Research context
This work contributes to international efforts to monitor extremes in the Asia–Pacific. It sits within the context of improving situational awareness for heavy precipitation and drought across Australia, where point-based observations are reliable but spatial interpolation is challenging due to the high variability of rainfall and the sparse, uneven gauge network in remote regions.
- Problem framing: surface gauges provide accurate records at their locations, but maps are hard to build because rainfall varies sharply in space and time.
- Research aims: (i) validate satellite-derived precipitation against BoM in‑situ gauges across Australia; (ii) evaluate performance for extremes at multiple spatial scales (e.g., 0.25°, 0.5°, 1.0°); (iii) examine links between heavy precipitation and floods.
- Study focus: regions with known extremes, including parts of NT, QLD and northern WA.
Validation design (satellite ⇄ gauges)
- In‑situ data: ~7,000 BoM gauges with daily observations (2000–2018).
- Satellite grid: high‑resolution daily fields at 0.25° × 0.25° (e.g., CMORPH).
- Collocation: nearest‑cell matching by great‑circle distance (Haversine) to pair gauges with satellite grid points.
- Skill metrics: Relative Bias (RB), Relative RMSE (RRMSE), Correlation Coefficient (CC), Probability of Detection (POD), Frequency of Hit (FOH), False Alarm Ratio (FAR), Critical Success Index (CSI), and Heidke Skill Score (HSS) for rain occurrence.
Note: The validation emphasises performance on the tail of the distribution (e.g., 95th–99th percentiles) and diagnostics at multiple grid scales.
Workflow — where things stand & what's next
0) Project setup & data QA Completed
- Goal: Assemble BoM daily gauges (harmonised time zones), select satellite products (CMORPH, GMAP/GSMaP, TRMM/GPM heritage).
- Methods: QC outliers, consistent accumulation windows, station metadata checks.
- Outputs: Clean gauge table; time‑aligned satellite stacks.
1) Satellite ⇄ gauge validation In progress
- Goal: Quantify skill overall and for extremes (≥P95/P99 and thresholds like ≥25/50 mm/day).
- Inputs: BoM gauges; CMORPH; GMAP/GSMaP; (TRMM-era records where applicable).
- Methods: Nearest‑cell collocation (great‑circle distance), continuous metrics (bias, RRMSE, CC, KGE), categorical skill (POD/FOH/FAR/CSI/HSS), reliability and Q–Q diagnostics.
- Handoffs: Site‑level scorecards; regime summaries by climate zone/topography.
2) Bias correction & fusion In progress
- Goal: Reduce systematic errors; improve extremes representation.
- Methods: Quantile mapping (local), local intensity scaling, regression‑kriging: satellite as covariate + kriged residuals to blend with gauges.
- Handoffs: Bias‑corrected satellite fields + uncertainty notes.
3) Extremes at gauges (EVA) In progress
- Goal: Estimate return levels and trends.
- Methods: GEV on annual/seasonal maxima; POT‑GPD with threshold selection (MRL/parameter stability) and runs‑based declustering; non‑stationary location/scale with ENSO/IOD (optionally MJO/SAM).
- Handoffs: Station‑wise return levels (e.g., 20‑/50‑/100‑year) with CIs; diagnostics.
4) Spatial mapping of return levels Next
- Goal: Produce national maps (e.g., 24‑hour 1‑in‑100‑year rainfall).
- Methods: Regression/universal kriging of return levels (or GEV parameters) using orography, distance‑to‑coast; variogram modelling (anisotropy checks); co‑kriging with bias‑corrected satellite where justified.
- Handoffs: Gridded rasters + cross‑validation scores (LOOCV / spatial CV).
5) Spatial‑extreme dependence Planned
- Goal: Better quantify joint tail risk over areas/catchments.
- Methods: Max‑stable processes (Brown–Resnick/Schlather) via pairwise likelihood; hierarchical models for spatially varying GEV parameters.
- Handoffs: Regional dependence diagnostics; comparison to kriging‑only approach.
6) Uncertainty & communication Planned
- Goal: Provide decision‑ready uncertainty.
- Methods: Bootstrap gauges; refit EVA; krige each bootstrap to form ensembles; scenario comparisons; IDF/ARI tables with intervals.
- Handoffs: Maps with uncertainty bands; LGA/catchment IDF tables; narrative report.
7) Integration & reporting Planned
- Reproducibility: versioned scripts, data dictionaries, parameter logs; clear provenance for products.
- Deliverables: validation dashboard, bias‑corrected datasets, return‑level atlas, technical memo.
Product primer — CMORPH, GMAP (GSMaP), TRMM
CMORPH (NOAA CPC Morphing Technique)
A high‑resolution precipitation analysis built from passive‑microwave (PMW) satellite estimates that are advected ("morphed") in time using motion fields, with IR imagery assisting temporal propagation. Daily products are available at 0.25° resolution, and Version 1.0 extends back to 1998 with homogeneous inputs; gauge‑blended variants are provided alongside satellite‑only fields.
GMAP / GSMaP (Global precipitation map)
A global precipitation mapping project based on satellite‑borne microwave radiometers (JAXA’s GSMaP initiative), providing PMW‑based retrievals and blended products suitable for near‑real‑time monitoring and climatological analysis.
TRMM (Tropical Rainfall Measuring Mission)
A joint NASA–JAXA satellite mission (1997–2015) that became a standard reference for tropical precipitation, improving understanding of cyclone structure and supporting flood/drought monitoring and forecasting. TRMM heritage underpins later products and the GPM era.
Expected outcomes
- Quantified skill of satellite products over Australian terrain and climates, including biases for extreme‑rain days.
- Bias‑ and gauge‑blended datasets that narrow the gauge–satellite gap, with room for further improvement as new sensors/products emerge.
- Region‑specific recommendations (metrics and model choices may vary by climate regime and topography).