WORKFLOW_LIST.md

Workflow List

We group FarmVibes.AI workflows in the following categories:

• Data Ingestion: workflows that download and preprocess data from a particular source, preparing data to be the starting point for most of the other workflows in the platform. This includes raw data sources (e.g., Sentinel 1 and 2, LandSat, CropDataLayer) as well as the SpaceEye cloud-removal model;
• Data Processing: workflows that transform data into different data types (e.g., computing NDVI/MSAVI/Methane indexes, aggregating mean/max/min statistics of rasters, timeseries aggregation);
• FarmAI: composed workflows (data ingestion + processing) whose outputs enable FarmAI scenarios (e.g., predicting conservation practices, estimating soil carbon sequestration, identifying methane leakage);
• ForestAI: composed workflows (data ingestion + processing) whose outputs enable ForestAI scenarios (e.g., detecting forest change, estimating forest extent);
• ML: machine learning-related workflows to train, evaluate, and infer models within the FarmVibes.AI platform (e.g., dataset creation, inference);

Below is a list of all available workflows within the FarmVibes.AI platform. For each of them, we provide a brief description and a link to the corresponding documentation page.

---

data_ingestion

• admag/admag_seasonal_field 📄: Generates SeasonalFieldInformation using ADMAg (Microsoft Azure Data Manager for Agriculture).

• admag/prescriptions 📄: Fetches prescriptions using ADMAg (Microsoft Azure Data Manager for Agriculture).

• airbus/airbus_download 📄: Downloads available AirBus imagery for the input geometry and time range.

• airbus/airbus_price 📄: Prices available AirBus imagery for the input geometry and time range.

• alos/alos_forest_extent_download 📄: Downloads Advanced Land Observing Satellite (ALOS) forest/non-forest classification map.

• alos/alos_forest_extent_download_merge 📄: Downloads Advanced Land Observing Satellite (ALOS) forest/non-forest classification map and merges it into a single raster.

• bing/basemap_download 📄: Downloads Bing Maps basemaps.

• bing/basemap_download_merge 📄: Downloads Bing Maps basemap tiles and merges them into a single raster.

• cdl/download_cdl 📄: Downloads crop classes maps in the continental USA for the input time range.

• dem/download_dem 📄: Downloads digital elevation map tiles that intersect with the input geometry and time range.

• gedi/download_gedi 📄: Downloads GEDI products for the input region and time range.

• gedi/download_gedi_rh100 📄: Downloads L2B GEDI products and extracts RH100 variables.

• glad/glad_forest_extent_download 📄: Downloads Global Land Analysis (GLAD) forest extent data.

• glad/glad_forest_extent_download_merge 📄: Downloads the tiles from Global Land Analysis (GLAD) forest data that intersect with the user input geometry and time range, and merges them into a single raster.

• gnatsgo/download_gnatsgo 📄: Downloads gNATSGO raster data that intersect with the input geometry and time range.

• hansen/hansen_forest_change_download 📄: Downloads and merges Global Forest Change (Hansen) rasters that intersect the user-provided geometry/time range.

• landsat/preprocess_landsat 📄: Downloads and preprocesses LANDSAT tiles that intersect with the input geometry and time range.

• modis/download_modis_surface_reflectance 📄: Downloads MODIS 8-day surface reflectance rasters that intersect with the input geometry and time range.

• modis/download_modis_vegetation_index 📄: Downloads MODIS 16-day vegetation index products that intersect with the input geometry and time range.

• naip/download_naip 📄: Downloads NAIP tiles that intersect with the input geometry and time range.

• osm_road_geometries 📄: Downloads road geometry for input region from Open Street Maps.

• sentinel1/preprocess_s1 📄: Downloads and preprocesses tiles of Sentinel-1 imagery that intersect with the input Sentinel-2 products in the input time range.

• sentinel2/cloud_ensemble 📄: Computes the cloud probability of a Sentinel-2 L2A raster using an ensemble of five cloud segmentation models.

• sentinel2/improve_cloud_mask 📄: Improves cloud masks by merging the product cloud mask with cloud and shadow masks computed by machine learning segmentation models.

