TYPE Review PUBLISHED 30 November 2022 DOI 10.3389/fclim.2022.938975 Can remote sensing enable a OPEN ACCESS Biomass Climate Adaptation EDITED BY Bao-Jie He, Index for agricultural systems? Chongqing University, China REVIEWED BY Dongxue Zhao, Amy Ferguson1, Catherine Murray1, Yared Mesfin Tessema1, The University of Queensland, Australia Hailong Wang, Peter C. McKeown1, Louis Reymondin2, Sun Yat-sen University, China Ana Maria Loboguerrero2, Tiany Talsma2, Brendan Allen1, *CORRESPONDENCE Charles Spillane Andy Jarvis2, Aaron Golden3 and Charles Spillane1* charles.spillane@nuigalway.ie 1Agriculture and Bioeconomy Research Centre, Ryan Institute, University of Galway, Galway, Ireland, SPECIALTY SECTION 2Climate Change Agriculture and Food Security Program (CCAFS), Alliance of Bioversity This article was submitted to International and CIAT, Cali, Colombia, 3Ryan Institute and School of Natural Sciences, University of Climate Adaptation, Galway, Galway, Ireland a section of the journal Frontiers in Climate RECEIVED 08 May 2022 Systematic tools and approaches for measuring climate change adaptation ACCEPTED 11 October 2022 at multiple scales of spatial resolution are lacking, limiting measurement of PUBLISHED 30 November 2022 progress toward the adaptation goals of the Paris Agreement. In particular, CITATION Ferguson A, Murray C, Mesfin there is a lack of adaptation measurement or tracking systems that are Tessema Y, McKeown PC, coherent (measuring adaptation itself), comparable (allowing comparisons Reymondin L, Loboguerrero AM, Talsma T, Allen B, Jarvis A, Golden A across geographies and systems), and comprehensive (are supported by the and Spillane C (2022) Can remote necessary data). In addition, most adaptation measurement eorts lack an sensing enable a Biomass Climate appropriate counterfactual baseline to assess the eectiveness of adaptation- Adaptation Index for agricultural systems? Front. Clim. 4:938975. related interventions. To address this, we are developing a “Biomass Climate doi: 10.3389/fclim.2022.938975 Adaptation Index” (Biomass CAI) for agricultural systems, where climate COPYRIGHT adaptation progress acrossmultiple scales can bemeasured by satellite remote © 2022 Ferguson, Murray, Mesfin sensing. The Biomass CAI can be used at global, national, landscape and farm- Tessema, McKeown, Reymondin, Loboguerrero, Talsma, Allen, Jarvis, level to remotelymonitor agri-biomass productivity associatedwith adaptation Golden and Spillane. This is an interventions, and to facilitate more tailored “precision adaptation”. The open-access article distributed under Biomass CAI places focus on decision-support for end-users to ensure that the terms of the Creative Commons Attribution License (CC BY). The use, the most eective climate change adaptation investments and interventions distribution or reproduction in other can be made in agricultural and food systems. forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the KEYWORDS original publication in this journal is cited, in accordance with accepted climate change, agriculture, resilience, adaptation, remote sensing, artificial academic practice. No use, distribution intelligence, machine learning or reproduction is permitted which does not comply with these terms. Introduction Unless greenhouse gas (GHG) emissions are curbed significantly within decadal timeframes to follow low emissions scenarios and allow us to remain within 1.5◦C by mid century, our social and ecological systems will experience more frequent and intense climate change impacts throughout the rest of this century and beyond (IPCC, 2018, 2021). Such climate change impacts will include temperature increases, sea level rise, changes to precipitation patterns, and an increased prevalence and intensity of extreme weather shocks (IPCC, 2018, 2021). The productivity of many agricultural systems is expected to be negatively impacted by climate change impacts (Mbow et al., 2019), potentially reducing yields of major staple crops by 3–12% by 2050, and by 11–25% by 2100 (Wing et al., 2021). Without effective adaptation measures at farm and Frontiers inClimate 01 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 landscape levels, climate change will negatively impact et al., 2012; Park et al., 2012; Rickards and Howden, 2012; livelihoods and food security dependant on farming (Morton, Rippke et al., 2016; Vermeulen et al., 2018). 