# Off-grid solar and battery sizing: 10 proposed Sanders County MeshCore sites
Generated Nov 22, 2026. Files: power_sizing.csv (all columns), power_chart.png, model.py / fetch.py (reproducible; raw API responses in raw/).

## Result
| Site | Elev m | Tilt | Dec PSH mean/severe | Horizon loss Dec | Design low °C | Wh/day nom/design | Panel W | Battery Ah | Autonomy d (actual) | Baseline 12Ah/20W |
|---|---|---|---|---|---|---|---|---|---|---|
| Patricks Knob | 2086 | 70° | 2.47 / 1.56 | 9.5% (S-sector 13.4°) | -35.9 | 3.9 / 10.56 | 30 | 20 | 10 (13.6) | FAIL (nominal load: PASS) |
| Clark Mountain | 1545 | 70° | 2.6 / 1.62 | 0.0% (S-sector 2.7°) | -33.2 | 3.9 / 10.56 | 20 | 20 | 10 (13.6) | FAIL (nominal load: PASS) |
| Eddy Mountain | 2121 | 70° | 2.62 / 1.65 | 0.0% (S-sector 1.1°) | -35.9 | 3.9 / 10.56 | 20 | 20 | 10 (13.6) | FAIL (nominal load: PASS) |
| Green Mountain (summit) | 1650 | 70° | 2.48 / 1.61 | 0.0% (S-sector 2.7°) | -35.0 | 3.9 / 10.56 | 20 | 20 | 10 (13.6) | FAIL (nominal load: PASS) |
| Rattlesnake Butte | 1132 | 70° | 2.73 / 1.74 | 0.4% (S-sector 6.5°) | -37.3 | 3.9 / 10.56 | 20 | 20 | 10 (13.6) | FAIL (nominal load: PASS) |
| Perma BNSF | 1477 | 70° | 2.78 / 1.74 | 0.7% (S-sector 6.9°) | -37.3 | 3.9 / 10.56 | 20 | 20 | 10 (13.6) | FAIL (nominal load: PASS) |
| Plains ridge tower (ASR A0866791) | 1167 | 70° | 2.65 / 1.69 | 0.7% (S-sector 3.8°) | -35.9 | 3.9 / 10.56 | 20 | 12 | 7 (8.1) | PASS (nominal load: PASS) |
| Noxon Rapids dam tower (ASR A1246210) | 711 | 65° | 2.01 / 1.34 | 18.6% (S-sector 22.9°) | -35.0 | 3.9 / 10.56 | 30 | 12 | 7 (8.1) | FAIL (nominal load: PASS) |
| Star Peak | 1874 | 70° | 2.27 / 1.4 | 0.0% (S-sector 0.0°) | -35.0 | 3.9 / 10.56 | 30 | 20 | 10 (13.6) | FAIL (nominal load: PASS) |
| Heron home node (hypothetical) | 705 | 70° | 2.24 / 1.38 | 1.3% (S-sector 6.1°) | -33.0 | 3.9 / 10.56 | 20 | 12 | 7 (8.1) | PASS (nominal load: PASS) |

- Baseline (12 V 12 Ah LiFePO4 + 20 W) **passes only at Plains ridge tower and Heron home node** (7-day target). It fails elsewhere mainly on **battery autonomy** (12 Ah gives 8.1 days against a 10-day target) and, at Patricks Knob, Noxon dam tower and Star Peak, on **panel** (severe-December harvest below load plus 14-day refill). Clark/Eddy/Green/Rattlesnake/Perma are within a few percent on the panel; battery is what fails.
- With the nominal (non-busy) load of 3.9 Wh/day the baseline is judged separately in the CSV (`baseline_ok_nominal_load`).
- Recommendation: 20 Ah + 20 W at Clark, Eddy, Green, Rattlesnake, Perma; 20 Ah + 30 W at Patricks Knob and Star Peak; 12 Ah + 30 W at Noxon dam tower; 12 Ah + 20 W at Plains ridge tower and Heron home node. Plains/Heron panel margin is thin (about 0 to 25%).
- Worst month is December at all 10 sites.

