Substantial declines in biodiversity over the past century demonstrate an immediate need to preserve ecosystems and further mitigate habitat loss. Monitoring changes in ecosystem condition at region thru continental to global scales can provide important information about biodiversity declines and help facilitate targeted intervention. Efforts to use satellite imagery to map ecosystem condition change have experienced challenges with distinguishing observed changes from the natural variation of ecosystems. In this study we use an innovative deep learning architecture to pair time series satellite imagery with locations of known on-ground condition. Our model was developed using 209,041 on-ground records of native species present in the landscape, as a surrogate measure of ecosystem condition, coupled with Landsat time series data and topographic and climatological datasets. We predict ecosystem condition across the Australian continent for several years (2010, 2015, 2020, 2021, 2022) at 100 m. Arid regions in Australia's interior had predicted condition scores close to reference condition (1) for all years. Comparatively, highly modified landscapes in Australia's southeastern and southwestern regions had predicted condition scores closer to fully degraded (0). Mean predicted ecosystem condition across Australia was greater than 0.65 for all years, suggesting greater overall presence of native species rather than absence, however this was spatially variable. Our results demonstrate that using deep learning techniques and time series data can provide quantitative information on ecosystem condition, accounting for temporal variability of vegetation phenology and spatial variability across bioregions. Ongoing efforts to collect essential biodiversity variables from space must consider integrating with deep leaning approaches that have capacity for context driven spatial modelling. This will help ensure mapping products can support policy and inform intervention strategies.

Satellite land surface temperature (Ts) provides valuable information on vegetation drought stress via its physical linkage to plant stomatal activity and transpiration. New-generation geostationary satellites offer opportunities to monitor sub-diurnal variations in Ts and thus track plant physiological stress response occurring at sub-daily timescales. Nevertheless, the potential of satellite Ts and its derived metrics for early detection of vegetation drought stress before visible canopy changes occur has not been widely assessed. Here, we developed a parsimonious Surface-Air Temperature Difference Anomaly (SATDA) method for tracking vegetation drought stress using the cumulative sub-diurnal difference from late-morning to early-afternoon between Ts from the Himawari8 geostationary satellite and hourly air temperature (Ta) from meteorological grids. SATDA utilised Ts - Ta as the physical driving gradient for sensible heat flux (H) to capture anomalous sensible heating due to reduced plant transpiration. We used SATDA to monitor the spatio-temporal patterns of the 2017-2019 Tinderbox Drought in southeast Australia. We benchmarked the skill of SATDA in forecasting visible drought-induced vegetation greenness decline against both conventional water availability-based indices (i.e., precipitation and soil moisture anomalies) and existing satellite Ts indices (i.e., Temperature Condition Index and Temperature Rise Index) across diverse climates and land covers. SATDA effectively captured a rapidly intensifying flash drought event at multi-week timescales (Jul to Sep 2019) embedded within the multi-year Tinderbox Drought, which contributed to detrimental impacts on agricultural production and increased wildfire risk. SATDA showed the best vegetation greenness forecast skill in the transitional semi-arid and sub-humid climates, with forecast correlation >0.5 at 32-day lead time. The advantage over water availability-based indices was more evident in woody-dominated ecosystems than herbaceous-dominated ecosystems, likely due to the importance of physiological regulations by trees during droughts such as deeper roots and stronger stomatal control. SATDA, based on Ts - Ta, showed overall better vegetation greenness forecasts than two Ts-only indices, especially in woody vegetation. Finally, SATDA showed consistently greater advantage over water availability-based and Ts-only indices in forecasting visible vegetation decline as the drought intensity increased. The parsimonious process-based SATDA method suits global-scale operational implementation to complement vegetation drought monitoring and early warning systems.

