair of years with columns:

Column	Description
`Year1`, `Year2`	The comparison endpoints
`FNConc_change`	Absolute change in FNConc
`FNConc_pct_change`	Percent change in FNConc
`FNConc_slope`	Annual slope (change / years)
`FNConc_pct_slope`	Annual percent slope
`FNFlux_change`	Absolute change in FNFlux
`FNFlux_pct_change`	Percent change in FNFlux
`FNFlux_slope`	Annual slope for flux
`FNFlux_pct_slope`	Annual percent slope for flux

The flux_factor parameter (default 0.00036525) converts flux from kg/day to 10^6^ kg/year.

Cross-validation error statistics based on leave-one-out predictions:

stats = w.error_stats()

Key	Description
`rsq_log_conc`	R-squared for log-concentration
`rsq_log_flux`	R-squared for log-flux
`rmse`	Root mean square error (log-space)
`sep_percent`	Standard error of prediction as a percentage

Requires cross_validate() or fit() to have been called first.

Evaluates the bias of flux estimates relative to observed values:

bias = w.flux_bias_stat()

Key	Description
`bias1`	Flux bias using `ConcHigh` as the observed value
`bias2`	Flux bias using `ConcLow` as the observed value (NaN treated as 0)
`bias3`	Average of `bias1` and `bias2`

The formula is: bias = (estimated - observed) / estimated

A value near zero indicates unbiased estimates. Positive values indicate overestimation; negative values indicate underestimation.