2007; Wheeler and Von Braun, 2013; Lipper et al., 2014; Mbow et al., 2019; Wing et al., 2021). To strengthen the resilience of rural and agricultural communities at risk of climate impacts, Resilience, Transformation and a wide ranging portfolio of “Climate-Smart Agriculture” (CSA) Climate Change practices are being deployed and scaled (Lipper et al., 2014; Rosenstock et al., 2019; FAO, 2013). Adaptation progress is The resilience of social-ecological systems has been critical to meeting the Paris Agreement goals, with financing extensively investigated, where early definitions refer to entities pledging investments (The World Bank Group, 2021) resilience as “the ability of a system to absorb changes of and over 131 parties now prioritizing adaptation of agricultural state variables, driving variables, and parameters, and still systems in their nationally determined contributions (NDCs) persist” (Holling, 1973) . This conceptualization of resilience (Strohmaier et al., 2016). recognizes that resilient systems can be unstable, shifting to a new state in different “basins of attraction” to promote resilience (Holling, 1973; Allen et al., 2019). The idea that systems can Building resilience through climate dramatically change state in a manner that leads to improved change adaptation resilience is referred to as “transformation” (Walker et al., 2004). Frameworks have been devised to help understand Conceptualization of climate change adaptation has resilience, however many complicate the definition further by juxtaposed incremental vs. transformational adaptation in introducing conflicting ideas (Allen et al., 2019). For example, agricultural systems (Howden et al., 2010; Kates et al., 2012; some approaches focus on the stability of systems in terms Park et al., 2012; Rickards and Howden, 2012; Rippke et al., of their “robustness, resistance and recovery”, insinuating 2016; Vermeulen et al., 2018). Incremental adaptation refers to that systems must remain in or return to their original state small changes made to an existing farming system to mitigate (Allen et al., 2019; Grafton et al., 2019). However, conflicting the impacts of climate shocks, such as changing planting times arguments should be considered together to move forward or varieties in accordance with weather projections (Howden toward improved decision-making for resilience building (Allen et al., 2010; Kates et al., 2012; Rickards and Howden, 2012). et al., 2019). This is critically important when considering Fitting with the concept that systems can migrate to different the threat of climate change, which could involuntarily push states with differing levels of resilience (Holling, 1973; Allen systems beyond an equilibrium threshold into a state that is et al., 2019), agricultural systems can undergo transformational maladapted to climate change impacts (Folke et al., 2010). For adaptation by fundamentally changing their system state when instance, some crops in regions of Sub-Saharan Africa could see climate change threatens the system’s existence (Howden et al., thresholds crossed before 2100, beyond which their cultivation 2010; Kates et al., 2012; Rickards and Howden, 2012). would not be feasible in the regions affected (Rippke et al., It is considered that transformational adaptation may be 2016). There remains an ongoing need for transformation of more appropriate than incremental adaptation for agricultural small scale systems (e.g., agricultural systems), to facilitate wider systems that face intense climate change projections (Howden scale Earth resilience (Folke et al., 2010). et al., 2010; Kates et al., 2012; Park et al., 2012; Rickards and Howden, 2012; Rippke et al., 2016; Vermeulen et al., 2018), with investment in incremental adaptation being criticized for Tracking climate change adaptation delaying the implementation of transformational adaptations in agriculture (Rickards and Howden, 2012). However, transformational adaptation efforts have significant inertia and path dependency Although they are frequently juxtaposed, incremental and effects to overcome, are often disorderly in practice (Vermeulen transformational adaptation do not necessarily need to be et al., 2018) and involve significant risk and barriers (Howden viewed as entirely separate pathways (Kates et al., 2012; Park et al., 2010; Kates et al., 2012; Park et al., 2012; Rickards and et al., 2012; Rickards andHowden, 2012; Vermeulen et al., 2018). Howden, 2012; Vermeulen et al., 2018). Both incremental Incremental and transformational adaptation can be considered and transformational adaptation processes require more as components of a broader “Adaptation Action Cycle” (Park robust planning processes with the involvement of multiple et al., 2012), where they occur in connected cycles across stakeholders (Howden et al., 2010; Kates et al., 2012; Park et al., four key stages: identification of the problem and development 2012; Vermeulen et al., 2018), financial support (Kates et al., of goals, creation of an adaptation