A pattern-oriented wavelet approach for defining the effective resolution of precipitation products for climate and hydrologic impact assessment studies
Abstract
Gridded precipitation products derived from gauge, radar, and/or satellite observations are widely used for hydrologic assessment studies. Typically, the grid size of the precipitation product is referred to as the "resolution" and assumed to be the scale at which the product resolves the precipitation variability and thus expected to resolve hydrologic variables and extremes via rainfall-runoff modeling. However, the gridded representation of the product is generally obtained by projection or interpolation of the observations and the choice of the grid may be partially arbitrary. Gridding a product at a fine resolution does not necessarily guarantee that the product resolves precipitation features at that fine scale.
Numerous published validation studies reveal that independent products may show very large discrepancies when compared at their full "nominal resolution". Multiscale assessments generally show decreasing retrieval performances at finer spatial and temporal scales. Assuming that the retrieval error is zero mean, additive, purely random, temporally and spatially uncorrelated, the variance of the error theoretically decreases with spatial and temporal aggregation proportionally to the number of averaged independent values. However, the hypothesis of additive spatially and temporally uncorrelated errors is generally not verified, the scale-dependence of the retrieval performance cannot be reduced to this simple theoretical model and should therefore be systematically assessed. For hydrologic applications, it is important to know down to which scale a given product can accurately resolve the spatial and temporal features of precipitation, as then discrepancies in hydrologic validation can be traced to model inadequacies and not to the input product. Here we develop a methodology to define the "effective resolution" of a product as the finest scale at which it can reproduce the spatial and temporal variations of the reference precipitation. The method consists of assessing the precipitation spatial patterns and temporal variations in a multiscale spectral domain (via wavelets) against a trusted reference. We evaluate a suite of gauge, radar, satellite and multisensor precipitation products among which the various TRMM and GPM retrieval products. The effective resolution is often found to be several orders of magnitude coarser than the grid resolution, and to vary with geographical areas, surface types, seasons and types of precipitation (see figure below). The products with the finest nominal grid resolution are not necessarily the ones having the highest effective resolution.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.H41H..14G
- Keywords:
-
- 3354 Precipitation;
- ATMOSPHERIC PROCESSESDE: 1840 Hydrometeorology;
- HYDROLOGYDE: 1854 Precipitation;
- HYDROLOGYDE: 4303 Hydrological;
- NATURAL HAZARDS