A super SDM (species distribution model) 'in the cloud' for better habitat-association inference with a 'big data' application of the Great Gray Owl for Alaska

The currently available distribution and range maps for the Great Grey Owl (GGOW; Strix nebulosa) are ambiguous, contradictory, imprecise, outdated, often hand-drawn and thus not quantified, not based on data or scientific. In this study, we present a proof of concept with a biological application for technical and biological workflow progress on latest global open access 'Big Data' sharing, Open-source methods of R and geographic information systems (OGIS and QGIS) assessed with six recent multi-evidence citizen-science sightings of the GGOW. This proposed workflow can be applied for quantified inference for any species-habitat model such as typically applied with species distribution models (SDMs). Using Random Forest-an ensemble-type model of Machine Learning following Leo Breiman's approach of inference from predictions-we present a Super SDM for GGOWs in Alaska running on Oracle Cloud Infrastructure (OCI). These Super SDMs were based on best publicly available data (410 occurrences + 1% new assessment sightings) and over 100 environmental GIS habitat predictors ('Big Data'). The compiled global open access data and the associated workflow overcome for the first time the limitations of traditionally used PC and laptops. It breaks new ground and has real-world implications for conservation and land management for GGOW, for Alaska, and for other species worldwide as a 'new' baseline. As this research field remains dynamic, Super SDMs can have limits, are not the ultimate and final statement on species-habitat associations yet, but they summarize all publicly available data and information on a topic in a quantified and testable fashion allowing fine-tuning and improvements as needed. At minimum, they allow for low-cost rapid assessment and a great leap forward to be more ecological and inclusive of all information at-hand. Using GGOWs, here we aim to correct the perception of this species towards a more inclusive, holistic, and scientifically correct assessment of this urban-adapted owl in the Anthropocene, rather than a mysterious wilderness-inhabiting species (aka 'Phantom of the North'). Such a Super SDM was never created for any bird species before and opens new perspectives for impact assessment policy and global sustainability.

Using an object-based machine learning ensemble approach to upscale evapotranspiration measured from eddy covariance towers in a subtropical wetland

Accurate prediction of evapotranspiration (ET) in wetlands is critical for understanding the coupling effects of water, carbon, and energy cycles in terrestrial ecosystems. Multiple years of eddy covariance (EC) tower ET measurements at five representative wetland ecosystems in the subtropical Big Cypress National Preserve (BCNP), Florida (USA) provide a unique opportunity to assess the performance of the Moderate Resolution Imaging Spectroradiometer (MODIS) ET operational product MOD16A2 and upscale tower measured ET to generate local/regional wetland ET maps. We developed an object-based machine learning ensemble approach to evaluate and map wetland ET by linking tower measured ET with key predictors from MODIS products and meteorological variables. The results showed MOD16A2 had poor performance in characterizing ET patterns and was unsatisfactory for estimating ET over four wetland communities where Nash-Sutcliffe model Efficiency (NSE) was less than 0.5. In contrast, the site-specific machine learning ensemble model had a high predictive power with a NSE larger than 0.75 across all EC sites. We mapped the ET rate for two distinctive seasons and quantified the prediction diversity to identify regions easier or more challenging to estimate from model-based analyses. An integration of MODIS products and other datasets through the machine learning upscaling paradigm is a promising tool for local wetland ET mapping to guide regional water resource management.

Bayesian machine learning ensemble approach to quantify model uncertainty in predicting groundwater storage change

Agricultural water demand, groundwater extraction, surface water delivery and climate have complex nonlinear relationships with groundwater storage in agricultural regions. As an alternative to elaborate computationally intensive physical models, machine learning methods are often adopted as surrogate to capture such complex relationships due to their high computational efficiency. Inevitably, using only one machine learning model is prone to underestimate prediction uncertainty and subjected to poor accuracy. This study presents a novel machine learning-based groundwater ensemble modeling framework in conjunction with a Bayesian model averaging approach to predict groundwater storage change and improve overall model predicting reliability. Three different machine learning models have been developed namely artificial neural network, support vector machine and response surface regression. To explicitly quantify uncertainty from machine learning model parameter and structure, Bayesian model averaging is employed to produce a forecast distribution associated with each machine learning prediction. Model weights and variances are obtained based on model performance to construct ensemble models. Then, the developed individual and Bayesian model averaging machine learning ensemble models are applied, evaluated and validated at different spatial scales including subregional and regional scales in an overdrafted agricultural region-the San Joaquin River Basin, through independent training and testing dataset. Results shows the machine learning models have remarkable predicting capability without sacrificing accuracy but with higher computational efficiency. Compared to a single-model approach, the ensemble model is able to produce consistently reliable predictions across the basin, yet it does not always outperform the best model in the ensemble. Additionally, model results suggest that groundwater pumping for agricultural irrigation is the primary driving force of groundwater storage change across the region. The modeling framework can serve as an alternative approach to simulating groundwater response, especially in those agricultural regions where lack of subsurface data hinders physically-based modeling.