Large-scale ecological restorations are recognized worldwide as an effective strategy to combat environmental degradation and promote sustainability. Remote sensing (RS) imagery, such as obtained from Landsat and Sentinel-2 satellites, can provide spatial, spectral, and temporal information on ecosystem service supply to support monitoring and evaluation of restoration interventions. However, because of the abundance of satellite data and methodological analysis options, choices in data selection and processing options need to be made. This study explored the effect of RS choices on the evaluation of changes in ecosystem services as a result of ecological restoration interventions. Using the ecosystem service of forage provision for wildlife as an example, we used a before-after-control-impact (BACI) analysis to compare how the following choices affected restoration evaluation outcomes: a) different number of control pixels; b) different spatial distribution of control pixels; c) intra-annual image selection; and d) different reference periods. In addition, e) we evaluated the effect of using two different satellite sensor types, using the ecosystem service 'erosion prevention' as an example. We explored the effect of these five choices for restoration sites in the Baviaanskloof, South Africa. Results showed that the choice of intraannual image selection, and the reference period describing the 'before state' had a strong effect on the outcomes, often leading to opposite BACI evaluation results. BACI results were less sensitive to choices related to the number of control points in the evaluation. The impact of methodological choices on the BACI outcomes was greater for the less degraded areas of our study site. Satellite sensor choice resulted in similar temporal trajectories of estimated supply. We demonstrated that RS choices have a strong effect on the evaluation results of restoration interventions. Therefore, we recommend that documenting the key RS choices results is essential when communicating restoration evaluation results in order to properly understand, manage and adapt restoration initiatives.

Satellite rainfall estimates (SRE) are an alternative rainfall data source to ground based gauged rainfall but require bias correction to better match gauged rainfall. This study proposes an innovative SRE bias correction procedure to serve now-casting applications of SRE in crop growth simulation. In the approach, effects of SRE biases on soil water storage expressed by the crop water requirement satisfaction index (WRSI) serve to identify an effective bias correction window size. WRSI was estimated for gauged rainfall, uncorrected and bias corrected SRE using the crop water balance model in Instat+. Gauged rainfall data from 20 stations (2012-2018) within the Lake Victoria basin of Kenya served to bias correct CHIRPS SRE. The proposed bias correction approach resulted in more favorable WRSI values. To determine the effective correction window length, a threshold shortage in soil water storage of 23.5% that can be attributed to the SRE bias was found effective. Bias correction improved correlation coefficients between gauged rainfall and SRE by 0.3 compared to uncorrected SRE. The root mean square error reduced by 3.5% for WRSI when corrected SRE served as input. The approach presented in this paper and the findings show the potential of bias-corrected SRE for crop growth now-casting.

Crop yield estimation provides information about crop production and contributes to food security monitoring systems. In the context of estimating yield in food insecure regions like sub-Sharan Africa, the large input requirements of crop growth models, fragmented landscape and small field sizes are substantial challenges. Multisensor data fusion can be a valuable source of high spatial and temporal resolution data to meet some of the requirements of crop growth models. In this study we estimated crop yield in fragmented agricultural landscapes of Ethiopia by assimilating Landsat and MODIS fused data in a simple algorithm for crop yield estimation (SAFY) model. An adapted version of the Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM) was used for data fusion. We estimated LAI using empirical relations between field-measured leaf area index (LAI) and enhanced vegetation index (EVI) from fused Landsat/MODIS and MODIS. LAI and phenology information determined from the data fusion, from MODIS and from field data were assimilated for comparison. The model was validated using in situ LAI, biomass and yield measured in rice and maize fields during the 2019 growing season. Data fusion minimized the yield estimation error (rRMSE= 16% for maize and rRMSE= 33% in rice) more than MODIS (rRMSE= 15% for maize and rRMSE= 37% in rice) because of its improved LAI and phenology estimation accuracy. Data fusion improved the calibration accuracy of the field and crop-specific model parameters and better captured the spatial variability of yield, which is vital for monitoring crop production and food security. Considering the promising results, further investigation into the transferability of the model in space and time to other fragmented agricultural landscapes is needed for large-area applications.

