Towards Global Radiative Energy Mapping: Integrating Scalable Computation and Earth Observation

Reliable and moderate to fine resolution estimates of radiative energy are required for mapping solar energy potentials and understanding land-atmosphere interaction as well as ecosystem water use. However, time-series data records of moderate resolution estimates of radiative energy are not available and consequently, it is challenging to develop multi-year estimates of radiation at scales that are relevant for policy and decision-making. This limitation is primarily associated with the intensive computational requirement to effectively use time-series Earth observation data. To support the retrieval of global net radiation and consequent analysis of large amounts of spatio-temporal data, we integrate interactive tensor factorization and decomposition techniques with MODIS (Moderate Resolution Imaging Spectroradiometer) satellite radiances. This approach offers unique advantages for activity characterization in spatio-temporal and multi-relational data analysis. A shortwave and longwave radiation balance model is implemented using Tensorflow, an open source software library developed by Google Brain Team within Google’s Machine Intelligence research organization for numerical computation conducting machine learning and deep neural networks research using data flow graphs. TensorFlow uses a tensor data structure to represent all data, but general enough to be applicable in a wide variety of other domains as well. Comparison with net-radiation measurements from globally distributed sites showed that the net-radiation product agreed well with measurements across seasons and climate types.