• sentinel2/improve_cloud_mask_ensemble 📄: Improves cloud masks by merging the product cloud mask with cloud and shadow masks computed by an ensemble of machine learning segmentation models.

• sentinel2/preprocess_s2 📄: Downloads and preprocesses Sentinel-2 imagery that covers the input geometry and time range.

• sentinel2/preprocess_s2_ensemble_masks 📄: Downloads and preprocesses Sentinel-2 imagery that covers the input geometry and time range, and computes improved cloud masks using an ensemble of cloud and shadow segmentation models.

• sentinel2/preprocess_s2_improved_masks 📄: Downloads and preprocesses Sentinel-2 imagery that covers the input geometry and time range, and computes improved cloud masks using cloud and shadow segmentation models.

• soil/soilgrids 📄: Downloads digital soil mapping information from SoilGrids for the input geometry.

• soil/usda 📄: Downloads USDA soil classification raster.

• spaceeye/spaceeye 📄: Runs the SpaceEye cloud removal pipeline, yielding daily cloud-free images for the input geometry and time range.

• spaceeye/spaceeye_inference 📄: Performs SpaceEye inference to generate daily cloud-free images given Sentinel data and cloud masks.

• spaceeye/spaceeye_interpolation 📄: Runs the SpaceEye cloud removal pipeline using an interpolation-based algorithm, yielding daily cloud-free images for the input geometry and time range.

• spaceeye/spaceeye_interpolation_inference 📄: Performs temporal damped interpolation to generate daily cloud-free images given Sentinel-2 data and cloud masks.

• spaceeye/spaceeye_preprocess 📄: Runs the SpaceEye preprocessing pipeline.

• spaceeye/spaceeye_preprocess_ensemble 📄: Runs the SpaceEye preprocessing pipeline with an ensemble of cloud segmentation models.

• user_data/ingest_geometry 📄: Adds user geometries into the cluster storage, allowing for them to be used on workflows.

• user_data/ingest_raster 📄: Adds user rasters into the cluster storage, allowing for them to be used on workflows.

• user_data/ingest_smb 📄: Adds user rasters into the cluster storage from an SMB share, allowing for them to be used on workflows.

• weather/download_chirps 📄: Downloads accumulated precipitation data from the CHIRPS dataset.

• weather/download_era5 📄: Hourly estimated weather variables.

• weather/download_era5_monthly 📄: Monthly estimated weather variables.

• weather/download_gridmet 📄: Daily surface meteorological properties from GridMET.

• weather/download_herbie 📄: Downloads forecast data for provided location & time range using herbie python package.

• weather/download_terraclimate 📄: Monthly climate and hydroclimate properties from TerraClimate.

• weather/get_ambient_weather 📄: Downloads weather data from an Ambient Weather station.

• weather/get_forecast 📄: Downloads weather forecast data from NOAA Global Forecast System (GFS) for the input time range.

• weather/herbie_forecast 📄: Downloads forecast observations for provided location & time range using herbie python package.

data_processing

• chunk_onnx/chunk_onnx 📄: Runs an Onnx model over all rasters in the input to produce a single raster.

• chunk_onnx/chunk_onnx_sequence 📄: Runs an Onnx model over all rasters in the input to produce a single raster.

• clip/clip 📄: Performs a soft clip on an input raster based on a provided reference geometry.

• gradient/raster_gradient 📄: Computes the gradient of each band of the input raster with a Sobel operator.

• heatmap/classification 📄: Utilizes input Sentinel-2 satellite imagery & the sensor samples as labeled data that contain nutrient information (Nitrogen, Carbon, pH, Phosphorus) to train a model using Random Forest classifier. The inference operation predicts nutrients in soil for the chosen farm boundary.

• index/index 📄: Computes an index from the bands of an input raster.

• linear_trend/chunked_linear_trend 📄: Computes the pixel-wise linear trend of a list of rasters (e.g. NDVI).

• merge/match_merge_to_ref 📄: Resamples input rasters to the reference rasters' grid.