plan, implementation of 2012; Rickards and Howden, 2012; Vermeulen et al., 2018), and the adaptation, and monitoring and evaluation (Park et al., better monitoring and decision-support tools to identify gaps 2012). After monitoring and evaluation (indicated by the arrows and develop an “evidence base” for decision-making (Kates in Figure 1), a system has the opportunity to switch from Frontiers inClimate 02 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 Adaptation tracking systems should be “consistent, coherent, comparable and comprehensive”—referred to as the “4Cs” (Ford and Berrang-Ford, 2016). Although the “4Cs” framework is largely applied to national level reporting for adaptation in all sectors, the “4Cs” are also important to consider for any tracking methodology. Firstly, a tracking system must have a “consistent” definition of adaptation (Ford and Berrang-Ford, 2016). The UNFCCC and the IPCC define adaptation as an “adjustment in natural or human systems in response to actual or expected climatic stimuli or their effects, which moderates harm or exploits beneficial opportunities” (Agard et al., 2014; UNFCCC, 2021). In the context of the agriculture sector, the broad concept of FIGURE 1 adaptation is now well understood. For example, adaptation Schematic of transformational and incremental adaptation in interventions under the banner of “Climate-Smart Agriculture” relation to the benefit gained from the adaptation and the risks involved. Redrawn from Howden et al. (2010) and Rickards and are considered as particular actions that increase productivity Howden (2012), taking the “Adaptation Action Cycle” (Park et al., and resilience to climate shocks (FAO, 2013; Lipper et al., 2012) into consideration. The black arrows represent 2014; Rosenstock et al., 2019). “Coherency” refers to the ability information provided by monitoring and evaluation tracking systems, allowing a transition between adaptation states (Park of an adaptation tracking system to appropriately measure a et al., 2012). successful adaptation, rather than for example, the quantity of adaptation interventions implemented (Ford and Berrang- Ford, 2016). A tracking system must also be “comparable” to enable assessment between different areas and across incremental to transformational adaptation, or vice versa (Park different time periods, involving metrics that are transparent et al., 2012). In this context, “adaptation tracking systems” are and easily collected throughout time to analyse adaptation critically important to monitor trends in adaptation progress progress (Ford and Berrang-Ford, 2016). “Comprehensive” through time and space, which is fundamental for (a) assessing tracking systems composed of good quality and abundant the performance of adaptation interventions and evaluating how data can facilitate comparability (Ford and Berrang-Ford, successful different adaptation investments are, (b) recognizing 2016). priorities, and (c) directing attention and investments toward Developing a comparable and comprehensive adaptation these priority areas (Ford et al., 2013; Ford and Berrang-Ford, tracking system faces difficulties due to methodological and 2016). If evidence shows that a transformation is required, empirical challenges associated with data collection (Ford et al., a system can enter a “preparatory phase” of decision-making 2015; Ford and Berrang-Ford, 2016; Adaptation Committee, (Rippke et al., 2016). 2021). For example, there are a lack of accepted indicators that Adaptation tracking systems have to date gained limited can be applied universally, compared to mitigation which can attention within the UNFCCC process (Ford et al., 2015). be measured universally using greenhouse gas concentrations Indeed, there are no adaptation tracking systems that are (Brooks et al., 2011; Ford et al., 2015; Ramasamy, 2017; Jacobs currently universally accepted or deployed by the global and Al-Azar, 2019; Adaptation Committee, 2021). It is also time adaptation community (Adaptation Committee, 2021). An consuming and expensive to collect data on multi- and wide- urgent need for the development of systematic tracking scales (FAO, 2013; Ramasamy, 2017; FAO and UNDP, 2019; procedures and methodologies that can assess global progress Jacobs and Al-Azar, 2019), i.e., data collected using surveys toward adaptation goals is recognized (Ford et al., 2013, which can take months to gather and process, and can cause 2015; Ford and Berrang-Ford, 2016; Berrang-Ford et al., a delay in obtaining findings that may be required quickly for 2019; Adaptation Committee, 2021). Investigations into further adaptation planning (FAO and UNDP, 2019). A further methodologies to track adaptation