Jns2183

I was at Hershey Park for 7 hours Saturday. This weather would have been much preferable. Mid 80's and the sun beating in those lines was not enjoyable. At least I didn't go Friday and have to deal with that giant fight. Sent from my SM-S731U using Tapatalk

My Magnus Opus of why I can never be mad at preceiptitation reports after a big storms.. We've all been there. NEXRAD shows 0.50" for your pixel. Your CoCoRaHS gauge reads 0.22". Your neighbor two blocks over posts 0.81". Both of you are right. The radar might also be right. Here's why, and why it gets genuinely alarming during intense convection. The Geometry First I'm 73.8 miles from KCCX at State College. At that range, the 1° beam has expanded to 6,803 feet in diameter — that's 1.29 miles across. The center of that beam is floating 7,882 feet above my roof. The super-resolution pixel covering my house is 0.830 square miles / 2.15 km². The legacy pixel was 1.660 square miles / 4.30 km². That single dBZ value stamped on that pixel represents the average backscatter of a volume of air roughly the size of 12 Disneylands hovering a mile and a half above the ground. Now imagine 100 top-of-the-line rain gauges evenly distributed across that pixel. Super-res spacing works out to one gauge every 481 feet. I stand dead-center in my grid cell. By Pythagoras, my four nearest gauges are at the corners — each 340 feet away from me. Here's Where It Gets Uncomfortable For well-behaved stratiform rain — your warm-frontal overrunning, your classic nimbostratus — everything is fine. CV of 0.10 to 0.16. My four nearest gauges within a few hundredths of each other. Radar is defensible. Life is good. But bump up into Category 7 — intense convective core directly over the pixel — and the math falls apart spectacularly. Radar reports 0.50". CV = 1.00. Standard deviation = 0.50". Lognormal distribution gives a pixel-wide range of 0.090" to 1.391" at the P5/P95 level. That's a 15:1 spread across my 0.830 square mile pixel, between gauges that are less than half a mile from each other. But here's the part that genuinely made me sit back: my four nearest gauges — 340 feet away from me, about the length of a city block — have a realistic P10/P90 spread of 0.250" to 0.810". Three-to-one variation between gauges I can practically see from my porch. Both are physically correct measurements of the same event. Scale that up to Category 8 — extreme/flash flood/training cells — and it becomes almost philosophical. Radar reports 0.50". Pixel range P5/P95: 0.052" to 1.613". That's a 31:1 ratio. My four nearest gauges: 0.157" to 0.965". Six-to-one. The gauge 340 feet north of me reads under two tenths. The gauge 340 feet east reads close to an inch. The radar says half an inch. Every single instrument is functioning perfectly. At 1.00" reported by the radar the numbers get worse: Cat 7 gives my nearest four a spread of 0.500" to 1.620" and Cat 8 gives 0.314" to 1.930". The gauge closest to me physically could read less than a third of an inch while the one a city block away reads nearly two inches. During the same event. At the same time. Why This Matters for Verification Work This isn't a measurement error problem. This isn't instrument quality. This is the fundamental physics of convective precipitation at sub-kilometer scales, documented by Peleg et al. (2013) with a dense gauge network and confirmed repeatedly in the literature. Rainfall within a single radar pixel during convection can vary by over 100% — and that finding came from a network covering a smaller area than my NEXRAD pixel at 73.8 miles. When someone posts their CoCoRaHS total and it's half what radar showed during a summer storm, the correct response isn't "your gauge is off." The correct response is "yes, that is entirely expected and physically consistent with everything we know about convective precipitation structure." The radar isn't wrong. Your gauge isn't wrong. The atmosphere just doesn't care about our need for tidy single-number summaries. The only honest statement you can make about precipitation during a Cat 7/8 event with a single gauge is: this is what fell at this exact point. Full stop. Extrapolating to even the next block over is an act of faith, not measurement. Methodology: Lognormal distribution fit to literature CV values by precip type (Ciach & Krajewski 1999, 2006; Peleg et al. 2013; Jensen & Pedersen 2005). Local near-gauge CV compressed by exponential spatial correlation model. KCCX beam geometry calculated using 4/3 Earth radius refraction. All pixel areas computed from actual Camp Hill–KCCX range of 73.8 miles. Sent from my SM-S731U using Tapatalk

