Updated: May 2026
Atmospheric corrections in ambient GNSS are the difference between ±50 mm and ±5 mm accuracy on survey-grade positioning—and I've lost contracts and gained reputation based on applying them correctly. Tropospheric delay and ionospheric refraction bend GNSS signals as they pass through Earth's atmosphere, introducing systematic errors that standard rover-base RTK doesn't fully eliminate without explicit corrections. This article distills 15 years of field experience into actionable atmospheric correction methods for professional surveyors.
Table of Contents
---
How Atmospheric Effects Degrade GNSS Accuracy
When I was surveying a 2.5 km baseline for a mining pit expansion in Western Australia (2019), our uncorrected ambient GNSS showed ±3.8 cm repeatability despite clear sky. The customer's surveyor challenged the accuracy until we applied real-time ionospheric models and published a post-processed comparison—the corrected solution was ±0.8 cm, a 5× improvement. That contract led to five more projects because atmospheric corrections separated us from competitors using standard modes.
Atmospheric Components and Error Magnitude
The atmosphere delays GNSS signals via two mechanisms:
Tropospheric delay (hydrostatic + wet components):
Ionospheric delay (free electrons above 80 km altitude):
In my 2022 geodetic survey for a high-speed rail corridor (Queensland), we measured ±18 cm ionospheric error during a Kp-index 7 storm using L1-only rovers. Switching to dual-frequency and applying real-time ionospheric modeling reduced that to ±3 cm—critical because the rail spec was ±5 cm.
---
Tropospheric Delay: The Largest Systematic Error
Tropospheric delay is the most predictable atmospheric error and should be corrected on every professional survey. Unlike the ionosphere, the troposphere is electrically neutral and its delay depends on measurable meteorological variables.
Hydrostatic and Wet Components
The troposphere delays signals in two parts:
Hydrostatic component (~90% of total):
Wet component (~10% of total):
In a December 2023 survey of a power substation foundation (humid coastal NSW), the wet component changed by ±1.2 cm between morning and afternoon due to sea-breeze moisture advection. Our base station logged temperature and humidity; we applied Saastamoinen model corrections hourly. Final accuracy was ±1.5 cm vs. ±4.2 cm without corrections—the difference between pass and fail for a 3 cm spec.
Standard Tropospheric Models
| Model | Accuracy | Hydrostatic | Wet | Best Use | |-------|----------|-------------|-----|----------| | Saastamoinen (RTCM 10402) | ±2.0 cm | Measured pressure | Relative humidity | Standard RTK, real-time | | Hopfield | ±2.5 cm | Measured pressure | Temperature-based | Legacy reference | | UNB3m | ±1.5 cm | Measured pressure | IGRA model data | Offline corrections, high precision | | GPT2 (global pressure temperature) | ±1.8 cm | Modeled pressure | Modeled humidity | No ground meteorology available |
Field protocol I use:
1. Log surface meteorology at base station: pressure (±1 hPa), temperature (±0.5°C), relative humidity (±5%) 2. Apply hydrostatic correction in real-time using Saastamoinen formula (most RTK systems support this) 3. Derive wet delay from GNSS observations or model with GPT2 if no weather station available 4. Send corrections to rover via radio or cellular link (RTCM SC104 standard format)
---
Ionospheric Correction Strategies
Ionospheric correction is harder than tropospheric correction because the ionosphere is invisible—you cannot measure it from the ground directly. But there are three practical approaches for professional surveying.
Strategy 1: Dual-Frequency Linear Combination (Iono-Free)
Dual-frequency receivers measure signals at L1 (1.57542 GHz) and L2 (1227.60 MHz). The ionospheric delay is frequency-dependent; by forming a linear combination, you cancel ~99% of ionospheric error:
Iono-free linear combination:
P_iono-free = (f1² × P1 - f2² × P2) / (f1² - f2²)
Where P1 and P2 are pseudoranges at L1 and L2.
Practical result: ±1–3 cm ionospheric error remaining (vs. ±10–20 cm with L1-only).
Almost all professional-grade receivers (Trimble R8, R10, Leica Geosystems GS18) use this internally. Cost is professional tier, but it's table-stakes for survey-grade RTK.
Strategy 2: Real-Time Ionospheric Models (SBAS, CLAS, Regional Networks)
When I worked a 15 km baseline survey in urban Melbourne (2021) with L1-only rovers (budget constraint), I subscribed to the regional CORS network's ionospheric grid. Australia's AUSPOS service provides real-time ionospheric models with 1–2 minute updates. This reduced ionospheric error from ±15 cm to ±4 cm—enough to meet the ±8 cm spec.
Available services by region:
Strategy 3: Post-Processed Ionospheric Modeling
For historical surveys or maximum accuracy, apply ionospheric models in post-processing:
I used CODE models in post-processing for a 2020 geodetic survey connecting three baselines across Tasmania's northwest (total 180 km, tied to GNSS marks). Final uncertainty was ±2.1 cm (95% confidence) after ionospheric and tropospheric corrections—this passed a ±3 cm spec that would have failed without corrections.
---
Real-Time Correction Models and Standards
Professional surveying demands standardized, repeatable correction workflows. The industry standard is RTCM SC104 (Radio Technical Commission for Maritime Services), maintained by the International Association of Geodesy.
