NPreciSe: NOAA Satellite Precipitation Validation System

NPreciSe

NOAA Satellite Precipitation Validation System

Documentation Satellite Products
control_point GPROF control_point bRR control_point MiRS control_point SCaMPR control_point CMORPH2 control_point IMERG control_point GSMaP control_point PERSIANN control_point SFR
Reference Products Validation Methods
control_point Detection Performance control_point Quantification Performance

Site Navigation Validation by Time Validation by Product Validation of Forecasts Documentation Contact
View as PDF

Validation Methods

Detection Performance

  • Rsat ≥ Rmin and Rref ≥ Rmin – Correct Detection/Agreement
  • Rsat ≥ Rmin and Rref = 0 – False Alarm/Satellite only
  • Rsat = 0 and Rref ≥ Rmin – Missed Detection/Reference Only

Where, Rsat – precipitation rate estimates by satellite-based product, Rref – precipitation rate estimates by ground-based reference product. Rmin is the minimum threshold. Minimum threshold depends on the satellite-based product.

Minimum threshold (Rmin) values considered:

GPROF blended Rain Rate MiRS SCaMPR SFR
0.5 mm/h 0.5 mm/h 0.5 mm/h 1.0 mm/h 0.2 mm/h
Note: SFR validation using Stage-IV does not provide detection performance maps since satellite-based SFR is used as a reference when identifying precipitation phase

Quantification Performance

Quantification performance are computed using histogram, scatterplot and statistical metrics. These are calculated only for correct detection/agreement pixels.

Statistical Metrics considered are as follows:

  1. Pearson Correlation: $$\mathit{r} = \frac{\sum \left ( R_{sat}-\overline{R_{sat}} \right )\left ( R_{ref}-\overline{R_{ref}} \right )}{\sqrt{\sum \left (R_{sat}-\overline{R_{sat}} \right )^{2} \sum \left (R_{ref}-\overline{R_{ref}} \right )^{2}}}$$
  2. Root Mean Square Deviation: $$\mathit{RMSD} = \sqrt{\sum \frac{\left ( R_{sat}-R_{ref} \right )^{2}}{\mathit{N}}}$$
  3. Bias: $$\frac{1}{\mathit{N}}\left ( \sum R_{sat} - \sum R_{ref} \right )$$
  4. Relative Bias: $$10\log\left ( \frac{\sum R_{sat}}{\sum R_{ref}} \right )$$
  5. Percent Bias: $$\frac{\sum R_{sat} - \sum R_{ref}}{\sum R_{ref}} \times 100$$
Note: Negative values of Bias, Relative Bias and Percent Bias signify underestimation while the positive values indicate overestimation.
arrow_circle_up