A Few Things That Should Make You Uncomfortable At 100 ft — 62,351 ft³. Roughly the size of a large house. This is the only range where the radar is sampling anything resembling a point. At 1 mile — 92 ft beam diameter, still smaller than a football field. Still reasonable. Still vaguely honest. At 10 miles — beam is now 922 ft wide. Bigger than the range gate itself. The pancake starts to make sense. At 65 miles (Camp Hill) — ~22 billion ft³. The beam diameter (~5,990 ft) is now wider than it is deep by a factor of 7. Your "data point" is a disc, not a cube. And it's floating at ~5,200 ft AGL. At 100 miles — 54.9 billion ft³, beam nearly 1.75 miles wide, center beam at 9,708 ft AGL. At this point the radar is essentially sampling the lower stratosphere and calling it precipitation data. Depth is always 822 feet. It took me a number of ever increasing pancake sizes and a traffic cone of unsavory origins to visualize this. This pulse and about 50;more with 97% overlap get run through a bunch of algorithms and out pops a pixel over us. So if radar makes you want to pull your hair out, this is why. America’s Boeing Everett Factory, 25 miles north of Seattle, remains the world’s largest building by volume and world’s largest factory. Workers assemble Boeing aircraft—including the new 787 Dreamliner—within this 472,000,000-cubic-foot factory. My single radar pixel contains roughly 47 Boeing Everett Factories worth of air Sent from my SM-S731U using Tapatalk

The atmosphere is a terrible heater and a spectacular refrigerator, and it's been this way for 4.5 billion years with zero apologies. Seriously, haha, what happens when air heats up. It rises and cools. One state requires energy in, another requires nothing Sent from my SM-S731U using Tapatalk

Here's the addition to drop into the message board post, after the site link and before the feedback section: --- **Lowest Radar Beam by County — South-Central PA Reference** For those who want to know what's actually looking at their county, here are the lowest 0.5° beam heights for the counties covered by this study, sorted alphabetically. Listed as best radar and beam height, with the runner-up in parentheses: **Adams** — KLWX: 4,680 ft MSL (KCCX: 9,241 ft) **Berks** — KDIX: 7,642 ft MSL (KCCX: 10,418 ft) **Cumberland** — KCCX: 7,235 ft MSL (KLWX: 7,377 ft) **Dauphin** — KCCX: 8,029 ft MSL (KLWX: 10,267 ft) **Franklin** — KLWX: 4,902 ft MSL (KCCX: 7,605 ft) **Lancaster** — KDIX: 9,420 ft MSL (KDOX: 8,523 ft) **Lebanon** — KCCX: 10,155 ft MSL (KDIX: 10,763 ft) **Perry** — KCCX: 5,498 ft MSL (KLWX: 8,789 ft) **Schuylkill** — KDIX: 10,008 ft MSL (KCCX: 10,418 ft) **York** — KLWX: 5,874 ft MSL (KDOX: 9,530 ft) A few things worth noting on these numbers. First, these are county-wide averages computed from five representative coordinate points distributed across each county they represent the typical beam height across the bulk of the county's land area, not any specific location within it. This matters especially in south-central PA where the terrain is anything but uniform and several counties are genuinely split between radar coverage zones. Adams, Cumberland, and Lancaster in particular have portions of their land area where a different radar may actually be lower than the county average suggests in some cases up to three different radar sites each hold the lowest beam advantage over some piece of a single county depending on which direction you're standing in it. Second, and this trips people up: these heights are above mean sea level, not above your head. To get the actual distance between you and the beam you need to subtract your own elevation. If you're at 500 feet in the Cumberland Valley and the KCCX 0.5° beam shows 7,235 ft MSL over Cumberland County, the beam is roughly 6,735 feet above you — not 7,235. In the ridge-and-valley terrain to the north and west where elevations run 1,200 to 1,600 feet, that same beam is considerably closer to the surface than the MSL number implies. It's the AGL figure that tells you whether the beam is actually seeing your weather. Sent from my SM-S731U using Tapatalk

