Developing Sub-Seasonal to Seasonal Vegetation Greenness Outlooks

Vegetation greenness is an indicator of fuel moisture and plant stress and health. A product such as satellite remotely sensed normalized difference vegetation index (NDVI) is used, for example, in assessing short- and long-term conditions when developing the Predictive Services National Significant Wildland Fire Potential Outlooks. Wildland fire practitioners have expressed interest in having a sub-seasonal to seasonal (S2S) greenness outlook product. A sub-seasonal product, e.g., three weeks, could inform short-term resource allocations and monthly to seasonal scales could provide guidance for situational awareness and strategic planning. In either case this applies to both wildfire and prescribed fire.

This study explores the development of a prediction product for vegetation greenness across the U.S. using a combination of statistical analysis and advanced machine learning models. Historical monthly and seasonal climate predictors (e.g., ENSO, drought indices, climate anomalies) are used as training data to develop predictive relationships with greenness for those variables having the most prediction skill. This approach leverages the power of machine learning to analyze large amounts of data, uncovering patterns and relationships that may not be immediately apparent through traditional analysis methods. Once developed, operational model output such as CFSv2 can be used to supply input for making S2S greenness outlooks. In this presentation, the development of a machine learning algorithm capable of predicting 21-day NDVI from relevant variables will be discussed. We focus on this sub-seasonal timescale to provide an example of the process and in response to fire agency recommendations as a more immediate need to inform decisions.