Micronutrient malnutrition is a global challenge affecting >2 billion people, in particular those with a crop-based diet and limited access to nutrient-rich food sources. Conventional methods for measuring the crop nutrients such as wet chemical analysis of grains are time-consuming and cost-prohibitive and, consequently, unsuitable for the consistent quantification of nutrients across space and time. In this study, we propose a new method that is using PRecursore IperSpettrale della Missione Applicativa (PRISMA) and Sentinel-2 images to estimate the nutrient concentrations of crop grains before harvest. We collected grain samples for corn, rice, soybean, and wheat from a farm situated in Italy and measured their nutrient concentrations in the lab. These measurements together with the PRISMA and Sentinel-2 images acquired at the main phases of crop development (vegetative, reproductive, maturity) were used as input for two-band vegetation indices (TBVIs) and Partial Least Squares Regression (PLSR) to predict Calcium (Ca), Iron (Fe), Potassium (K), Magnesium (Mg), Nitrogen (N), Phosphorus (P), Sulphur (S) and Zinc (Zn). Models' performances were assessed using the coefficient of determination (R2) and Root Mean Square Error (RMSE). For PRISMA images, the best prediction results were obtained for P in soybean (R2 = 0.69), K in soybean (R2 = 0.66), Mg in soybean (R2 = 0.58), Fe in soybean (R2 = 0.57), K in wheat (R2 = 0.57), K in corn (R2 = 0.55), P in wheat (R2 = 0.51), S in rice (R2 = 0.58) using TBVIs. In contrast to PRISMA, PLSR outperformed TBVIs when Sentinel-2 images were used as input. For Sentinel-2, the best predictions were obtained for P in soybean (R2 = 0.73), K in wheat (R2 = 0.67), Mg in soybean (R2 = 0.62), Zn in wheat (R2 = 0.56), Fe in soybean (R2 = 0.52), P in wheat (R2 = 0.52). Our study showed that estimating the nutrient composition of crops using remote sensing images has the potential to change how we approach a cost-effective, timely, and spatially explicit representation of the crops' nutritional quality.

Emphasizing connectedness among the many elements of plant-based systems that are collectively addressed in One Health is a logical outcome of debates generated by the International Year of Plant Health. The notion of One Health was implemented through the joint use of the concepts of production situation (PS)-the "rice way of life"-and injury profile (IP)-rice health (pathogens, animal pests, and weeds)-in a population of 1051 rice farmers' fields surveyed from 1987 to 2011. Seven associated PSs and nine IPs are characterized. An ending Green Revolution (Period 1) and a post-Green Revolution (Period 2) are considered. Major changes in PSs and IPs occurred, with increased mineral fertilizer inputs, pesticide applications, and landscape uniformity, along with shortening fallow periods, and dropping labour. Increases or decreases in some injuries lead to shifts in IPs towards more frequent bacterial, flower, and panicle diseases. Yield gain between the two periods was marginal but yield losses to IPs remained unchanged. Yet there was an increase by 77.0%, 21.6%, and 439% for insecticides, herbicides, and fungicides, respectively, along with a 27.9% increase in mineral fertilizer inputs. Total factor productivity in rice production dropped by over 40% during the 24 intervening years, not accounting for the long-term effects of these changes on soil fertility, biodiversity, and scarcer water resource. These results call for urgently required reassessment of public efforts to safeguard these agrosystems that are of critical importance economically, socially, and culturally for roughly half the world's population.