Central PA In-Depth Series — NEXRAD Beam Coverage & Precipitation Analysis** For those who haven't been following along, I've been working on a series of data-driven deep dives into the weather observation landscape of central Pennsylvania, the kind of local, granular analysis that doesn't get done because it doesn't scale nationally. Each project starts with a question that's been nagging at me as a 35-year observer of this stretch of the Susquehanna Valley and Ridge-and-Valley terrain. The two earlier projects in the series are still up and active. The first is the **MDT vs. CXY Bias Deep Dive** a two-era instrumentation study confirming the ~11% precipitation bias between Harrisburg International and Capital City Airport is real and physical, not an instrumentation artifact, spanning cooperative observer records back to 1948 through 2025 ASOS data: https://jns182wx.github.io/kmdt-cxy-analysis/ The second is the **South-Central PA 25 Years of Winter Weather Deep Dive** a comprehensive climatological look at winter precipitation across our region covering storm catalogs, snow ratio verification, RSI maps, and terrain-aware elevation bias decomposition across the ridge-and-valley province: https://jns182wx.github.io/winter-weather/ --- **The New Project — NEXRAD Beam Geometry & Precipitation Coverage** The question that drove this one: when it rains in south-central Pennsylvania, how much of that rainfall does NEXRAD radar actually see? The geometry alone is striking, the 0.5° beam from KCCX clears 8,000 to 9,000 feet above ground level by the time it reaches the Harrisburg corridor, sitting well above the cloud base during the shallow stratiform rain, drizzle, and cold air damming events that define our winter and early spring precipitation climatology. To answer this properly I pulled three months of hourly ASOS data from ten south-central PA stations, paired each precipitation hour with rawinsonde thermodynamics from IEM, and computed precise beam heights using the standard 4/3 Earth-radius WSR-88D formula for all four radars covering us, KCCX, KLWX, KDIX, and KBGM, selecting the best radar per station by actual geometry rather than assumed proximity. Along the way I found a critical data quality issue that affects anyone using IEM's ASOS archive: the p01i field is a running cumulative total since the last routine observation, not a per-period amount. Naive summation of sub-hourly SPECI observations inflates event totals by 1.5 to 2.5 times. Every precipitation number in this project uses the corrected method. The findings are instructive but also honest about the limits of what surface observations alone can tell us. For the Harrisburg corridor stations the beam is provably below the cloud base, sampling clear air, during a meaningful fraction of rainy hours. For the majority of precipitation hours the beam is at or above the cloud base, but without cloud-top data we can't say whether it's sampling the precipitating layer or overshooting it entirely. What the rawinsonde thermodynamics add is character, the February 20th CAD event shows the freezing level within 300 feet of the beam height, textbook conditions for bright band contamination. January 25th shows the opposite: deep warm frontal overrunning with 105-knot bulk shear and the beam solidly inside a saturated column, exactly when radar performs well. The five storm deep dives each pull hourly RAP BUFKIT soundings for all ten stations, compute hodographs, storm tilt, raindrop trajectory, dry air intrusion, and wet-bulb zero, and render a west-to-east atmospheric cross-section showing where the beam intersects the thermodynamic structure of the storm. Full site with all interactive tools and downloadable data here: https://jns182wx.github.io/radar_site/ --- **Feedback, Corrections, and What's Next** If you spot an error in the beam geometry, a station assignment that looks wrong, a precipitation total that doesn't match your records, or a methodology choice you'd push back on, please say so. Same goes for the earlier projects. Local knowledge improves local science and if you've been observing in this part of PA for any length of time your instincts about what the data should look like are worth hearing. If you have ideas for sections to add to any of these , additional storm events, verification against CoCoRaHS or COOP records, cross-sections to other radar sites, those conversations are welcome too. The next project is going to need community input to be worthwhile so I want to put it out there now. I just got my **Ecowitt GW90 "Fatboy"** up and running and in that spirit of personal weather stations, the next deep dive is going to look at the PWS landscape across south-central PA: station distribution, hardware and software variety, network density, and most importantly bias estimation and quality weighting against official observations. The core idea is to take any location, wherever you live or work, drop a 5, 10, 15, and 20km radius around your coordinates, and map every PWS, ASOS, CoCoRaHS, COOP, and DCP station in each ring with historical bias estimates where the record supports them. Think miniaturized personal MADIS for your specific spot. Long term the goal is to integrate that localized ground truth with NBM (which is taking over NWS short-range forecasting) and MRMS (radar-derived QPE) to arrive at a reliable blended daily ground truth that's actually calibrated to where you are rather than the nearest official station. This requires a significant amount of manual data collection to do right, so I want to direct the effort toward people for whom it will be genuinely useful. **If you want to be part of this, respond here or DM me with your town name or coordinates.** At minimum you'll get a full report on your local station landscape with whatever bias estimates the historical record supports, within the next month. And if you have ideas for other locally-rooted projects , anything about the weather in your corner of central PA that nobody seems to have a good quantitative answer for, I'd genuinely love to hear them. Sent from my SM-S731U using Tapatalk

