geoskill-evapotranspiration-estimation

Priestley-Taylor 与简化 SEBAL 蒸散发估算，从净辐射/气温/LST/NDVI 计算 ET (mm/day)。Priestley-Taylor and simplified SEBAL evapotranspiration estimation from net radiation/air temperature/LST/NDVI. 输出 ET 栅格 + 统计 JSON。

蒸散发估算 | Evapotranspiration Estimation

Estimates regional evapotranspiration (ET, mm/day) from net radiation, air temperature, land surface temperature, and vegetation indices. Suitable for farmland irrigation water-demand assessment, watershed water-consumption analysis, drought monitoring, and land-surface process validation. Two methods are implemented:

• pt (Priestley-Taylor, 1972):

  text

  ET = α × Δ/(Δ + γ) × Rn × 0.408
  

  α ≈ 1.26 (empirical coefficient for adequately watered conditions); Δ is the slope of the saturation vapor pressure–temperature curve (kPa/°C, derived from air temperature via the Tetens formula es = 0.6108·exp(17.27T/(T+237.3))); γ ≈ 0.066 kPa/°C is the psychrometric constant; Rn is net radiation (MJ/m²/day); 0.408 converts MJ/m² to mm of water depth (latent heat). This method has a clear physical basis, requires only radiation and air temperature, and is suitable for large-area estimation.

• sebal (simplified empirical SEBAL): the evaporative fraction EF is built from NDVI and LST (EF = clip(NDVI_norm × (1 − LST_norm), 0, 1)), then ET = EF × Rn × 0.408. Dense vegetation with a cool surface yields high EF and high ET; bare land/urban heat islands yield low ET.

Outputs an ET raster (mm/day, EPSG:4326) and statistics JSON. --synthetic mode generates physically consistent Rn/T/LST/NDVI fields (vegetation on the left with high ET, bare land on the right with low ET); estimates should fall within the physically plausible range of 0–10 mm/day, and ET should be positively correlated with net radiation.

Dependencies / 依赖

bash

pip install numpy rasterio scipy

Usage / 使用方法

Basic usage (synthetic data, offline)

bash

python geoskill-evapotranspiration-estimation.py --bbox 116.0 39.0 117.0 40.0 --method pt --output-dir ./output

Example 1: Priestley-Taylor (synthetic data)

bash

python geoskill-evapotranspiration-estimation.py \
    --bbox 116.0 39.0 117.0 40.0 \
    --method pt \
    --synthetic \
    --output-dir ./et_pt

Example 2: simplified SEBAL

bash

python geoskill-evapotranspiration-estimation.py \
    --bbox 116.0 39.0 117.0 40.0 \
    --method sebal \
    --synthetic \
    --output-dir ./et_sebal

Example 3: real net radiation raster

bash

python geoskill-evapotranspiration-estimation.py \
    --input net_radiation.tif \
    --method pt \
    --output-dir ./et_real

(The input is net radiation Rn in MJ/m²/day; the accompanying meteorological/surface fields are generated synthetically to demonstrate the workflow.)

Example 4: custom Priestley-Taylor coefficient

bash

python geoskill-evapotranspiration-estimation.py \
    --bbox 116 39 117 40 \
    --method pt --alpha 1.10 \
    --synthetic --output-dir ./et_a11

Output / 输出

File	Format	Description
evapotranspiration.tif	GeoTIFF (float32)	Evapotranspiration ET (mm/day), EPSG:4326
et_stats.json	JSON	Method, parameters, ET mean/extremes/standard deviation, Rn mean
output-manifest.json	JSON	Run manifest (inputs/outputs/QA/software versions)

Data Source / 数据源 / Source

• Net radiation Rn: local GeoTIFF, or from MODIS / Landsat radiation balance and reanalysis data
• Air temperature T / LST / NDVI: generated in synthetic mode; real applications can ingest weather stations and MODIS LST/NDVI
• Synthetic mode: generated locally, no external data sources

Privacy / 隐私声明 / Privacy

• Runs fully offline by default and makes no network requests
• --synthetic mode reads no external data
• All computation is done locally; user data is never uploaded