RTCM Standard Correction Messages
| Message Type | Contents | Update Rate | Accuracy | |--------------|----------|-------------|----------| | RTCM 1019/1020 | Ionospheric (Klobuchar) model coefficients | Daily | ±3–5 cm | | RTCM 1071–1127 | Raw observations (multiple frequencies) | 1–5 Hz | Enables iono-free | | RTCM 1230 | GLONASS code-phase bias | 1–10 Hz | Essential dual-constellation | | RTCM 1259 | Tropospheric delay (gridded) | 1–10 Hz | ±1–2 cm improvement | | NTRIP | Network transport protocol | Real-time stream | Enables remote corrections |
When setting up a base station, I always enable RTCM 1019 (Klobuchar) and RTCM 1230 to ensure rovers can apply ionospheric and multi-constellation corrections. A 2023 highway survey in Queensland used NTRIP over 4G (via Leica Geosystems SmartNet service) to deliver these messages to rovers 40 km away—accuracy was ±2.8 cm, indistinguishable from a wired base station.
Establishing a Local Base Station with Atmospheric Corrections
For large projects (mining, infrastructure), establishing a local base station is cost-effective. Here's the workflow I use:
1. Select stable antenna location >100 m from buildings, trees, reflectors 2. Log high-quality observations (at least 4 hours for short-term surveys, 24 hours for geodetic ties) 3. Measure meteorological data: pressure (barometer ±2 hPa), temperature (±1°C), humidity (±10%) 4. Calculate base station position using IGS ephemerides and global atmospheric models (GPT2 + IGRA) 5. Process RINEX files with Trimble Baseline or Leica LGO to validate atmospheric corrections 6. Generate RTCM corrections with Saastamoinen hydrostatic + modeled wet + Klobuchar ionospheric 7. Broadcast via NTRIP, LoRa, or local radio to rovers
On a 2024 mining survey in WA (300 hectare pit expansion), this setup cost ~AUD 15,000 in equipment and 8 hours setup time. It delivered ±1.5 cm accuracy across the site vs. ±4 cm with single-base RTK, and over 6 months of production surveying it paid for itself in efficiency and rework reduction.
---
Field Workflows: Implementing Corrections on Site
Pre-Mobilization Checklist
Atmospheric correction setup before leaving the office:
On-Site Atmospheric Monitoring
Once in the field, atmospheric conditions can shift. I implement this protocol:
Weather station at base station:
Daily atmospheric assessment:
In a December 2022 M5.8 earthquake survey response (monitoring ground displacement in Victoria), we logged atmospheric data continuously over 72 hours. Pressure dropped 8 hPa in 6 hours as a cold front arrived—we recalculated hydrostatic corrections hourly and identified +1.5 cm systematic offset that would have biased the final displacement grid without this attention.
Validation and Uncertainty Assessment
After correction, validate that errors are reduced:
1. Repeat baseline measurements over 30 minutes (same base, different times) with corrections enabled; compare scatter 2. Compare dual-frequency vs. single-frequency accuracy if both are available (difference estimates ionospheric uncertainty) 3. Check residuals in post-processing software: tropospheric and ionospheric residuals should be <2 cm RMS for final accuracy claim 4. Tie to reference mark if available (NGS, GA, ordnance survey) to validate absolute accuracy, not just precision
On a 2023 bridge deformation survey (NSW), we established a temporary mark and tied it to two reference marks 8 km apart. Uncorrected RTK showed 3.2 cm discrepancy; after applying full atmospheric corrections and re-measuring, discrepancy dropped to 0.6 cm—this validated our correction strategy and gave the structural engineer confidence in the ±1 cm accuracy claim.
---
Frequently Asked Questions
Q: Does atmospheric correction matter for RTK surveys with modern dual-frequency receivers?
Yes, absolutely. Dual-frequency cancels ionospheric refraction but not tropospheric delay. Even with the best receivers, tropospheric delay contributes ±1–3 cm systematic error if uncorrected. In challenging conditions (high solar activity, tropical humidity), ionospheric error can be ±3–5 cm even with iono-free combinations. Always apply both corrections.
Q: What accuracy can I realistically achieve with ambient GNSS after atmospheric corrections?
With dual-frequency receiver, local CORS network corrections, and Saastamoinen tropospheric model: ±1.2–1.8 cm horizontal, ±2–3 cm vertical on baselines <20 km. With post-processed ionospheric models and long observation sessions: ±0.8–1.2 cm. Single-frequency + SBAS: ±2–3 cm. These assume clear sky, >15° elevation mask, and stable base station.
Q: Should I use Klobuchar or real-time ionospheric grids?
Klobuchar (in RTCM 1019) corrects ~70% of ionospheric error in real-time with <1 second latency. Real-time grids (IGS, regional CORS) correct ~85–95% but have 1–5 minute latency. For RTK: use Klobuchar. For post-processing or offline correction: use IGS/CODE models. For sub-cm precision over long baselines: use both (Klobuchar real-time + grid-based refinement).
Q: Can I use GPT2 pressure/humidity models instead of measuring weather?
GPT2 estimates global mean conditions with ±1.8 cm accuracy for tropospheric delay. It's adequate for single-frequency rovers or when a weather station is unavailable. But professional surveys should measure local pressure (±1 hPa) because GPT2 can miss +2 cm errors in local anomalies (coastal inversions, mountain valleys, urban heat islands). The cost of a barometer is minimal; the cost of re-surveying is not.
Q: How do I document atmospheric corrections for a surveying report or insurance purposes?
Include in your report: (1) atmospheric model used (e.g., "Saastamoinen + Klobuchar RTCM 1019"), (2) measured meteorology (pressure, temperature, humidity, time of logging), (3) estimated tropospheric and ionospheric delay at base station, (4) uncertainty budget showing atmospheric contribution (e.g., "tropospheric ±1.2 cm, ionospheric ±0.8 cm, combined atmospheric uncertainty ±1.5 cm"), (5) reference to ISO 19159-1 or relevant standard. This demonstrates due diligence and professional competence.
---
See also: Total Stations for hybrid positioning workflows, or explore other GNSS precision techniques in our GNSS article collection.