Cloud detection is a requisite step of almost all terrestrial applications using optical remote sensing imagery, as many applications are sensitive to cloud contamination. In this research, a new algorithm has been developed to detect nocturnal cloud (from sunset to sunrise) in Himawari-8/9 AHI (Advanced Himawari Imager) imagery over land, with simultaneous aims of maximising accuracy, simplicity and efficiency. The algorithm consists of two cloud detection methods: (i) proxy emissivity temporal variation, measured by pixel wise standard deviation within an hour; and (ii) monthly clear surface proxy emissivity database which is updated daily. Results from the two components are combined based on their respective confidence. A validation was conducted against 6 years of CALIPSO LiDAR data over the Australian continent, showing an overall accuracy of 96 %. The algorithm requires no ancillary data. It is also computationally efficient and so is suitable for near real-time (i.e., within 2 h) operation and can be readily adopted to similar operational geostationary sensors.

Background Knowledge of the distribution of wildland fuels across the landscape is necessary for the appropriate application of models used to support a broad range of fire management activities.Aims To develop an automated and nationally consistent method that generates up-to-date spatial fuel type information across Australia.Methods Data from various space-borne broad-band optical, LiDAR and radar sensors were combined with land use data to generate structural descriptions of vegetation that were then converted into fuel types.Key results An Australian fuel type spatial layer was generated using the Bushfire Fuel Classification fuel typology. Evaluation against field measurements revealed accuracies of 89 and 71% for native forest and non-forest fuel types, respectively. This product provides a higher level of spatial and structural detail than previously obtained by other national-level fuel classification approaches in Australia.Implications The developed fuel type layer is made available and can be readily used in research applications. The data also have use in supporting jurisdictional-level fuel mapping for a range of fire management applications, such as fire behaviour prediction, fire danger forecasting and risk assessment.

Context Understanding the functional role of vegetation across landscapes requires the ability to monitor tree and grass foliage cover dynamics. Several satellite-derived products describe total and woody foliage cover across Australia. Few of these are suitable for monitoring changes in woody foliage cover and only one can currently describe subseasonal dynamics in both woody and grass cover. Aims (1) To improve the accuracy of woody and grass foliage cover estimates in Australia's arid environments, around major disturbances and in perennially green pastures. (2) To gain a detailed understanding of the accuracy of woody and grass foliage cover estimates for Australia. Methods Satellite-derived greenness data were converted to total foliage cover fraction (0.0-1.0), accounting for differences in background soil affects. Total cover was split into component woody and grass cover by using a modified persistent-recurrent splitting algorithm. Results were compared with 4214 field measurements of cover. Key results Accuracy varied between woody and grassland vegetation types, with total, woody and grass foliage cover having low errors (of similar to 0.08) and near-zero biases across all woody vegetation types. Across grasslands, errors were higher (up to 0.28), and biases were greater (and negative), with both scaling with foliage density. Conclusions Foliage cover was accurately estimated for forested through to sparsely wooded ecosystems. Foliage cover of pure, dense grasslands was systematically underpredicted. Implications This is the only Australian cover product that can generate temporally dense woody and grass foliage cover data and is invaluable for monitoring vegetation dynamics, particularly across Australia's mixed tree-grass landscapes.

Nitrogen cycles control the structure, function, and composition of ecosystems globally. Despite their importance, our understanding of long‐term changes in global nitrogen cycles remains limited. The foliar nitrogen stable isotope ratio (δ 15 N) serves as a valuable metric for assessing changes in nitrogen cycling and potentially in plant nitrogen availability. However, existing observations of δ 15 N suffer from spatial bias and temporal discontinuity with contradictory findings across biomes, hindering our ability to detect and attribute drivers of change. Leveraging ground‐based observations as our calibration source, we derived annual maps of foliar δ 15 N spanning from 1984 to 2022 globally from Landsat spectra. We found that the Landsat‐derived δ 15 N effectively captured the observations, with an R 2 of 0.77 and a normalized root mean square error of 0.15. Globally, we found widespread temporal changes in δ 15 N with significant decreases for 44% and increases for 16% of vegetated ecosystems. Foliar δ 15 N mostly declined in forest ecosystems but increased in non‐forest land cover types. Gross primary productivity and its trend consistently explained spatiotemporal variation of δ 15 N globally, indicating increasing plant demand could lead to decreasing δ 15 N. Our study presents an innovative approach to effectively monitor and track potential changes in global nitrogen cycles over the past four decades, setting the stage for more impactful management and conservation strategies.

