NDVI Time Series Module¶

File: computing/misc/ndvi_time_series.py

Overview¶

This module generates NDVI (Normalized Difference Vegetation Index) time series data for micro-watersheds. It processes vegetation index data by land cover class (crop, tree, shrub) and creates temporal aggregations for analysis.

Purpose¶

Creates vegetation index time series for: - Agricultural monitoring - Forest health assessment - Vegetation change detection - Distributes to GEE, GeoServer, and database

Architecture¶

flowchart TD
    A[Celery Task Triggered] --> B[ee_initialize]
    B --> C{Has state/district/block?}
    C -->|Yes| D[Build Standard ROI]
    C -->|No| E[Use Custom ROI]
    D --> F[Load MWS FeatureCollection]
    E --> G[Load Custom FeatureCollection]
    F --> H{Asset Exists?}
    G --> H
    H -->|Yes| I[Get Last Date from Existing]
    H -->|No| J[Use Start Date]
    I --> K{Need Update?}
    K -->|Yes| L[Generate New Data]
    K -->|No| M[Skip Generation]
    J --> L
    L --> N[Build Class Assets]
    N --> O[For Each Class: crop/tree/shrub]
    O --> P[save_layer_info_to_db]
    P --> Q[make_asset_public]
    Q --> R[sync_fc_to_geoserver]
    R --> S[Update Sync Status]
    S --> T[Return True]

Components¶

Main Task: `ndvi_timeseries()`¶

Location: Line 34

Parameters: | Parameter | Type | Description | |-----------|------|-------------| | state | str, optional | State name | | district | str, optional | District name | | block | str, optional | Block/tehsil name | | roi | ee.FeatureCollection, optional | Custom ROI | | asset_suffix | str, optional | Custom asset suffix | | asset_folder_list | list, optional | Custom asset folder | | start_year | int | Start year (fiscal: July 1) | | end_year | int | End year (fiscal: June 30) | | app_type | str | Application type (default: "MWS") | | gee_account_id | int, optional | GEE account identifier |

Returns: bool - Whether layer was successfully synced to GeoServer

Helper Functions¶

`extract_class_fc()`¶

Location: Line 172

Extracts features for a specific land cover class from yearly NDVI asset.

`build_final_class_asset()`¶

Location: Line 196

Merges yearly NDVI assets into a single time series per class.

`_generate_data()`¶

Location: Line 243

Generates yearly NDVI data with LULC masking.

`get_last_date()`¶

Location: Line 365

Extracts the last date from existing NDVI asset for incremental updates.

Processing Logic¶

1. Fiscal Year Definition¶

start_date = f"{start_year}-07-01"   # July 1 of start year
end_date = f"{end_year+1}-06-30"     # June 30 of end year

2. LULC Masking¶

NDVI is masked by land cover class:

flowchart LR
    A[LULC Image] --> B[Extract Class Masks]
    B --> C[Crop Mask: classes 8,9,10,11]
    B --> D[Tree Mask: class 6]
    B --> E[Shrub Mask: class 12]
    C --> F[Apply to NDVI]
    D --> F
    E --> F

LULC Classes: | Class | Description | |-------|-------------| | 6 | Tree cover | | 8, 9, 10, 11 | Crop types | | 12 | Shrub cover |

3. NDVI Processing¶

# Get interpolated NDVI time series
ndvi = get_padded_ndvi_ts_image(f_start_date_str, f_end_date_str, roi, 14)

# Add masked bands for each class
def add_masked_bands(img):
    nd = img.select("gapfilled_NDVI_lsc")
    date = img.date().format("YYYY-MM-dd")

    return ee.Image.cat([
        nd.updateMask(crop_mask).rename(ee.String("crop_").cat(date)),
        nd.updateMask(tree_mask).rename(ee.String("tree_").cat(date)),
        nd.updateMask(shrub_mask).rename(ee.String("shrub_").cat(date)),
    ])

4. Zonal Statistics¶

reduced = ndvi_band_stack.reduceRegions(
    collection=roi.select(["uid"]),
    reducer=ee.Reducer.mean(),
    scale=30,
    tileScale=4
)

5. Yearly Processing¶

For multi-year time series, processes each year separately:

flowchart TD
    A[Start Date] --> B{Year Loop}
    B --> C[Process Year]
    C --> D[Load LULC for Year]
    D --> E[Generate NDVI TS]
    E --> F[Export Yearly Asset]
    F --> G{More Years?}
    G -->|Yes| H[Advance 365 days]
    H --> B
    G -->|No| I[Merge All Years]

Integration Points¶

computing/misc/ndvi_time_series.py
├── computing.misc.hls_interpolated_ndvi
│   └── get_padded_ndvi_ts_image()  # NDVI time series generation
├── computing.utils
│   ├── get_layer_object()          # Layer metadata retrieval
│   ├── save_layer_info_to_db()     # Database persistence
│   ├── sync_layer_to_geoserver()   # GeoServer sync
│   ├── update_layer_sync_status()  # Status tracking
│   └── sync_fc_to_geoserver()      # GeoServer sync
├── utilities.gee_utils
│   ├── ee_initialize()             # GEE authentication
│   ├── valid_gee_text()            # Text sanitization
│   ├── get_gee_dir_path()          # Directory path generation
│   ├── export_vector_asset_to_gee() # Vector export
│   ├── check_task_status()         # Task monitoring
│   ├── is_gee_asset_exists()       # Asset existence check
│   ├── merge_fc_into_existing_fc() # FeatureCollection merge
│   ├── make_asset_public()         # ACL management
│   ├── create_gee_dir()            # GEE directory creation
│   └── build_gee_helper_paths()    # Path building
├── utilities.constants
│   └── GEE_PATHS                   # GEE path configurations
└── gee_computing.models
    └── GEEAccount                  # GEE account model

Output¶

Three separate assets are generated per block:

Class	Asset Name	Description
Crop	`ndvi_timeseries_{district}_{block}_crop`	Agricultural vegetation
Tree	`ndvi_timeseries_{district}_{block}_tree`	Forest/tree vegetation
Shrub	`ndvi_timeseries_{district}_{block}_shrub`	Shrubland vegetation

GeoServer Workspace: ndvi_timeseries

Dataset Name: NDVI Timeseries

Feature Properties¶

Each MWS feature contains time series properties:

{
  "uid": "unique_mws_id",
  "crop_2023-07-01": 0.65,
  "crop_2023-07-15": 0.68,
  "crop_2023-08-01": 0.72,
  ...
  "tree_2023-07-01": 0.45,
  "tree_2023-07-15": 0.47,
  ...
  "shrub_2023-07-01": 0.32,
  ...
}

Usage¶

from computing.misc.ndvi_time_series import ndvi_timeseries

result = ndvi_timeseries.delay(
    state="Rajasthan",
    district="Jaipur",
    block="Sanganer",
    start_year=2022,
    end_year=2023,
    gee_account_id=1
)

Incremental Updates¶

The module supports incremental updates: 1. Checks if asset already exists 2. Retrieves last date from existing asset 3. Only generates data for new time periods 4. Merges new data with existing data

Dependencies¶

ee (Google Earth Engine Python API)
Celery - Distributed task queue
datetime - Date handling
gee_computing.models.GEEAccount - GEE account management
computing.misc.hls_interpolated_ndvi - NDVI processing utilities