A NEXRAD beam leaves the dish at half a degree and keeps climbing while the Earth curves away beneath it. Over Harrisburg the lowest beam from State College is already looking about 8,300 feet up. This atlas shows how high the radar is looking everywhere in Pennsylvania, and teaches what that means for the rain and snow you measure on the ground.
Each section answers one question. The Beam Lab and Learn tabs teach the geometry; the rest are reference tools you can look things up in.
Pick a radar and a place. The plot shows every tilt you switch on, drawn to scale with the standard 4/3-Earth refraction model, against a freezing level and cloud deck you control.
Six short lessons on why radar sees less than you think, each with a live example from Pennsylvania radars. Then a five-question check.
A radar beam travels in a nearly straight line, but the Earth curves down beneath it. Two things raise the beam: the tilt angle, and the drop of the Earth's surface. Close in, tilt dominates. Curvature grows with the square of distance, and for the 0.5° beam it overtakes the tilt at about 150 km (93 mi).
The WSR-88D beam is 0.95° wide at half power. That sounds narrow, but width grows linearly with range. By Harrisburg, a KCCX pulse fills a slab of air more than a mile thick, and anything smaller than that, like a shallow snow band, gets averaged with empty air.
Air density falls with height, so the beam bends slightly back toward the ground. The standard model treats this as an Earth 4/3 larger than real (k = 4/3). On nights with a strong low-level inversion, bending increases (super-refraction) and the beam can clip the ground, producing false echoes. Choose a condition to compare heights 150 km out.
Snowflakes falling through the freezing level get a coat of meltwater. They are still large, but now they reflect like water, which returns about five times more power than ice (dielectric factor |K|² ≈ 0.93 versus 0.18). The result is a ring of inflated reflectivity just below the 0 °C level. If the beam sits in it, radar overestimates rain. If the beam is above it in dry snow, radar underestimates.
A METAR reports the cloud base, never the top. If the beam center is below the base, the radar is sampling clear air under the cloud: a definite miss. If it is above the base, the radar could be in the cloud, but it could also be above a shallow deck. In the January–April 2026 study, the definite-miss share was small. The larger and harder problem is the precipitation that fell with no ceiling reported at all.
ASOS hourly precipitation (p01i) is a running total since the last routine :56 observation. Special reports (SPECIs) during heavy rain repeat the growing total. Add them all up and you count the same rain two or three times. The correct hourly value is the largest p01i in the hour. Add SPECIs to watch the error grow.
183,349 grid cells, 800 m on a side, each checked against every radar within the 460 km reflectivity range. Pick a layer, then click or tap any cell to see its full radar stack. Drag to pan, scroll or pinch to zoom, arrow keys to step.
Heights are above mean sea level for the 0.5° tilt. “Extended” means inside the 460 km product range, not useful low-level coverage: at that range the beam is several kilometers up.
County-average height of the 0.5° beam center above the county's mean ground elevation, from eight nearby radars (5-point integration per county). Color by the radar with the lowest beam, or by the height from any one radar.
Twenty-six towns from Waynesboro to Tamaqua, computed live for any tilt. Heights are the beam center above each town's ground. The five lowest radars are shown for each town.
The ten ASOS stations used in the precipitation study, with the beam center height above each station at four low tilts from the four closest radars. The shaded block is each station's best radar, the one with the lowest 0.5° beam.
Tap a station row to open it in the Beam Lab with its best radar. Heights use the 4/3 Earth model: beam height above the antenna, plus radar site elevation, minus station elevation.
Every WSR-88D that reaches at least one Pennsylvania grid cell. Select a site to draw its 230 km core ring and 460 km reflectivity ring. The first percentage is the useful one.
January 1 to April 4, 2026. Ten ASOS stations, 3,014 corrected hourly precipitation reports, each checked against its best radar's 0.5° beam, the METAR cloud base, and the nearest IAD or OKX balloon sounding.
Each dot is one precipitation hour, colored by the temperature at the beam. The dashed line is the selected station's beam. Dots left of it had the beam above the cloud base.
Share of each station's precipitation that fell with the beam below the melting layer (liquid), inside the 2,000 ft bright-band zone, or above the freezing level (ice).
Hourly RAP model soundings from the IEM BUFKIT archive, taken at each event's peak hour, show where each station's radar beam sat relative to the freezing level, the cloud, and the wind.
Wind vectors at the surface, 1, 3 and 6 km. A clockwise turn with height (veering) signals warm-air advection.
Corrected hourly totals (max p01i per clock hour). The outlined column is the peak sounding hour.
How every number in the atlas is computed, what the model leaves out, and where the data came from.
All beam heights use the standard WSR-88D 4/3 effective-Earth-radius model:
with k = 4/3, Rₑ = 6,371 km, θ the elevation angle and r the range. Height above ground at a target is h plus the radar's site elevation minus the target's ground elevation. The half-power beam edges use θ ± 0.475° (beam width 0.95°), and the physical beam diameter is 2·r·tan(0.475°).
Terrain blockage. Ridges near a radar can block the lowest tilt. The atlas assumes an unobstructed beam; it does not model blockage.
Non-standard refraction. Real refraction varies with the temperature and moisture profile. The Beam Lab lets you compare other values of k.
Cloud tops. METAR reports the cloud base only. “Beam at or above cloud base” is necessary for the radar to see precipitation, but not sufficient; the beam can pass over a shallow deck.
Drop drift. The storm pages integrate the sounding wind between beam height and the ground with a 7 m/s fall speed, typical of 2 mm raindrops. Snowflakes fall near 1 m/s and drift several times farther.
The Beam Lab, county table, town table and precipitation study use the research site elevations (KCCX 2,486 ft, KLWX 282 ft, KDIX 149 ft, KBGM 1,606 ft, KPBZ 1,213 ft, KDOX 50 ft, KBUF 693 ft, KCLE 763 ft MSL). The 800 m coverage map uses NCEI station-list elevations, which run about 80 to 120 ft higher for most sites. The difference shifts heights by the same amount and does not change which radar is lowest.
ASOS p01i is a running total since the last routine :56 observation. Hourly totals here are the maximum p01i within each clock hour. Summing every report inflated totals 1.5 to 2.5 times in this dataset.
ASOS observations: Iowa Environmental Mesonet. Radiosondes: IEM RAOB archive (IAD, OKX; 00Z and 12Z). RAP BUFKIT soundings: IEM mtarchive. Radar sites: NCEI NEXRAD station list. Boundaries: U.S. Census Bureau cartographic boundaries (2025). Grid: 800 m cells in EPSG:5070 (NAD83 Conus Albers), centroid-in-state rule.
These files sit in the data/ folder next to this page on the website.
The atlas is one self-contained HTML file. Upload it next to the data/ folder and link to it, or embed it inside an existing page with this snippet.
Deep links open a tab directly, for example radar-atlas.html#storms or #lab.