are increasing, driven partly challenge noted by many in the adaptation tracking sphere is by the climate financing community calling for evidence behind the difficulty in devising a “counterfactual baseline” to compare the outcomes of adaptation investments (Ford et al., 2013; adaptation progress to, especially as baselines shift with climate Ford and Berrang-Ford, 2016; Jacobs and Al-Azar, 2019), change (Brooks et al., 2011; FAO, 2013; Ford et al., 2013, 2015; and the demand for national reporting and transparency to Dinshaw et al., 2014; Ford and Berrang-Ford, 2016; Ramasamy, meet the Paris Agreement’s global goal on adaptation (Ford 2017; FAO and UNDP, 2019; Jacobs and Al-Azar, 2019). A et al., 2015; UNFCCC, 2015; Ramasamy, 2017; Rosenstock tracking system that is “coherent, standardized and relevant” is et al., 2017; Berrang-Ford et al., 2019; Jacobs and Al-Azar, required, while also being inexpensive and accurate (Rosenstock 2019). et al., 2017). Frontiers inClimate 03 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 Remote sensing technology for 2011; Damian et al., 2020; Lebrini et al., 2020). NDVI has further climate change adaptation tracking been used to estimate net primary productivity of vegetation (Vrieling et al., 2011; Anchang et al., 2019), and to track climate To develop next-generation adaptation tracking systems, change impacts on vegetation (Liu et al., 2015; Piedallu et al., there is an opportunity to establish new methods of data 2019). collection building on technological innovation (Ford et al., Remote sensing is also being investigated as a method to 2016; Rosenstock et al., 2017; FAO and UNDP, 2019). monitor the resilience of ecosystems (Ndungu et al., 2019; Jones In particular, remote sensing technology presents a major et al., 2021). For example, NDVI has been used to measure opportunity to improve climate change adaptation tracking the resilience of ecosystems following exposure to shocks such systems (Ford et al., 2016; Rosenstock et al., 2017; FAO and as a drought (Washington-Allen et al., 2008; Ndungu et al., UNDP, 2019; Schiavon et al., 2021), with the potential to track 2019; Von Keyserlingk et al., 2021). However, limited research a range of indicators at high spatial and temporal resolution has to date been conducted to assess the use of NDVI to track (Rosenstock et al., 2017). Indeed, the IPCC have acknowledged climate change adaptations in the agriculture sector. Recently, the key role of remote sensing formonitoring land based systems the potential of MODIS satellite derived NDVI as a tool for (IPCC, 2020). Satellite remote sensing (SRS) demonstrates the monitoring adaptations has been demonstrated focusing on ability to consistently monitor agriculture systems across all sites in Burkina Faso (Nyamekye et al., 2021) and Kenya geospatial locations globally, providing data that can generate (Ndungu et al., 2019). Using the RESTREND method (Evans more impartial and reliable evidence to direct decision-making and Geerken, 2004; Ibrahim et al., 2015), it is possible to (Atzberger, 2013; Yang et al., 2013). Remote sensing tools monitor adaptation interventions compared to a counterfactual are consequently becoming more commonplace within the (Nyamekye et al., 2021). For this method, NDVI is predicted agriculture sector, such as the Copernicus programme in the using linear regression modeling with precipitation or soil European Union which is being used for tracking progress and moisture data, which is then subtracted from observed NDVI to issuing payments to farmers under the Common Agricultural show land degradation independent of climatic influence (Evans Policy (European Commision., 2018; Schiavon et al., 2021). and Geerken, 2004; Ibrahim et al., 2015). In the context of biomass (natural or agricultural), vegetation Climatic factors, evapotranspiration, water quality, and indices are commonly used metrics derived from SRS, such as topography all have an impact on crop growth (Jia et al., the Normalized Difference Vegetation Index. 