320 likes | 405 Views
11/1/2011. Summary statement on runoff generation You summarized the classic “named” mechanisms My view – 2 requisite conditions 1 2 Key task: incorporate process knowledge into predictive models Next Project. How to Apply Process Information to Improve Prediction. flow. Precipitation.
E N D
11/1/2011 • Summary statement on runoff generation • You summarized the classic “named” mechanisms • My view – 2 requisite conditions • 1 • 2 • Key task: incorporate process knowledge into predictive models • Next Project
flow Precipitation time time • Improved prediction and improved process understanding are mutually reliant
REW 2 REW 3 REW 4 REW 1 REW 5 REW 7 REW 6 Small Large Coarser Parametric Fine Physics-Based Modified from Mukesh Kumar DistributedModel, Physics based Semi-DistributedModel, Conceptual Lumped Model p q q q Process Representation: Predicted States Resolution: Data Requirement: Computational Requirement: Small Large Perceived Intellectual Value:
REW 2 REW 3 REW 4 REW 1 REW 5 REW 7 REW 6 ? Mathematical Lumping Right for Wrong Reasons Process Understanding Wrong for Right Reasons Process Understanding Modified from Mukesh Kumar DistributedModel, Physics based Semi-DistributedModel, Conceptual Lumped Model p q q q Outcome: History: Future:
How do we use Process Knowledge or data in this scene? flow Precipitation time time
How do we use Process Knowledge or data in this scene? flow Precipitation time time • Calibration • Assumes “model” is correct, forces parameters to give the right answer • Rewrite model to properly represent processes
In Defense of Hydrologic Reductionism … an approach to understand the nature of complex things by reducing them to the interactions of their parts… …a philosophical position that a complex system is nothing but the sum of its parts, and that an account of it can be reduced to accounts of individual constituents … My Past Berkely Catchment Science Symposium 2009
My Past:In Defense of Reductionism • Newton was right • Model failures result from poor characterization of heterogeneous landscapes leads to • No emergent properties • Our community struggles to identify grand, overarching questions because…there are no grand unknowns • Hydrology is a local science
The Response Ciaran Harman, Catchment Science Symposium, EGU 2011
The Response Ciaran Harman, Catchment Science Symposium, EGU 2011
Catchments Lump Processes Emergent Behavior Recent work has identified many Physically Lumped Properties that are manifestations of the system of states and fluxes -A physical basis for lumped parameter modeling Decades of case studies have documented the many ways that water moves downhill
Physically Lumped Properties(emergent behavior) • Connectivity • Thresholds • Residence Time
Physically Lumped Properties • Connectivity • Thresholds • Residence Time
Threshold responses 1 Satellite Tarrawarra Runoffratio 0 0 10 20 30 40 50 Moisture content (%) Courtesy of Roger Grayson Roger Grayson, pers. Com.
Physically Lumped Properties • Connectivity • Thresholds • Residence Time
Figure from Jim Kirchner C ( t ) in ( t ) C out This approachsimplified
Model Theory: The Convolution Integral Predicted or simulated output d18O signature Input Function: Derived from precipitation d18O signal Represents d18O in water that contributes to recharge System Response Function: Time distribution of water flow paths
Soil water residence time Precipitation Soil Water d18O‰ Soil water Residence Time -4 -8 -12 -16 -4 -8 -12 -16 d18O‰ Average -9.4‰ Amplitude 0.1‰ Std Dev. 3.4 ‰ Average -9.4‰ Amplitude 1.2‰ Std Dev. 0.6 ‰ Annual Data P 2250 mm Q 1350 mm E 850 mm Average Data Slope 34o Relief 100-150m Ksat 5 m/hr Soils Data Depth 1 m Strong catenary sequence
If bedrock quite impermeableMRT and distance from the divide Vache and McD WRR 2005
REW 2 REW 3 REW 4 REW 1 REW 5 REW 7 REW 6 ? Mathematical Lumping Process Understanding Process Understanding Modified from Mukesh Kumar DistributedModel, Physics based Semi-DistributedModel, Conceptual Lumped Model p q q q History: Future:
Mathematical Lumping Process Understanding Process Understanding Physically lumped properties Modified from Mukesh Kumar DistributedModel, Physics based Physically Lumped Model Physically lumped properties q History: Future:
How to Apply Process Information to Improve Prediction • Retain the computationally efficiency and lumped philosophy of systems models • Observe how catchments create physically lumped properties • Replace mathematical lumping approaches with physically lumped properties • Use as “processes” , not data as validation targets • Build “processes” into new model structures
What do we do with this awareness? Connectivity Thresholds Residence Time
Lump the lumpsIt’s about Storage P-ET-Q =dS/dt Storage Connectivity Thresholds Residence Time
A Natural Storage Experiment Storage Capacity
A Natural Storage Experiment P-ET-Q =dS/dt Storage Capacity We should focus on Runoff Prevention mechanisms in addition to runoff generation mechanisms We should concern ourselves with how catchments Retain Water in addition to how they release water
The Storage Problem • Storage is not commonly measured • Storage is often estimated as the residual of a water balance • Storage is treated as a secondary model calibration target
Improved storage characterization will lead to improved prediction Dry Creek Reynolds Creek