Integration of Landsat time series and field plots for forest productivity estimates in decision support models

Abstract

Forests provide an array of services ranging from forest products to greenhouse gas absorption. The reliability and appropriateness of management decisions concerning ecosystem services and resources are directly correlated with the accuracy and extent of the available information on forests and related dynamics. Historically, forest resource management activities informed by decision support systems, have largely relied on field measurements, often with sparse geographical, temporal, and ecological coverage. In contrast, remotely-sensed data can provide dense temporal coverage over large areas. Landsat data in particular have spatial and temporal characteristics well-suited to capturing forest changes on the landscape. Since 2008 free and open access to analysis-ready data representing more than 40years of data in the Landsat archive has created new opportunities for forest monitoring. There is however, a need to link forest information generated from remotely-sensed data to the data input needs of decision support systems. Here, we develop such a linkage between field and remotely-sensed observations to characterize one of the main drivers of forest change: growth. The goal of this research is to produce above-ground biomass (AGB) change estimates that integrate the information of re-measured field plots and that of the Landsat time series. Over a large test area (∼3.3Mha) in Saskatchewan, Canada, we demonstrate how information from a dense Landsat time series (1984–2012) can be used with a network of field plots to estimate change in AGB over the broad geographical, temporal, and ecological range of these forests. Our approach expands forest growth estimation both spatially and temporally, and results in information layers that are directly ingestible in decision support models. Our results show that trends match between the field plots and the remote-sensing-field-combined estimate, but that the more encompassing spatial estimates give more varied and lower average estimates of AGB change for our study region. We hypothesize that integrating both remotely-sensed information and repeated field measurements covers a wider range of ecological conditions, forest productivity, and forest developmental stages present on the landscape and through time than the field plots alone. The next step in our research is to integrate our findings into the forest resource management decision support tools, with the aim of increasingly spatializing what are usually spatially-referenced models. Our findings highlight emergent synergies between complementary data sources that foster the generation of new information products for forest resources management decision support.