GPS snow sensing: results from the EarthScope Plate Boundary Observatory Kristine Larson, Felipe Nievinski Department of Aerospace Engineering Sciences University of Colorado at Boulder
Why this paper? Non-standard GPS application Multipath application Implications for – Expanded / novel applications of NGS CORS – “one NOAA” … snow and soil moisture = water cycle
Outline Multipath / reflections basics Specifics of snow sensing technique – How it’s done – Why it’s useful Paper overview Tie in to NOAA / NGS work that could be done with CORS
Definition of some terms Multipath: signal travels by multiple paths SNR: signal-to-noise ratio (one type of GPS observable, derived from phase tracking; sensitive to phase multipath) PBO: Plate Boundary Observatory (CORS-type network for plate tectonics) SNOTEL: network of in situ snow sensors (truth measure for snow) PRN: pseudorandom noise code (satellite ID number)
GPS Multipath Satellite signals become a plane wave by the time they reach Earth’s surface. Some ray paths are received by antenna on the ground: Direct path After bouncing off objects
GPS Multipath – SNR from simple ray tracing Slide modified from K Larson PdPd PrPr
GPS Multipath : remove direct + gain pattern Slide modified from K Larson Ignore the direct Analyze only the multipath part of SNR 10 deg Elevation Angle 30 deg
Each rising or setting GPS satellite samples a ground track Interference pattern in SNR data Slide courtesy of K Larson
Snow Depth/Height estimate reflector height from frequency of interference pattern P101 (Utah) Slide courtesy of K Larson
Snow Depth/Height sensing footprint (Fresnel zone) also changes with snow height Fresnel zone = function of Height above surface Elevation angle of satellite
Result: 20 observations per day (maximum) Sampling area: ~1000 m 2 Slide modified from K Larson Specular reflection point (bounce point off of snow/ground)
Spatial resolution cm 3 soil moisture probes 1000 m 2 GPS km Satellite missions 10 m 2 SNOTEL
Terrain Slope Effect Error in snow depth < 2-5 cm for ground slopes of 8 degrees or less
Why you can’t apply this technique everywhere DateSnow Depth d300/10No snow (bare ground) d330/10~ 50 cm snow d080/11~ 100 cm snow
P101 (works great) Local slope calculations show low-angle terrain to NW and due S Terrain is wide open, no trees.
P101 … most of the time Complex reflections to the NE, observed with PRN31. not all satellite tracks at a station will work equally well
P360 (also pretty good) Absence of data makes peak overly wide Site is very flat over the sensing zoneOpen scrubland
P711 (where are the wiggles?) … Lots of trees in the sensing footprint Overwhelms the relatively flat terrain… Sometimes a station is “great on paper” but other factors will override potential signal
P720 (nada) Trees, trees and more trees
How to make a snow depth time series 1.Pick satellite tracks (strong LSP peaks) 2.Analyze rising and setting separately (different hunks of ground) 3.Get ground height from summertime data 4.Every day 1.Peak LSP (LSP height minus ground height = snow depth) 2.Average all good satellites into site mean (scatter of satellites = error bar)
Snow Depth Over 2 Years Utah Idaho Wyoming 15 km apart
Snow Depth Over 2 Years Utah Idaho Wyoming
Cautionary Tale: SNR data quality Not all GPS receivers / data types provide good SNR! Breaking the multipath “rules” Poor resolution (firmware) P101 Trimble NetRS PRN25 vs time L2C, PRN25 L1C, PRN25 L2P, PRN18 L2C with 1.0 dB resolution
What about CORS?
CORS case study: MNDOT Network All Trimble receivers Operating at 5-sec rates Mix of NetRS NetR3 NetR5 NetR9 Stations often next to farm fields Asked VRS operator to turn on L2C in January 2011 COOP stations across state Home of SNOWDAS
ASHL Looking West
MRTA
To integrate CORS into PBO H2O Collect SNR data Evaluate SNR quality of non-Trimble receivers (Leica, Topcon, etc) – Varies by model, firmware release – Varies for phase observable (L1C, L2C, L2P, L5) Manually sort stations – Exclude 100% urban – Flat terrain – Find rural, or semi-rural (side of building next to field) Not Good Good
Adding CORS to PBO H2O would get our logos here
Signal-to-Noise Ratio (SNR) Measure of signal strength for each satellite Total SNR = direct plus reflected signal(s) – Direct amplitude = dominant trend – Multipath signal = superimposed on direct
Multipath: One man’s error is another man’s signal Multipath = direct + reflected signals Error: increases apparent distance from satellite Signal: amplitude and phase of GPS signal affected by reflecting surface
Multipath Geometry amplitudesangles A d direct signal amplitude A m multipath signal amplitude h reflector distance angle of reflection satellite elevation angle path delay
Simplified Multipath Model and SNR Recorded SNR = direct + multipath signal Carrier phase error: Code (pseudorange) error (short delay) : multipath direct composite direct multipath MP composite
What do these equations tell us? Oscillations in SNR, phase MP, and pseudorange MP all have common frequency MP frequency – Key to determining – Function of Reflector distance h Reflection angle GPS wavelength Fast MP = far away Slow MP = nearby For a fixed reflector, satellite motion generates time-varying signature
SNR, after removing direct signal Wiggles… Sinusoidal Result from interference of direct and reflected signals Repeat daily Have amplitude, frequency, and phase sensitive to soil moisture