AusENDVI: A long-term NDVI dataset for Australia

AusENDVI (Australian Emprical NDVI) is a monthly, 5-km gridded estimate of NDVI across Australia from 1982-2022. It is built by calibrating and harmonising NOAA's Climate Data Record AVHRR NDVI data to MODIS MCD43A4 NDVI using a gradient boosting ensemble decision tree method.  Additionally, the datasets are gapfilled using a synthetic NDVI dataset.  The methods are extensively described in an Earth System Science Data publication.

AusENDVI consists of several datasets, each dataset has a description in the attributes of the NetCDF file that describes its provenance.  The naming convention is "AusENDVI___.nc". 

AusENDVI-clim_gapfilled_1982_2013. Calibrated and harmonised Climate Data Record AVHRR NDVI data from Jan. 1982 to Dec. 2013. This version of the dataset used climate data in the calibration and harmonisation process and has the best agreement statistics with MODIS MCD43A4 NDVI. The dataset has been gap filled using the methods described in the accompanying publication.

AusENDVI-clim_MCD43A4_gapfilled_1982_2022. This dataset consists of calibrated and harmonised NOAA Climate Data Record AVHRR NDVI data from Jan. 1982 to Feb. 2000, joined with MODIS-MCD43A4 NDVI data from Mar. 2000 to Dec. 2022. This version of the dataset _used climate data_ in the calibration and harmonisation process. The dataset has been gapfilled using the methods described in the accompanying publication

AusENDVI-noclim_1982_2013. Calibrated and harmonised Climate Data Record AVHRR NDVI data from Jan. 1982 to Dec. 2013. This version of the dataset did not use climate data in the calibration and harmonisation process and the dataset has not been gap filled.

AusENDVI-synthetic_1982_2022. This dataset consists of synthetic NDVI data that was built by training a model on the joined _AusENDVI-clim_ and _MODIS-MCD43A4 NDVI_ timeseries using climate, woody-cover-fraction, and atmospheric CO2 as predictors. The synthetic NDVI is used for gap filling.

All datasets are in 'EPSG:4326' projection, and have a spatial resolution of 0.05 degrees. Geographic coordinate information is contained in the `spatial_ref` variable. 

A Jupyter Notebook is also provided that shows how to load, plot, QC mask, reproject, and gap-fill AusENDVI datasets. The notebook is effectively a 'readme' file.

The notebook is also available to view/download here

An open-source github repository details the methods used to create these datasets

https://github.com/cbur24/AusENDVI

A few small changes to the datasets were implemented in version 0.2.0:

All datasets now have their values clipped to the range 0-1

The AusENDVI-clim dataset is now gapfilled, and includes a QC layer

The merged AusENDVI-noclim_MCD43A4_1982_2022 dataset was removed to simplify the number of datasets included in the repository. Users who want to join the 'noclim' and MODIS datasets can do so by clipping out MCD43A4 from the AusENDVI-clim_MCD43A4_gapfilled_1982_2022 dataset.

The accompanying Jupyter Notebook 'readme' has been updated.