Integrating remotely sensed phenology observations in a multi-model land data assimilation system

Water, energy, and nutrients are the primary determinants of productivity in terrestrial ecosystems. In a changing climate, dynamic monitoring and local adaptation of management practices to changing resource levels are crucial to socio-ecosystem sustainability. A better characterization of terrestrial water, energy, and carbon cycles through the integration of observations into models at spatial and temporal scales conducive to decision making and adaptation responses are essential. Recent advances in remote sensing, terrestrial carbon and phenology modeling, and data assimilation techniques provide the tools we need to make significant progress towards enhancing our understanding of terrestrial cycles and phenology dynamics.

In this project we implement an innovative terrestrial phenology data assimilation technique to integrate carbon-cycle observations into an ensemble modeling framework that will produce terrestrial carbon-water-energy reanalyses over the North American Land Data Assimilation System (NLDAS) domain at 1/8° and hourly spatial and temporal resolution. Specifically, we evaluate the potential of assimilating phenology observations in land data assimilation system by constraining the modeled terrestrial carbon dynamics with remotely sensed observations of vegetation condition, water and energy availability (e.g., albedo, leaf area index, fraction of photosynthetically active radiation, and gross primary production). We also assess the efficiency of a multi-model ensemble assimilation technique by including four land surface models that have a dynamic vegetation phenology component within the assimilation system. The multi-model ensemble will allow incorporating the uncertainty in the physical processes, which is still an unresolved assimilation issue. Furthermore, we use remotely sensed water, energy, and vegetation observations to dynamically minimize systematic biases, using automatic parameter calibration procedures, and state initialization errors, using an ensemble land data assimilation system.

Besides answering outstanding science questions, this work will generate terrestrial carbon reanalyses from 1980 to present to be used in weather and climate forecast models and will provide sensitivity and uncertainty analyses to understand the limitations and constraints of the proposed approach. The Land Information System (LIS) is used to integrate remotely sensed phenology observations, ensemble modeling, uncertainty/sensitivity analysis, and data assimilation capabilities to quantify the contemporary terrestrial carbon budget and associated uncertainties and their contributions from different sources.

Maps of Leaf Area Index (LAI) across the Contigous United States as estimated by a land surface model (top), a satellite product (bottom), and a land data assimilation system that merges the model estimate with the satellite observation (center).

Funded by NASA ROSES 2016 Modeling, Analysis, and Prediction (MAP)

Team: Dr. Maggioni (PI); Dr. Houser (co-I); Dr. Sauer (co-I); Dr. Zhang (Postdoctoral Fellow); Ms. Rahman (Graduate Research Assistant)

​Performance Period: August 2017 – July 2021