Research

Derin Precipitation Lab
From sensors to ensembles: precipitation science that holds up in practice

Research Thrust Areas

We bridge precipitation science and real-world hydrologic needs—advancing radar QPE, uncertainty-aware extremes, and scalable ensembles for flood and infrastructure-relevant decisions.

Thrust 1 — Radar QPE and Radar Observing Systems

We develop and evaluate radar-based quantitative precipitation estimation (QPE)—linking retrieval physics, quality control, and uncertainty to hydrologic relevance.

Key questions
– What uncertainty representation is honest and useful for downstream modeling?
– When and why do radar QPE errors emerge across regimes (convective/stratiform, bright band, mixed phase)?
– How do retrieval choices (Z–R, dual-pol, VPR handling, attenuation correction) translate into hydrologic error at basin scales?

Thrust 2 — Uncertainty and Extremes

We quantify precipitation uncertainty and extremes by diagnosing error mechanisms in satellite/radar products and modeling tail behavior in ways aligned with hazard-relevant thresholds.
Uncertainty and error mechanisms
– How do errors depend on terrain, storm type, season, and atmospheric drivers (moisture flux, instability, orographic lift)?
– How should uncertainty be expressed so it’s meaningful for hydrologic modeling?
Extremes and tails
– Do products preserve tail behavior and event structure, or do they “clip” extremes?
– How does uncertainty in extremes propagate into return levels / design metrics?
– How do extremes cluster in space/time, and what does that imply for hazard estimation?

Thrust 3 — Stochastic Rainfall Generation, Ensembles, and Downscaling

We generate scalable ensemble precipitation fields that preserve realistic space–time structure and extremes—especially when forecasts are coarse, mis-timed, or sparse.
Key questions
– How do we preserve spatial dependence and temporal coherence while generating ensembles?
– How do we handle phase/timing errors in S2S forecasts without pretending the model is perfectly synchronized?
– When does downscaling improve decision-relevant skill, and when is it cosmetic?

Thrust 4 — Hazard Applications: Flood Risk, PMP, and Convective Hazards

We translate precipitation science into hazard-relevant metrics and workflows—supporting flood resilience, infrastructure safety, and scalable convective hazard assessment.
Flood hazard assessment and resilience
– Probabilistic precipitation inputs for hydrologic modeling and threshold exceedance analysis
– Event-based evaluation tied to flood-relevant outcomes (timing tolerance, intensity-duration relevance)
PMP modernization and design-relevant extremes
– Research workflows that integrate physical understanding, ensembles, and climate context
– Uncertainty on design metrics (return levels, bounds) rather than single deterministic numbers
Convective hazard emulation (hail/convective risk)
– Data-driven emulators that reproduce CPM-like skill at large ensemble/scenario scale
– Benchmarking and uncertainty characterization (what the emulator captures vs misses)