Effective satellite-based monitoring of ecosystem integrity or condition needs to address four key challenges: (a) context dependency; (b) alternative ecological states; (c) short-term temporal ecosystem dynamics; and (d) scarcity of reference data where ecosystems retain high levels of integrity. Here we present a typology, and outline strengths and weaknesses, of different approaches to mapping and monitoring ecosystem integrity across entire regions or continents using time series satellite data. We then describe how one of these approaches, the Habitat Condition Assessment System (HCAS), addresses all of the above challenges, and provide an outline of the evolved method which includes annual outputs, and Australian continent applications. HCAS requires three readily available inputs (i.e., representative examples of relatively natural areas as reference sites, remotely sensed ecosystem characteristics, and environmental covariate data) and could be easily adapted and applied by other countries to provide an effective indicator of ecosystem integrity for nature-based decisions.

Abstract Nitrogen (N) availability regulates the productivity of terrestrial plants and the ecological services they provide. There is evidence for both increasing and decreasing plant N availability in different biomes, but the data are fragmentary. How plant N availability responds to climate change, N deposition and increasing atmospheric CO2 concentration remains a major uncertainty in the projection of the terrestrial carbon sink. The foliar N stable isotope ratio (δ15N) is an indicator of plant N availability but its usefulness to infer long-term global patterns has been limited by data scarcity. Combining ground-based δ15N and Landsat spectra, we derived annual global maps of Landsat-based foliar δ15N as estimates of plant N availability during 1984-2022. We found significant decreases in plant N availability for 44% and increases in 16% of vegetated Earth’s surface with large spatial heterogeneity. Plant N availability declined in woody-dominated ecosystems but increased in herbaceous-dominated ones. These δ15N trends were consistently and negatively correlated with the trends of Normalised-Difference-Vegetation-Index as they varied across ecosystems, suggesting increasing plant cover could have led to decreasing plant N availability. Our results indicate possible future reductions in plant N availability in many terrestrial ecosystems and provide a useful way to monitor those changes globally.

The Australian dryland grain-cropping landscape occupies 60 Mha. The broader agricultural sector (farmers and agronomic advisors, grain handlers, commodity forecasters, input suppliers, insurance providers) required information at many spatial and temporal scales. Temporal scales included hindcasts, nowcasts and forecasts, at spatial scales ranging from sub-field to the continent. International crop-monitoring systems could not service the need of local industry for digital information on crop production estimates. Therefore, we combined a broad suite of satellite-based crop-mapping, crop-modelling and data-delivery techniques to create an integrated analytics system (Graincast™) that covers the Australian cropping landscape. In parallel with technical developments, a set of user requirements was identified through a human-centred design process, resulting in an end-product that delivered a viable crop-monitoring service to industry. This integrated analytics solution can now produce crop information at scale and on demand and can deliver the output via an application programming interface. The technology was designed to underpin digital agriculture developments for Australia. End-users are now using crop-monitoring data for operational purposes, and we argue that a vertically integrated data supply chain is required to develop crop-monitoring technology further.

