Spatially explicit information on cropland use intensity is vital for monitoring land and water resource demands in agricultural systems. Cropping practices underlie substantial spatial and temporal variability, which can be captured through the analysis of image time series. Temporal binning helps to overcome limitations concerning operability and repeatability for mapping large areas and can improve the thematic detail and consistency of maps in agricultural systems. We here assessed the use of annual, quarterly, and eight-day temporal features for mapping five cropping practices on annual croplands across Turkey. We used 2403 atmospherically corrected and topographically normalized Landsat Collection 1 L1TP images of 2015 to compute quarterly best-pixel composites, quarterly and annual spectral-temporal metrics, as well as gap-filled eight-day time series of Tasseled Cap components. We tested 22 feature sets for binary cropland mapping, and subsequent discrimination of five cropping practices: Spring and winter cropping, summer cropping, semi-aquatic cropping, double cropping, and greenhouse cultivation. We evaluated area-adjusted accuracies and compared cropland area estimates at the province-level with official statistics. We achieved overall accuracies above 90%, when using either all quarterly features or the eight-day Tasseled Cap time series, indicating that temporal binning of intra-annual image time-series into multiple temporal features improves representations of cropping practices. Class accuracies of winter and spring, summer, and double cropping were robust, while omission errors for semi-aquatic cropping and greenhouse cultivation were high. Our mapped cropland extent was in good agreement with province-level statistics (r2 = 0.85, RMSE = 7.2%). Our results indicate that 71.3% (±2.3%) of Turkey’s annual croplands were cultivated during winter and spring, 15.8% (±2.2%) during summer, while 8.5% (±1.6%) were double-cropped, 4% (±1.9%) were cultivated under semi-aquatic conditions, and 0.32% (±0.2%) was greenhouse cultivation. Our study presents an open and readily available framework for detailed cropland mapping over large areas, which bears the potential to inform assessments of land use intensity, as well as land and water resource demands.

Open and analysis-ready data, as well as methodological and technical advancements have resulted in an unprecedented capability for observing the Earth’s land surfaces. Over 10 years ago, Landsat time series analyses were inevitably limited to a few expensive images from carefully selected acquisition dates. Yet, such a static selection may have introduced uncertainties when spatial or inter-annual variability in seasonal vegetation growth were large. As seminal pre-open-data-era papers are still heavily cited, variations of their workflows are still widely used, too. Thus, here we quantitatively assessed the level of agreement between an approach using carefully selected images and a state-of-the-art analysis that uses all available images. We reproduced a representative case study from the year 2003 that for the first time used annual Landsat time series to assess long-term vegetation dynamics in a semi-arid Mediterranean ecosystem in Crete, Greece. We replicated this assessment using all available data paired with a time series method based on land surface phenology metrics. Results differed fundamentally because the volatile timing of statically selected images relative to the phenological cycle introduced systematic uncertainty. We further applied lessons learned to arrive at a more nuanced and information-enriched vegetation dynamics description by decomposing vegetation cover into woody and herbaceous components, followed by a syndrome-based classification of change and trend parameters. This allowed for a more reliable interpretation of vegetation changes and even permitted us to disentangle certain land-use change processes with opposite trajectories in the vegetation components that were not observable when solely analyzing total vegetation cover. The long-term budget of net cover change revealed that vegetation cover of both components has increased at large and that this process was mainly driven by gradual processes. We conclude that study designs based on static image selection strategies should be critically evaluated in the light of current data availability, analytical capabilities, and with regards to the ecosystem under investigation. We recommend using all available data and taking advantage of phenology-based approaches that remove the selection bias and hence reduce uncertainties in results.

Pixel quality (PQ) products delivered with Analysis Ready Data (ARD) provide users with information about the conditions of the surface, atmosphere, and sensor at the time of acquisition. Knowing whether an observation was affected by clouds or sensor saturation is crucial when selecting data to include in automated analysis, as imperfect or erroneous observations are undesirable for most applications. There is, however, a certain rate of commission error in cloud detection, and saturation may not affect all spectral bands at a time, which can lead to suitable observations being excluded. This can have a substantial impact on the amount of data available for analysis. To understand how different surface types can affect cloud commission and saturation, we analyzed cloud and per-band saturation PQ flags for 31 years of Landsat data within Digital Earth Australia. Areas showing substantial reduction in observation density compared to their surroundings were investigated to characterize how specific surface types impact on the temporal density of observations deemed desirable. Using Fmask 3.2 by way of example, our approach demonstrates a method that can be applied to summarize the characteristics of cloud-screening algorithms and sensor saturation. Results indicate that cloud commission and sensor saturation rates show specific characteristics depending on the targets under observation. This potentially leads to an imbalance in data availability driven by surface type in a given study area. Based on our findings, the level of detail in PQ flags delivered with ARD is pivotal in maximizing the potential of EO data.

No abstract

No abstract