Lightning strike trend prediction with GBM in R
Lightning activity is projected to increase with climate change. Lightning activity is interesting to model with stochastic gradient boosting (GBM: generalized boosted regression models/gradient boosting machine) in R. One use I have for this at SNAP is in the context of landscape fire modeling with SNAP’s ALFRESCO model. The simulations from the model can be enhanced by incorporating information about lightning strike activity over Alaska which varies both spatially and temporally.
The plot above reveals the upward projected trend in lightning strike frequency over Alaska, predominantly interior boreal forest in this case. A preliminary model suggests a 17% increase in lightning strikes per decade on average. More will be shared in a future update.