Last year the single worst day for Storms were the ones that developed as a backdoor cold front came through the area. Storm motion was to the south west. It did a bunch of wind damage. Does anyone have the dates for the storm? Sent from my SM-S731U using Tapatalk

The Fatboy awakens!!! Plus a manual rain gauge. I maybe have the worst yard ever for a weather station. But she will do. Even with a garage blocking me to the north I still managed a 18.8mph gust last night Sent from my SM-S731U using Tapatalk

That 68 wind guest should be 28:for Harrisburg, haha Sent from my SM-S731U using Tapatalk

MOS WHACK A MOLE I've learned way way too much about this over the last couple months so now you get to also. If you ever wonder why a new version of the GFS is in training forever it's MOS, it is always MOS MOS is basically not “an equation.” It’s a bloated empire of site-by-site statistical band-aids. At one station like MDT, a single MOS package can involve hundreds to over a thousand equations once you break out forecast hour, variable, and category thresholds. Then double it for warm vs. cool season. Now spread that across 2,000 sites and multiple guidance systems, and you’re dealing with millions of regression equations. So when model physics gets changed upstream, MDL doesn’t get a fun little tune-up. They get to play nationwide statistical whack-a-mole because half their bias-correction patchwork may now be off. That’s the whole game. MY next little project until I recover from winter storm data burnout is one I'm almost halfway through. A fun little deep dive into the various MOS for the the GFS, NAM, HRRR, NBM and a 10 year (8 year for NBM);statistical analysis of forcasting skill by them for KMDT using about 25 metrics from NWS and Research Papers for every variable that can be seen, computer, derived from data record of KMDT. And think. This skill is after a couple thousand linear regression equations to brow beat NAMs drunken benders Sent from my SM-S731U using Tapatalk

Thank you. I have to fix the stuff on the RSI page to actually show all records I have. I downloaded our top 10 storms from dive in CSV format but git hub co-pilot wanted to just keep the 4 station theme going and I was in no mood to fix after 5 hours at 1:30am. Eventually I'll try to get all the storms I can between RSI and Cocorahs. All accumulation I have are from kmdt for most part due to them actually keeping track of snow there. Small airports don't really have a point person for that. Other than that I downloaded every single hourly and daily record I could from those 4 sites over the years. 500mb of csv files, thank God for paraquet formats. Sent from my SM-S731U using Tapatalk

I finally finished cateloging all winter storms greater than 3" since 2000 till end of 2025 77 storms A bunch of other winter stuff is there too. I think it's a good start https://jns182wx.github.io/winter-weather/ Sent from my SM-S731U using Tapatalk

I guess I should add York. Sent from my SM-S731U using Tapatalk

You say that now, but just wait till we get locked in the 50's half of May. On the years ahead of us from work they finish a plus 1.7 and exactly normal for April and May. Although they did have April's with mean in 40s and may mean In 50's Sent from my SM-S731U using Tapatalk

If you ever asked the question "I wonder what the local ASOS Stations at KCXY, KMDT, KTHV, KLNS record for wind when they have an kind of Snow" I have the fun answers below. Local terrain and microclimates making themselves known. Especially the wind tunnel that is the Susquehanna River Sent from my SM-S731U using Tapatalk