Satellite image data deliver consistent and frequent information for crop yield estimation over large areas. Hyperspectral narrowbands are more sensitive spectrally to changes in crop growth than multispectral broad bands but few studies quantified the gains in the former over the later. The PRecursore IperSpettrale della Missione Applicativa (PRISMA) mission offers narrow (<= 10 nm) band capability across the full optical range. The multispectral broadband Sentinel-2 mission carries four experimental red-edge and near infrared (NIR) hyper spectral narrow (<= 20 nm) bands. We compared the performance of PRISMA and Sentinel-2 spectral bands at important phases of crop development (vegetative, reproductive, maturity) in estimating field-level biomass and yield for corn, rice, soybean, and wheat. We selected three data-driven methods: two-band vegetation indices (TBVIs), partial least squares regression (PLSR), and random forest (RF). The PRISMA and Sentinel-2 models on average explained approximately 20% more variability in biomass and yield with RF than TBVIs and PLSR. The mean RMSE of the PRISMA RF models was 0.42 and 0.17 kg m(-2), which was lower than the Sentinel-2 RF models (0.48 and 0.18 kg m(-2)). Multidate image (seasonal) model performance was generally higher than single-date image model performance. PRISMA shortwave infrared narrowbands and Sentinel-2 red-edge and near infrared bands were among the top-performing spectral regions. The results highlight potential complementarity between the PRISMA and Sentinel-2 missions for predicting crop biomass and yield. The results also show the benefits, limitations, and pitfalls of hyperspectral imaging in agricultural monitoring, which is important for upcoming operational hyperspectral missions, such as ESA CHIME and NASA Surface Biology and Geology (formerly HysPIRI).

Reversing ecological degradation through restoration activities is a key societal challenge of the upcoming decade. However, lack of evidence on the effectiveness of restoration interventions leads to inconsistent, delayed, or poorly informed statements of success, hindering the wise allocation of resources, representing a missed opportunity to learn from previous experiences. This study contributes to a better understanding of spatial and temporal dynamics of ecosystem services at ecological restoration sites. We developed a method using Landsat satellite images combined with a Before-After-Control-Impact (BACI) design, and applied this to an arid rural landscape, the Baviaanskloof in South Africa. Since 1990, various restoration projects have been implemented to halt and reverse degradation. We applied the BACI approach at pixel-level comparing the conditions of each intervened pixel (impact) with 20 similar control pixels. By evaluating the conditions before and after the restoration intervention, we assessed the effectiveness of long-term restoration interventions distinguishing their impact from environmental temporal changes. The BACI approach was implemented with Landsat images that cover a 30-year period at a spatial resolution of 30 meter. We evaluated the impact of three interventions (revegetation, livestock exclusion, and the combination of both) on three ecosystem services; forage provision, erosion prevention, and presence of iconic vegetation. We also evaluated whether terrain characteristics could partially explain the variation in impact of interventions. The resulting maps showed spatial patterns of positive and negative effects of interventions on ecosystem services. Intervention effectiveness differed across vegetation conditions, terrain aspect, and soil parent material. Our method allows for spatially explicit quantification of the long-term restoration impact on ecosystem service supply, and for the detailed visualization of impact across an area. This pixel-level analysis is specifically suited for heterogeneous landscapes, where restoration impact not only varies between but also within restoration sites.

Evapotranspiration (ET) accounts for water movements from land to air and plays a vital role in the terrestrial water, energy, and carbon cycles. Reliable estimates of ET for agricultural landscapes can facilitate water resources management and food security analysis. The widely used Priestley-Taylor Jet Propulsion Laboratory (PT-JPL) model has the most potential to operationally simulate ET over large areas, but its inability to fully track soil evaporation dynamics using atmospheric humidity limits its application in agroecosystems. In this study, we isolated the uncertainties resulting from soil evaporation and assessed three Earth observation-based alternatives - apparent thermal inertia (ATI), microwave soil moisture (SM), and optical spectral indices based on shortwave infrared (SWIR) to formulate soil evaporation. Our results illustrate that the incorporation of the SWIR-based soil moisture divergence index (SMDI) and microwave-based SM into monthly soil evaporation led to 6% and 5% increase in explained ET variances and reduced RMSE by 23.2% and 13.1% for cropland and grassland, respectively, as compared to PT-JPL using atmospheric reanalysis data only. Further analyses demonstrated that PT-SMDI explained more observed ET variances than PT-JPL using in-situ measurements of atmospheric humidity during the crop growing season, particularly for irrigated cropland (R2=0.65 for PT-SMDI; R2=0.62 for PT-JPL). On the other hand, the use of microwave SM outperformed other indices for ET assessment in grasslands but had lower performance in croplands. Our results suggest that a combination of optical SWIR and microwave SM has good potential to improve the PT-JPL model accuracy for agricultural landscapes.

