Atmosphere Demo 1 Documentation (Fourth Edition · Demo 1 Final)

1. Project Overview

Atmosphere is a smart monitoring and early-warning system for flash floods and debris flows at outdoor streams / drifting sites.

Demo 1 implements: soil moisture, flow, and wind sensing + pump PWM speed control + web monitoring + early-warning model.

2. Hardware Connections

Main controller: Arduino Nano (ATmega328P, new bootloader)

Soil moisture sensor: AO->A2 (analog), DO->D2 (digital / disconnection detect), VCC->5V, GND->GND

Flow sensor YF-S401: Red->5V, Yellow->D8 (signal), Black->GND

Pump: IO to D9 (PWM pin, soft-start / speed control, higher duty = faster); VCC/GND to power, D7 pulled LOW as fallback

Wind sensor (3-cup, 0~5V): on another computer's Arduino UNO A0, forwarded to this machine via HTTP

3. Software Implementation

Two files (both included in this folder):

FlowSensor.ino — Arduino firmware: three-channel sensing + pump PWM speed control (soft start), outputs one JSON line per second over serial (9600)

server.py — Python web service (port 8080): serial reading + wind HTTP integration + warning model + web page (with pump slider speed control + total power display)

Serial command: send speed value 0~255 + newline (0 = stop, 255 = full speed).

Pump speed control: D9 PWM + soft start (0→255 ramps smoothly over ~0.5s, smoothing out inrush current).

Web slider: drag 0~100% for real-time speed control (150ms debounce); total power = fixed 0.4W + duty cycle × pump full-load power (default 1.25W estimate, calibratable).

Wind endpoint: remote computer http://192.168.1.200:8001/wind returns JSON (wind_ms / wind_kmh / temp / hum).

4. Warning Model

Debris flow (fully implemented):

Danger index D = 0.40·S̄ + 0.30·Q̄ + 0.20·dQ̄ + 0.10·ΔS̄

S̄ normalized soil moisture, Q̄ normalized flow, dQ̄ flow spike, ΔS̄ soil-moisture rise

Gate: soil moisture < 60% (not saturated) is judged safe directly, does not enter D leveling

Special rules: high soil moisture + sustained flow rise -> escalate level

Flash flood (partially implemented):

Manning formula 70% (needs ultrasonic water-level sensor, not yet implemented)

Spike detection 30% (approximated by per-second flow change rate, sliding-baseline adaptive, >local P95 = high risk)

Sliding baseline (climate-adaptive):

A 24-hour sliding window uses P10/P95 percentiles to judge "how far above local normal"; activates after 2 hours of data.

Warning levels (Level 1 = most severe):

🟢 Normal: D < 0.45 (soil moisture < 60% is judged safe directly)

🟡 Level 3 Alert: 0.45 ≤ D < 0.55

🟠 Level 2 Alert: 0.55 ≤ D < 0.75

🔴 Level 1 Alert: D ≥ 0.75

5. How to Run

Compile and upload FlowSensor.ino to the Arduino Nano

Run: python3 -u server.py

Open http://localhost:8080 in a browser and drag the slider to adjust pump speed

6. Known Issues

Pump powered from the Arduino causes unstable supply and repeated USB resets at high speed (soft start only eases inrush) — needs an independent power supply

Total power is an estimate (no current sensor); full-load power needs calibration from the pump nameplate

Flash-flood Manning 70% missing ultrasonic water-level sensor, pending

Spike sliding baseline implemented; fallback fixed threshold (2.0 L/min/s) still needs field calibration