Abstract The observed spatial and temporal dynamics in landscape and ecosystem resources are the net effects of natural processes and management activities. Monitoring the impact that humans have on these resources requires that these two sources of variability be partitioned, removing the natural variability to reveal variability due to management activities. Here, we present Compere, a relative benchmarking framework for monitoring the management‐driven impacts on ecosystem resources. The framework identifies locations in a region that share similar biophysical properties to a target location. Taking an attribute of interest, it then compares the value of the target location with those of all its biophysically equivalent locations, with any differences being attributed to variations in management. We provide an example application of Compere, using satellite‐derived vegetation cover data to examine the impact of hydrocarbon extraction activities on land condition (as described by vegetation cover) across the 490,000‐km2 Cooper Creek region in arid central Australia. Validation was performed by comparing land condition estimates to known disturbance patterns across the region. We found that the establishment of well sites was associated with between 12% and 41% decreases in land condition but that condition recovered to original levels by 6 years after establishment. We also found that fires had much larger and longer lasting effects on land condition across these landscapes, indicating that mining‐related activities that change fire patterns are likely to affect land condition most. Compere is a generic, flexible framework, providing a new capability for monitoring management impacts on multiple types of ecosystem resources.

Foresight of grain yields prior to harvest would be empowering for many stakeholders along the supply chain from farmers through to bulk handlers, banks and insurance companies. Estimating Australian grain production ahead of harvest is difficult for many reasons including the highly variable year to year rainfall. The rainfall in the final months prior to harvest, can be crucial to final harvest totals. Here we explore the importance of rainfall from September 1, which broadly corresponds to the close of the top-dressing fertilizer application window, for the remaining cropping season in determining final yield. This is assessed via sensitivity analysis of water-limited wheat potential yield totals from historical climate in the APSIM crop model. At locations where the rainfall influences wheat yield, we compare three methods to forecast wheat yields that differ based on the climate data input: 1) climatology approach, which uses 30 years of observed climate data, 2) analogue climatology, which uses information from climate drivers (El-Nino Southern Oscillation and Indian Ocean Dipole) to create analogue years; and 3) dynamical climate forecasts from a general circulation model (ACCESS-S). We find that potential yields strongly depend on in-season plant available water (PAW) where years with high PAW are unaffected by the late season rainfall. Predicting the potential yield from analogue climatology (climate drivers) had the greatest skill, with smallest Root Mean Squared Error of 0.45 t/ha. This approach ranked first for 42% of the study locations compared to the climatology and ACCESS-S forecasting methods. This knowledge can help inform decision makers about the need to incorporate seasonal climate forecasts and the most appropriate climate forecasting method.

This paper reviews information about field observations of vegetation productivity in Australia’s rangeland systems and identifies the need to establish a national initiative to collect net primary productivity (NPP) and biomass data for rangeland pastures. Productivity data are needed for vegetation and carbon model parameterisation, calibration and validation. Several methods can be used to estimate pasture productivity at various spatial and temporal scales, ranging from in situ measurements to satellite-based approaches and biogeochemical modelling. However, there is a barrier to implementing national vegetation and carbon modelling schemes because of the lack of digitised and readily available data derived from field observations, not because of the lack of modelling expertise. Our main goal in this paper is to explore the potential for consolidation of existing NPP and biomass databases for Australian rangelands. A protocol structure was proposed to establish a productivity database for Australia. The TERN (Terrestrial Ecosystems Research Network) national field data network for rangeland pasture productivity monitoring and modelling team could potentially coordinate the database. Government agencies and national and international research institutions could use the outputs from productivity models to inform greenhouse gas emissions and in measuring mitigation activities relevant for reporting against the United Nations’ Sustainable Development Goals and other international obligations. Other applications include monitoring fire danger, tracking ecological restoration and protection, and estimating fodder availability. Australian researchers have the tools needed to succeed in creating such a national database and a robust community of practice to curate it, enhance it and benefit from its availability.