## Method
1. **Sun.** PVGIS API v5.3 PVcalc (ERA5 radiation, `usehorizon=1`, DEM-calculated horizon), 1 kWp, loss 0, azimuth south, tilts 45/55/60/65/70/75/80/90. Used `H(i)_d` (kWh/m²/day = peak sun hours) and the PVGIS interannual `SD_m`. Best December tilt: 70° (65° at Noxon, where the canyon horizon penalizes steeper angles).
   - PVGIS-ERA5 horizontal December is 0.98-1.16, NASA POWER (SYN1DEG, 25-yr climatology) 0.82-0.89. To be conservative, PVGIS tilted values are multiplied by NASA/PVGIS horizontal (0.78-0.85) ("bias factor" column).
   - "Severe" December = mean × bias × (1 − 1.28·SD/mean) (about the 10th-percentile December, assuming normal).
   - Horizon shading = PVGIS with horizon vs. without, Dec, same tilt.
2. **Load** (battery-side, 12 V bus):
   - Nominal: node 7.5 mA base at 3.7 V (powersaving; 6-9 mA, MeshCore PRs #1238/#2286 and IoTThinks fork) + 1% TX; buck 75%.
   - Busy/design: 20 mA continuous RX (RAK4631 datasheet LoRa RX 17 mA, plus MCU) + 5% TX at 130 mA; buck 85%.
   - TX current: RAK4631 datasheet 125 mA @ 20 dBm; 130 mA at 22 dBm is an **assumption, verify** (T114 datasheet 164 mA @ 20 dBm module-level; board choice changes this).
   - Charge controller 10 mA @ 12 V (**assumption**, user range 5-20 mA; verify datasheet of the controller chosen): 2.88 Wh/day.
   - Cold-charge allowance +5 Wh/day (**assumption**) where design low ≤ −20 °C (all sites).
   - Design load: 5.56 + 5 = 10.56 Wh/day; nominal 3.90 Wh/day.
3. **Panel.** W = Wh/day ÷ (severe PSH × 0.72 × 0.80) × (1 + N/14). 0.72 = system losses (**assumption**, 28%: MPPT, wiring, cold/hot, battery charge efficiency); 0.80 = snow/rime/soiling (**assumption**); the N/14 factor refills the autonomy bank in 14 days (**assumption**). Round to 10/20/30/50/80/100 W.
4. **Battery.** Wh = Wh/day × N ÷ (0.80 DoD × cold derate); Ah at 12.8 V. N = 10 days (remote summits), 7 days for Plains tower, Noxon dam tower, Heron home node (**assumption**: easier access). Cold derate 0.70 (design low ≤ −25 °C), 0.80 (≤ −15), else 0.90 (**assumptions, unverified**; get the pack's low-temperature discharge curve). Round to 6/12/20/30/50/100 Ah.
5. **Temperature.** Design low = colder of (a) minimum of PVGIS ERA5 hourly T2m 2005-2023 and (b) NASA POWER monthly T2M_MIN at the point. These are grid-cell values and are **not elevation-corrected** (summits are likely colder in radiation nights; valleys and Heron have inversions). **Verify** with NRCS SNOTEL stations and NOAA climate normals near each site. The −35 °C figures also agree with the hardware report's −30 °F design assumption.
6. **Cold charging.** LiFePO4 must not be charged below 0 °C (vendor pages, hardware report: LiTime 6 Ah cuts charge at 0 °C; Epoch heated pack charges to −20 °C and needs ≥0.5 A charger; Victron 75/10 default 5 °C low-temp cutoff). Use a low-temp-charge or self-heating pack in an insulated box, with a battery temperature sensor. Heater energy: raising a ~1.3 kg pack by 25 K ≈ 9 Wh (assumed mass, c ≈ 1 kJ/kg·K), so a few warm-ups per day fit the +5 Wh/day allowance only if the box is well insulated. Continuous heating of an insulated box (order 5-8 W for a 25 cm EPS box at ΔT 30 K, k = 0.035 W/m·K assumed) is not affordable on 20-30 W panels in December. Quantity per pack: **verify with the vendor**.

## Shading and caveats
- Horizon from PVGIS is terrain only: no trees, buildings, towers or snow-laden branches. Noxon dam tower loses 18.6% of Dec sun to terrain (south horizon up to 22.9°). Patricks Knob loses 9.5% (up to 13.4°). Heron home node loses only 1.3% to terrain but is at valley floor with a hypothetical house site: trees and buildings could dominate, **verify on site**.
- Interannual December SD is 14-20% of the mean; a stormy year or weeks of rime ice can be worse than modeled.
- Existing tower sites (Plains ridge, Noxon dam, Perma BNSF) may have a host that offers power or restricts panels; the assumption here is no commercial power.
- NREL PVWatts (DEMO_KEY) gave no response from the box; cross-check at pvwatts.nrel.gov. Controller, battery and panel part prices are not quoted: re-quote before budgeting.
- Panels are specified at steep tilt (65-70°) and south-facing; check wind loading and mounting on each tower or mast.

## Sources
- PVGIS API v5.3: https://re.jrc.ec.europa.eu/api/v5_3/PVcalc, /seriescalc, /printhorizon (ERA5, DEM horizon); docs https://joint-research-centre.ec.europa.eu/pvgis-photovoltaic-geographical-information-system_en
- NASA POWER climatology API: https://power.larc.nasa.gov/api/temporal/climatology/point (ALLSKY_SFC_SW_DWN, T2M_MIN)
- NREL PVWatts v8 (not retrieved): https://developer.nrel.gov/docs/solar/pvwatts/v8/
- RAK4631 datasheet (via hardware report): https://docs.rakwireless.com ; Heltec T114: https://resource.heltec.cn
- MeshCore PRs #1238, #2286: https://github.com/meshcore-dev/MeshCore ; IoTThinks EasySkyMesh: https://github.com/IoTThinks/EasySkyMesh
- Battery and controller specs: vendor pages cited in /workspace/meshcore-hw/report.md (LiTime, Epoch, Enjoybot, Victron manual)
- To verify: NRCS SNOTEL https://www.nrcs.usda.gov/wps/portal/wcc/home/snowClimateMonitoring/snowpack/snotelSites ; NOAA normals https://www.ncei.noaa.gov/products/land-based-station/us-climate-normals

## Budget
Repriced from this sizing: see BUDGET.md and bom.csv. Note the battery is upsized to 20 Ah at every site (no 12 Ah cold-charge LiFePO4 found).