• outlier/detect_outlier 📄: Fits a single-component Gaussian Mixture Model (GMM) over input data to detect outliers according to the threshold parameter.

• threshold/threshold_raster 📄: Thresholds values of the input raster if higher than the threshold parameter.

• timeseries/timeseries_aggregation 📄: Computes the mean, standard deviation, maximum, and minimum values of all regions of the raster and aggregates them into a timeseries.

• timeseries/timeseries_masked_aggregation 📄: Computes the mean, standard deviation, maximum, and minimum values of all regions of the raster considered by the mask and aggregates them into a timeseries.

farm_ai

• agriculture/canopy_cover 📄: Estimates pixel-wise canopy cover for a region and date.

• agriculture/change_detection 📄: Identifies changes/outliers over NDVI across dates.

• agriculture/emergence_summary 📄: Calculates emergence statistics using thresholded MSAVI (mean, standard deviation, maximum and minimum) for the input geometry and time range.

• agriculture/green_house_gas_fluxes 📄: Computes Green House Fluxes for a region and date range

• agriculture/heatmap_using_classification 📄: The workflow generates a nutrient heatmap for samples provided by user by downloading the samples from user input.

• agriculture/heatmap_using_classification_admag 📄: This workflow integrate the ADMAG API to download prescriptions and generate heatmap.

• agriculture/heatmap_using_neighboring_data_points 📄: Creates heatmap using the neighbors by performing spatial interpolation operations. It utilizes soil information collected at optimal sensor/sample locations and downloaded sentinel satellite imagery.

• agriculture/methane_index 📄: Computes methane index from ultra emitters for a region and date range.

• agriculture/ndvi_summary 📄: Calculates NDVI statistics (mean, standard deviation, maximum and minimum) for the input geometry and time range.

• agriculture/weed_detection 📄: Generates shape files for similarly colored regions in the input raster.

• carbon_local/admag_carbon_integration 📄: Computes the offset amount of carbon that would be sequestered in a seasonal field using Microsoft Azure Data Manager for Agriculture (ADMAg) data.

• carbon_local/carbon_whatif 📄: Computes the offset amount of carbon that would be sequestered in a seasonal field using the baseline (historical) and scenario (time range interested in) information.

• land_cover_mapping/conservation_practices 📄: Identifies conservation practices (terraces and grassed waterways) using elevation data.

• land_degradation/landsat_ndvi_trend 📄: Estimates a linear trend over NDVI computer over LANDSAT tiles that intersect with the input geometry and time range.

• land_degradation/ndvi_linear_trend 📄: Computes the pixel-wise NDVI linear trend over the input raster.

• segmentation/segment_basemap 📄: Downloads basemap with BingMaps API and runs Segment Anything Model (SAM) over them with points and/or bounding boxes as prompts.

• segmentation/segment_s2 📄: Downloads Sentinel-2 imagery and runs Segment Anything Model (SAM) over them with points and/or bounding boxes as prompts.

• sensor/optimal_locations 📄: Identify optimal locations by performing clustering operation using Gaussian Mixture model on computed raster indices.

• water/irrigation_classification 📄: Develops 30m pixel-wise irrigation probability map.

forest_ai

• deforestation/alos_trend_detection 📄: Detects increase/decrease trends in forest pixel levels over the user-input geometry and time range for the ALOS forest map.

• deforestation/ordinal_trend_detection 📄: Detects increase/decrease trends in the pixel levels over the user-input geometry and time range.

ml

• crop_segmentation 📄: Runs a crop segmentation model based on NDVI from SpaceEye imagery along the year.

• dataset_generation/datagen_crop_segmentation 📄: Generates a dataset for crop segmentation, based on NDVI raster and Crop Data Layer (CDL) maps.

• driveway_detection 📄: Detects driveways in front of houses.

• segment_anything/basemap_prompt_segmentation 📄: Runs Segment Anything Model (SAM) over BingMaps basemap rasters with points and/or bounding boxes as prompts.

• segment_anything/s2_prompt_segmentation 📄: Runs Segment Anything Model (SAM) over Sentinel-2 rasters with points and/or bounding boxes as prompts.