Remote sensing data are used to map the extent of croplands. They are especially useful in sub-Saharan Africa (SSA) where landscapes are complex and farms are small, i.e. less than two ha. In this study, a hierarchical remote sensing approach was developed to estimate field fractions at 30 m spatial resolution in a highly fragmented agricultural region of Ethiopia. The landscape was stratified into crop production system (CPS) zones with ten-day SPOT Proba-V 1 km normalized difference vegetation index (NDVI) composites. The CPS zones were used to disaggregate agricultural census statistics to 1 km field fractions and mask "wet" and "dry" seasons. Long-term average wet-dry season NDVI and topographic information derived from 30 m Landsat-8 (OLI) surface reflectance and the SRTM digital elevation model were combined with 1 km field fractions in a Generalized Additive Model (GAM) to produce the field fractions. Sample dot grids were manually interpreted from very high-resolution DigitalGlobe imagery on the Google Earth platform for training and testing. The model yielded an Area Under the Curve (AUC) of 0.71 and R-2 of 0.65 in the holdout sample set. The high AUC reveals the model was effective at classifying 30 m pixels as "crop" or "not crop" while the high R-2 indicated leveraging at the extremes (100 and 0% probability), meaning at 30 m resolution, subpixel variations were difficult to discern. The improved model skill compared to previous cropland mapping studies using GAMs can be attributed to the stratification and decomposition of the Landsat time series using CPS-defined phenology. Additional remote sensing model inputs, such as Sentinel-1 radar backscatter and Sentinel-2 red-edge reflectance, could provide additional explanatory power. Wall-to-wall national coverage for agricultural production estimation or other food security related application could be achieved by manually digitizing additional sample data in other regions of Ethiopia or using existing crowd-sourced databases, such as Geo-Wiki.

Planting a cover crop between the main cropping seasons is an agricultural management measure with multiple potential benefits for sustainable food production. In the maize production system of the Netherlands, an effective establishment of a winter cover crop is important for reducing nitrogen leaching to groundwater. Cover crop establishment after maize cultivation is obliged by law for sandy soils and consequently implemented on nearly all maize fields, but the winter-time vegetative ground cover varies significantly between fields. The objectives of this study are to assess the variability in winter vegetative cover and evaluate to what extent this variability can be explained by the timing of cover crop establishment and weather conditions in two growing seasons (2017–2018). We used Sentinel-2 satellite imagery to construct NDVI time series for fields known to be cultivated with maize within the province of Overijssel. We fitted piecewise logistic functions to the time series in order to estimate cover crop sowing date and retrieve the fitted NDVI value for 1 December (NDVIDec). We used NDVIDec to represent the quality of cover crop establishment at the start of the winter season. The Sentinel-2 estimated sowing dates compared reasonably with ground reference data for eight fields (RMSE = 6.6 days). The two analysed years differed considerably, with 2018 being much drier and warmer during summer. This drought resulted in an earlier estimated cover crop sowing date (on average 19 days) and an NDVIDec value that was 0.2 higher than in 2017. Combining both years and all fields, we found that Sentinel-2 retrieved sowing dates could explain 55% of the NDVIDec variability. This corresponded to a positive relationship (R2 = 0.50) between NDVIDec and the cumulative growing degree days (GDD) between sowing date and 1 December until reaching 400 GDD. Based on cumulative GDD derived from two weather stations within Overijssel, we found that on average for the past three decades a sowing date of 19 September (± 7 days) allowed to attain these 400 GDD; this provides support for the current legislation that states that from 2019 onwards a cover crop should be sown before 1 October. To meet this deadline, while simultaneously ascertaining a harvest-ready main crop, in practice implies that undersowing of the cover crop during spring will gain importance. Our results show that Sentinel-2 NDVI time series can assess the effectiveness and timing of cover crop growth for small agricultural fields, and as such has potential to inform regulatory frameworks as well as farmers with actionable information that may help to reduce nitrogen leaching.

