Modeling Satellite Precipitation Errors Over Mountainous Terrain: The Influence of Gauge Density, Seasonality, and Temporal Resolution

• Published in: IEEE transactions on geoscience and remote sensing Vol. 55; no. 7; pp. 4130 - 4140
• Main Authors: Maggioni, Viviana, Nikolopoulos, Efthymios I., Anagnostou, Emmanouil N., Borga, Marco
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.07.2017; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Alarm systems; Algorithm design and analysis; Artificial neural networks; Atmospheric precipitations; Calibration; Climate; Climate prediction; Cold season; Cold storage; Complex terrain; Correlation; Density; Detection; Error analysis; Error correction; Error detection; Errors; Estimation; Forecasting; Hydrologic models; Hydrology; Imagery; Mathematical models; Modelling; Morphing; Neural networks; Parameters; Precipitation; Predictions; Probability theory; Products; Properties; Rain; Rain gauges; Rainfall; Rainfall intensity; Real time; Remote sensing; Resolution; satellite observations; Satellites; Scale (ratio); Seasonal variations; Seasons; Spatial resolution; Storms; Systematic errors; Temporal resolution; Terrain; Variability
• ISSN: 0196-2892; 1558-0644
• DOI: 10.1109/TGRS.2017.2688998

This paper contributes to the predictive understanding of satellite precipitation estimation errors over complex terrain, which is fundamental to the development of error models for improving hydrological applications. This paper focuses on the Trentino-Alto Adige region of the eastern Italian Alps. Rainfall observations over a 10-year period (2000-2009) from a dense rain gauge network in the region are used as reference precipitation. A number of satellite precipitation error properties (probability of detection, false alarm rates, missed events, spatial correlation of the error, and hit biases) are investigated in terms of seasonality, satellite algorithm, rainfall intensity, gauge density, and temporal resolution dependencies. These error parameters are typically used in error models (e.g., SREM2D) and provide the basis for enhancing error scheme development. Three widely used satellite-based precipitation products are employed: 1) the Climate Prediction Center morphing product; 2) the precipitation estimation from remotely sensed imagery using artificial neural networks; and 3) the Tropical Rainfall Measuring Mission multisatellite precipitation analysis 3B42 near-real-time product. The three products show similar performances, with larger errors during the warm season, characterized by convective storms, and less variability in the cold season, characterized by more organized stratiform systems. Lower biases are depicted at the daily scale with respect to the 3-hourly resolution. The SREM2D error model has the ability to correct the satellite precipitation products, even though attention is needed for potential systematic errors when applying the calibrated model to independent periods or regions.