This paper presents a novel spatio-temporal LSTM (SPATIAL) architecture for time series forecasting applied to environmental datasets. The framework was applied for three different ocean datasets: current speed, temperature, and dissolved oxygen. Network implementation proceeded in two directions that are nominally separated but connected as part of a natural environmental system – across the spatial (between individual sensors) and temporal dimensions of the sensor data. Data from twenty ocean sensors were used to train the model. Results were compared against four baseline models: two machine learning algorithms generated by robust autoML frameworks, and two deep neural networks based on CNN and LSTM, respectively. Results demonstrated ability to accurately replicate complex signals and provide comparable performance to state-of-the-art benchmarks. Learning from multiple sensors simultaneously increased robustness to missing data. This paper addresses two fundamental challenges related to environmental applications of machine learning: 1) data sparsity, particularly in a challenging ocean environment, and 2) environmental datasets are inherently connected in the spatial and temporal directions while classical ML approaches only consider one of these at a time. Furthermore, sharing of parameters across all input steps makes SPATIAL a fast, scalable, and easily-parameterized forecasting framework.