Considerable efforts and resources are being invested in integrated conservation and restoration interventions in rural arid areas. Empirical research for quantifying ecosystem services - nature's benefits to people - is essential for evaluating the range of benefits of ecological restoration and to support its use in natural resource management. Satellite remote sensing (RS) can be used to monitor interventions, especially in large and remote areas. In this study we used field measurements, RS-based information from Sentinel-2 imagery together with soil and terrain data, to estimate ecosystem service supply and evaluate integrated ecological restoration interventions. We based our research on the arid, rural landscape of the Baviaanskloof Hartland Bawarea Conservancy, South Africa, where several integrated interventions have been implemented in areas where decades of small livestock farming has led to extensive land degradation. Interventions included i) long term livestock exclusion, ii) revegetating of degraded areas, iii) a combination of these two, and iv) essential oil production as alternatives to goat and sheep farming. We assessed six ecosystem services linked to the objectives of the interventions: erosion prevention, climate regulation, regulation of water flows, provision of forage, biomass for essential oil production, and the sense of place through presence of native species. We first estimated the ecosystem service supply based on field measurements. Secondly, we explored the relationships between ecosystem services quantities derived from the field measurements with 13 Sentinel-2 indices and four soil and terrain variables. We then selected the best fitting model for each ecosystem service. Finally, we compared the supply of ecosystem services between intervened and non-intervened sites. Results showed that models based on Sentinel-2 indices, combined with slope information, can estimate ecosystem services supply in the study area even when the levels of field-based ecosystem services supplies are low. The RS-based models can assess ecosystem services more accurately when their indicators mainly depend on green vegetation, such as for erosion prevention and provision of forage. The agricultural fields presented high variability between plots on the provision of ecosystem services. The use of Sentinel-2 vegetation indices and terrain data to quantify ecosystem services is a first step towards improving the monitoring and assessment of restoration interventions. Our results showed that in the study area, livestock exclusion lead to a consistent increase in most ecosystem services.

Landscape processes fluctuate over time, influencing the intra-annual dynamics of ecosystem services. However, current ecosystem service assessments generally do not account for such changes. This study argues that information on the dynamics of ecosystem services is essential for understanding and monitoring the impact of land management. We studied two regulating ecosystem services (i. erosion prevention, ii. regulation of water flows) and two provisioning services (iii. provision of forage, iv. biomass for essential oil production) in thicket vegetation and agricultural fields in the Baviaanskloof, South Africa. Using models based on Sentinel-2 data, calibrated with field measurements, we estimated the monthly supply of ecosystem services and assessed their intra-annual variability within vegetation cover types. We illustrated how the dynamic supply of ecosystem services related to temporal variations in their demand. We also found large spatial variability of the ecosystem service supply within a single vegetation cover type. In contrast to thicket vegetation, agricultural land showed larger temporal and spatial variability in the ecosystem service supply due to the effect of more intensive management. Knowledge of intra-annual dynamics is essential to jointly assess the temporal variation of supply and demand throughout the year to evaluate if the provision of ecosystem services occurs when most needed