The onus for monitoring crop growth from space is its ability to be applied anytime and anywhere, to produce crop yield estimates that are consistent at both the subfield scale for farming management strategies and the country level for national crop yield assessment. Historically, the requirements for satellites to successfully monitor crop growth and yield differed depending on the extent of the area being monitored. Diverging imaging capabilities can be reconciled by blending images from high-temporal-frequency (HTF) and high-spatial-resolution (HSR) sensors to produce images that possess both HTF and HSR characteristics across large areas. We evaluated the relative performance of Moderate Resolution Imaging Spectroradiometer (MODIS), Landsat, and blended imagery for crop yield estimates (2009–2015) using a carbon-turnover yield model deployed across the Australian cropping area. Based on the fraction of missing Landsat observations, we further developed a parsimonious framework to inform when and where blending is beneficial for nationwide crop yield prediction at a finer scale (i.e., the 25-m pixel resolution). Landsat provided the best yield predictions when no observations were missing, which occurred in 17% of the cropping area of Australia. Blending was preferred when <42% of Landsat observations were missing, which occurred in 33% of the cropping area of Australia. MODIS produced a lower prediction error when ≥42% of the Landsat images were missing (~50% of the cropping area). By identifying when and where blending outperforms predictions from either Landsat or MODIS, the proposed framework enables more accurate monitoring of biophysical processes and yields, while keeping computational costs low.

Since about 2010 there has been an explosion in the interest and expectations for data-driven agriculture, often dubbed ‘digital agriculture‘. Digital agriculture is often used interchangeably with the term ‘smart farming’, which refers to the use of data to inform farm decisions and then automation and actuation to execute those decisions. Several technological drivers have converged to bring about this interest (Koch 2017):

There is considerable demand for nationwide grain yield estimation during the cropping season by growers, grain marketers, grain handlers, agricultural businesses, and market brokers. In this paper, we developed a semi-empirical model (Crop-SI) to estimate the yield of the three major crops in the dryland Australian wheatbelt by combining a radiation use efficiency approach with meteorology driven Stress Indices (SI) at critical crop growth stages (e.g., anthesis and grain filling). These crop-specific SI (e.g., drought, heat and cold stress) help explain the impact of high spatial agro-environmental heterogeneity, which lead to substantial improvement in grain yield prediction. Crop-SI explains 87%, 69% and 83% of the observed field-scale grain yield variability with root mean square error of ~0.4, 0.4 and 0.5 t/ha for canola, wheat, and barley, respectively. At the pixel-level, Crop-SI reduces the relative error in grain yield estimation to 34%, 25%, and 20% for canola, wheat, barley, respectively, compared to two benchmark models. By incorporating water- and temperature-driven stresses, Crop-SI's predictive skill in highly variable environments is enhanced. As such, it paves the way for the next generation of agricultural systems models, knowledge products and decision support tools that need to operate at various scales.

Existing agricultural grain yield models predict yield at the field scale, or at regional scales (like districts and countries), but not both with consistent accuracy. Here we describe a scalable, satellite-based yield model called C-Crop. It is calibrated locally and so has field-scale accuracy. Its input data can be inferred remotely (namely crop type, foliage cover and air temperature) and so it can be potentially applied at any regional scale. We calibrated C-Crop using harvester-derived yield data for canola (31 field-years) and wheat (160 field-years), across the Australian cropping zone. C-Crop explained 69 and 68% of the observed variability in field-scale canola and wheat yields, respectively, with errors in the order of 33% and 32% of total yield. Given its simplicity, C-Crop is an effective model for estimating field-scale crop yields and has the potential to be applied across large regions.

Tree–grass savannas are a widespread biome and are highly valued for their ecosystem services. There is a need to understand the long‐term dynamics and meteorological drivers of both tree and grass productivity separately in order to successfully manage savannas in the future. This study investigated the interannual variability ( IAV ) of tree and grass gross primary productivity ( GPP ) by combining a long‐term (15 year) eddy covariance flux record and model estimates of tree and grass GPP inferred from satellite remote sensing. On a seasonal basis, the primary drivers of tree and grass GPP were solar radiation in the wet season and soil moisture in the dry season. On an interannual basis, soil water availability had a positive effect on tree GPP and a negative effect on grass GPP . No linear trend in the tree–grass GPP ratio was observed over the 15‐year study period. However, the tree–grass GPP ratio was correlated with the modes of climate variability, namely the Southern Oscillation Index. This study has provided insight into the long‐term contributions of trees and grasses to savanna productivity, along with their respective meteorological determinants of IAV .