Evaluating the impacts of models, data density and irregularity on reconstructing and forecasting dense Landsat time series

Abstract

We evaluated the performance of a variety of time series models, that include the harmonic (HR) model, autoregressive (AR) model, linear Gaussian state-space (LGSS) model, cubic spline (SP) model, double logistic (DL) model, and asymmetric Gaussian (AG) model, for reconstructing (all six models) and forecasting (HR, AR, and LGSS models) dense Landsats 5–7 time series based on 4562 samples. To remove the impact of land change and human interventions on data reconstruction and forecasting, this evaluation excluded croplands and samples changed between 2000 and 2011. Results show that the widely used HR model is not a good model for data reconstruction but outperforms other models in data forecasting. The DL and AG models have the best performance in data reconstruction but cannot forecast Landsat observations. The AR and LGSS models shared similar performance in reconstructing Landsat data but are less ideal for data forecasting, particularly for the LGSS model. Integrating the HR (for forecasting) and DL or AG (for reconstruction) is recommended to improve land change detection and land cover classification results. We also evaluated the impact of data density and irregularity on reconstructing and forecasting Landsat observations. When the data density is low (<7 clear observations per year), the increase of data density can substantially improve the performance of data reconstruction and forecasting, but when the data density is higher than 7 and less than 17, the noise in the data dominates the results, and slightly lower reconstruction and forecasting accuracy is observed. When the data density is higher than 17, model performance improves with the increase of data density again. Therefore, we recommend analyzing Landsat time series for places with data density higher than 7 clear observations per year if possible. On the other hand, data irregularity has a moderate impact on data reconstruction and forecasting. When the irregularity is less than 1, the smaller the irregularity the better the performance in both data reconstruction and forecasting, and when the irregularity is higher than 2.5, it will have a more substantial negative impact. Therefore, time series analysis using Landsat data with an irregularity less than 1 is generally recommended, and time series with an irregularity larger than 2.5 should be avoided if possible.