Accurate and up-to-date spatial agricultural information is essential for applications including agro-environmental assessment, crop management, and appropriate targeting of agricultural technologies. There is growing research interest in spatial analysis of agricultural ecosystems applying satellite remote sensing technologies. However, usability of information generated from many of remotely sensed data is often constrained by accuracy problems. This is of particular concern in mapping complex agro-ecosystems in countries where small farm holdings are dominated by diverse crop types. This study is a contribution to the ongoing efforts towards overcoming accuracy challenges faced in remote sensing of agricultural ecosystems. We applied time-series analysis of vegetation indices (Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI)) derived from the Moderate Resolution Imaging Spectrometer (MODIS) sensor to detect seasonal patterns of irrigated and rainfed cropping patterns in five townships in the Central Dry Zone of Myanmar, which is an important agricultural region of the country has been poorly mapped with respect to cropping practices. To improve mapping accuracy and map legend completeness, we implemented a combination of (i) an iterative participatory approach to field data collection and classification, (ii) the identification of appropriate size and types of predictor variables (VIs), and (iii) evaluation of the suitability of three Machine Learning algorithms: Support Vector Machine (SVM), Random Forest (RF), and C5.0 algorithms under varying training sample sizes. Through these procedures, we were able to progressively improve accuracy and achieve maximum overall accuracy of 95% When a small sized training dataset was used, accuracy achieved by RF was significantly higher compared to SVM and C5.0 (P < 0.01), but as sample size increased, accuracy differences among the three machine learning algorithms diminished. Accuracy achieved by use of NDVI was consistently better than that of EVI (P < 0.01). The maximum overall accuracy was achieved using RF and 8-days NDVI composites for three years of remote sensing data. In conclusion, our findings highlight the important role of participatory classification, especially in areas where cropping systems are highly diverse and differ over space and time. We also show that the choice of classifiers and size of predictor variables are essential and complementary to the participatory mapping approach in achieving desired accuracy of cropping pattern mapping in areas where other sources of spatial information are scarce.

Mapping the extent and location of field boundaries is critical to food security analysis but remains problematic in the Global South where such information is needed the most. The difficulty is due primarily to fragmentation in the landscape, small farm sizes, and irregular farm boundaries. Very high-resolution satellite imagery affords an opportunity to delineate such fields, but the challenge remains of determining such boundaries in a systematic and accurate way. In this paper, we compare a new crowd-driven manual digitization tool (Crop Land Extent) with two semi-automated methods (contour detection and multi-resolution segmentation) to determine farm boundaries from WorldView imagery in highly fragmented agricultural landscapes of Ethiopia. More than 7000 one square-kilometer image tiles were used for the analysis. The three methods were assessed using quantitative completeness and spatial correctness. Contour detection tended to under-segment when compared to manual digitization, resulting in better performance for larger (approaching 1 ha) sized fields. Multi-resolution segmentation on the other hand, tended to over-segment, resulting in better performance for small fields. Neither semi-automated method in their current realizations however are suitable for field boundary mapping in highly fragmented landscapes. Crowd-driven manual digitization is promising, but requires more oversight, quality control, and training than the current workflow could allow.

Crop pathogens and pests reduce the yield and quality of agricultural production. They cause substantial economic losses and reduce food security at household, national and global levels. Quantitative, standardized information on crop losses is difficult to compile and compare across crops, agroecosystems and regions. Here, we report on an expert-based assessment of crop health, and provide numerical estimates of yield losses on an individual pathogen and pest basis for five major crops globally and in food security hotspots. Our results document losses associated with 137 pathogens and pests associated with wheat, rice, maize, potato and soybean worldwide. Our yield loss (range) estimates at a global level and per hotspot for wheat (21.5% (10.1-28.1%)), rice (30.0% (24.6-40.9%)), maize (22.5% (19.5-41.1%)), potato (17.2% (8.1-21.0%)) and soybean (21.4% (11.0-32.4%)) suggest that the highest losses are associated with food-deficit regions with fast-growing populations, and frequently with emerging or re-emerging pests and diseases. Our assessment highlights differences in impacts among crop pathogens and pests and among food security hotspots. This analysis contributes critical information to prioritize crop health management to improve the sustainability of agroecosystems in delivering services to societies.