License / License

MIT

---

---

name: geoskill-evapotranspiration-estimation description: 'Priestley-Taylor 与简化 SEBAL 蒸散发估算，从净辐射/气温/LST/NDVI 计算 ET (mm/day)。Priestley-Taylor and simplified SEBAL evapotranspiration estimation from net radiation/air temperature/LST/NDVI. 输出 ET 栅格 + 统计 JSON。'

蒸散发估算 | Evapotranspiration Estimation

从净辐射、气温、地表温度与植被指数估算区域蒸散发（ET，mm/day），适用于农田 灌溉需水评估、流域耗水分析、干旱监测、陆面过程验证等场景。实现两种方法：

• pt（Priestley-Taylor，1972）：

  text

  ET = α × Δ/(Δ + γ) × Rn × 0.408
  

  α≈1.26（充分供水经验系数），Δ 为饱和水汽压—温度曲线斜率（kPa/°C，由气温 经 Tetens 公式 es = 0.6108·exp(17.27T/(T+237.3)) 求得），γ≈0.066 kPa/°C 为干湿表常数，Rn 为净辐射（MJ/m²/day），0.408 为 MJ/m² → mm 水深的潜热换算。 该方法物理基础清晰，只需辐射与气温，适合大范围估算。

• sebal（简化 SEBAL 经验版）：用 NDVI 与 LST 构建蒸发比 EF （EF = clip(NDVI_norm × (1 − LST_norm), 0, 1)），再 ET = EF × Rn × 0.408。 植被茂密、地表凉爽处 EF 高、ET 高；裸地/城市热岛处 ET 低。

输出 ET 栅格（mm/day，EPSG:4326）与统计 JSON。支持 --synthetic 模式生成 物理一致的 Rn/T/LST/NDVI 场（左侧植被高 ET、右侧裸地低 ET），估算结果应落在 0–10 mm/day 的物理合理区间，且 ET 与净辐射正相关。

依赖

bash

pip install numpy rasterio scipy

使用方法

基本用法（合成数据，离线）

bash

python geoskill-evapotranspiration-estimation.py --bbox 116.0 39.0 117.0 40.0 --method pt --output-dir ./output

示例 1：Priestley-Taylor（合成数据）

bash

python geoskill-evapotranspiration-estimation.py \
    --bbox 116.0 39.0 117.0 40.0 \
    --method pt \
    --synthetic \
    --output-dir ./et_pt

示例 2：简化 SEBAL

bash

python geoskill-evapotranspiration-estimation.py \
    --bbox 116.0 39.0 117.0 40.0 \
    --method sebal \
    --synthetic \
    --output-dir ./et_sebal

示例 3：真实净辐射栅格

bash

python geoskill-evapotranspiration-estimation.py \
    --input net_radiation.tif \
    --method pt \
    --output-dir ./et_real

（输入为净辐射 Rn，MJ/m²/day；配套气象/地表场由合成生成以演示流程。）

示例 4：自定义 Priestley-Taylor 系数

bash

python geoskill-evapotranspiration-estimation.py \
    --bbox 116 39 117 40 \
    --method pt --alpha 1.10 \
    --synthetic --output-dir ./et_a11

输出

文件	格式	说明
evapotranspiration.tif	GeoTIFF (float32)	蒸散发 ET（mm/day），EPSG:4326
et_stats.json	JSON	方法、参数、ET 均值/极值/标准差、Rn 均值
output-manifest.json	JSON	运行清单（输入/输出/QA/软件版本）

数据源 / Source

• 净辐射 Rn：本地 GeoTIFF，或来自 MODIS / Landsat 辐射平衡、再分析资料
• 气温 T / LST / NDVI：合成模式生成；真实应用可接入气象站、MODIS LST/NDVI
• 合成模式：本地生成，无外部数据源

隐私声明 / Privacy

• 默认完全离线运行，不发起任何网络请求
• --synthetic 模式不读取任何外部数据
• 所有计算在本地完成，不上传用户数据

License

MIT