Global food security is negatively affected by drought. Climate projections show that drought frequency and intensity may increase in different parts of the globe. Early season forecasts on drought occurrence and severity could help to better mitigate the negative consequences of drought. The objective of this study was to assess if interannual variability in agricultural productivity in Chile can be accurately predicted from freely-available, near real-time data sources. As the response variable, we used the standard score of seasonal cumulative NDVI (zcNDVI), based on 2000-2017 data from Moderate Resolution Imaging Spectroradiometer (MODIS), as a proxy for anomalies of seasonal primary productivity. The predictions were performed with forecast lead-times from one- to six-month before the end of the growing season, which varied between census units in Chile. Predictor variables included the zcNDVI obtained by cumulating NDVI from season start up to prediction time; standardised precipitation indices, derived from satellite rainfall estimates, for time-scales of 1, 3, 6, 12 and 24 months; the Pacific Decadal Oscillation and the Multivariate ENSO oscillation indices; the length of the growing season, and latitude and longitude. We used two prediction approaches: (i) optimal linear regression (OLR) whereby for each census unit the single predictor was selected that best explained the interannual zcNDVI variability, and (ii) a multi-layer feedforward neural network architecture, often called deep learning (DL), where all predictors for all units were combined in a single spatio-temporal model. Both approaches were evaluated with a leave-one-year-out cross-validation procedure. Both methods showed good prediction accuracies for small lead times and similar values for all lead times. The mean R2cv values for OLR were 0.95, 0.83, 0.68, 0.56, 0.46 and 0.37, against 0.96, 0.84, 0.65, 0.54, 0.46 and 0.38 for DL, for one, two, three, four, five, and six months lead time, respectively. Given the wide range of climates and vegetation types covered within the study area, we expect that the presented models can contribute to an improved early warning system for agricultural drought in different geographical settings around the globe.

Good access to resources and opportunities is essential for sustainable development. Improving access, especially in rural areas, requires useful measures of current access to the locations where these resources and opportunities are found. Recent work has developed a global map of travel times to cities with more than 50,000 people in the year 2015. However, the provision of resources and opportunities will differ across the broad spectrum of settlements that range from small towns to megacities, and access to this spectrum of settlement sizes should also be measured. Here we present a suite of nine global travel-time accessibility indicators for the year 2015, at approximately one-kilometre spatial resolution, for a range of settlement size classes. We validated the travel-time estimates against journey times from a Google driving directions application across 1,511 2° × 2° tiles representing 47,812 journeys. We observed very good agreement, though our estimates were more frequently shorter than those from the Google application with a median difference of −13.7 minutes and a median percentage difference of −16.9%.

Global food security is negatively affected by drought. Climate projections show that drought frequency and intensity may increase in different parts of the globe. These increases are particularly hazardous for developing countries. Early season forecasts on drought occurrence and severity could help to better mitigate the negative consequences of drought. The objective of this study was to assess if interannual variability in agricultural productivity in Chile can be accurately predicted from freely-available, near real-time data sources. As the response variable, we used the standard score of seasonal cumulative NDVI (zcNDVI), based on 2000-2017 data from Moderate Resolution Imaging Spectroradiometer (MODIS), as a proxy for anomalies of seasonal primary productivity. The predictions were performed with forecast lead times from one- to six-month before the end of the growing season, which varied between census units in Chile. Predictor variables included the zcNDVI obtained by cumulating NDVI from season start up to prediction time; standardised precipitation indices derived from satellite rainfall estimates, for time-scales of 1, 3, 6, 12 and 24 months; the Pacific Decadal Oscillation and the Multivariate ENSO oscillation indices; the length of the growing season, and latitude and longitude. For each of the 758 census units considered, the time series of the response and the predictor variables were averaged for agricultural areas resulting in a 17-season time series per unit for each variable. We used two prediction approaches: (i) optimal linear regression (OLR) whereby for each census unit the single predictor was selected that best explained the interannual zcNDVI variability, and (ii) a multi-layer feedforward neural network architecture, often called deep learning (DL), where all predictors for all units were combined in a single spatio-temporal model. Both approaches were evaluated with a leave-one-year-out cross-validation procedure. Both methods showed good prediction accuracies for small lead times and similar values for all lead times. The mean R-cv(2) values for OLR were 0.95, 0.83, 0.68, 0.56, 0.46 and 0.37, against 0.96, 0.84, 0.65, 0.54, 0.46 and 0.38 for DL, for one, two, three, four, five, and six months lead time, respectively. Given the wide range of climates and vegetation types covered within the study area, we expect that the presented models can contribute to an improved early warning system for agricultural drought in different geographical settings around the globe.