2020). NDVI can be used to examine crop growth and its relationship with various factors to reveal the important factors The Normalized Difference Vegetation Index (NDVI) is a for intervention and tracking climate adaptation (Phan et al., popular vegetation index used to monitor the productivity 2021; Shen and Evans, 2021; Yadav and Geli, 2021; Rigden et al., (Pettorelli et al., 2005) and therefore the “greenness” 2022). For instance, in China, precipitation was found to be of vegetation (Reed et al., 1994). As chlorophyll within the leading cause of agricultural failure over other factors (Peng photosynthetic organisms absorbs red light (between 0.6 and et al., 2008), whereas Lamchin et al. (2018) found temperature 0.7µm) rather than near infrared light (between 0.75 and to be the most influential factor in vegetation growth in the 1.35µm), a “contrast” is created (Myneni et al., 1995; Pettorelli Asia region. Agricultural yields, such as of tea in Vietnam et al., 2005; Dinan et al., 2015). Using SRS, the reflectance of red (Phan et al., 2021), and corn, sorghum, alfalfa, and wheat in and near infrared light and the contrast between these can be New Mexico, USA (Yadav and Geli, 2021) have been predicted calculated using the NDVI equation (Equation 1), with values by calculating the deviation of historical mean NDVI to the falling between −1 and 1 (Huete et al., 1994; Reed et al., 1994; current and assessing the correlation with water stress, extreme Myneni et al., 1995; Pettorelli et al., 2005; Dinan et al., 2015). weather events, and soil moisture. From time series NDVI Equation 1: (Rigden et al., 2022) and (Wei et al., 2015) have depicted the NIR− Red NDVI = (1) cropping calendar and explored climatic variability’s impact on NIR+ Red crop yield (Rigden et al., 2022). Wei et al. (2015) define onset A range of metrics, such as maximum NDVI and cumulative and end of growth as the dates when the reconstructed NDVI NDVI, can be derived from an NDVI time-series spanning a time-series curve increases and decreases to 20% of the overall vegetation growing season, indicating temporal phenological level, respectively, and the peak of growth is defined as the changes (Pettorelli et al., 2005). NDVI provides an indication of date when the reconstructed NDVI time-series curve reaches photosynthetic activity (Tucker, 1979; Asrar et al., 1984; Huete the maximum. This capability endorses NDVI for developing et al., 1994; Reed et al., 1994; Myneni et al., 1995; Pettorelli et al., biomass-based climate adaptation index and provides quasi- 2005; Dinan et al., 2015), and has therefore been extensively real-time information for different farming systems. Such applied to the agricultural sector to monitor crop productivity research serves as a basis for our current “Tracking Adaptation and yield (Moriondo et al., 2007; Huang et al., 2013; Lopresti Progress in Agricultural Systems” (TAPAS) program, which et al., 2015), in addition to identification and monitoring of crop is investigating the use of SRS and deep learning derived systems and management zones on a wide scale (Vrieling et al., NDVI to track agri-biomass resilience fostered by climate Frontiers inClimate 04 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 TABLE 1 Selection of adaptation tracking indicators from the TAAS deep learning (DL) models, convolutional neural networks framework (Ramasamy, 2017). (CNNs) and long-short-term memory (LSTM), and one more Category Indicator Method of Outcome/ traditional machine learning model (i.e., random forest -RF) for collection process spatiotemporal data fusion of Landsat 8 and Sentinel-2 NDVI datasets for predicting vegetation growth in China. Htitiou Natural Percentage of the Quantitative Outcome et al. (2021) have also used deep learning-based Very Deep Resources and population employed in (Gender Super-Resolution (VDSR) for spatiotemporal data fusion of Ecosystems agriculture that own land disaggregated) NDVI retrieved from Sentinel-2 and Landsat 8 images for crop Procedures in place to Qualitative Process monitoring. Their study indicated that VDSR performed better ensure species diversity than the enhanced spatial and temporal adaptive reflectance conservation fusion model (ESTARFM) and the flexible spatiotemporal data Agricultural Percentage change in Quantitative Outcome fusion (FSDAF) spatiotemporal image fusion algorithms in Production yield from the baseline terms of producing the least blurred images and predictions Systems of NDVI values. To produce a high spatial resolution NDVI Percentage GDP loss Quantitative Outcome dataset for investigating vegetation dynamics in heterogeneous associated with crop loss landscapes, Liao et al. (2016) proposed the NDVI-Bayesian Socio- Percentage of the Quantitative Outcome Spatiotemporal Fusion Model (NDVI-BSFM), which integrates economics population that are the Moderate Resolution Imaging Spectroradiometer (MODIS) undernourished and Landsat 8 NDVI. Percentage of the Quantitative Outcome Various machine learning (ML) algorithms are presented population under safety (Gender for predicting vegetation conditions using NDVI data. For nets disaggregated) instance, Huang et al. (2017) proposed the use of Multiple Institutions Operational capacity of Qualitative Process Linear Regression (MLR), Artificial Neural Network (ANN), and Policies climate adaptation funds and Support Vector Machine (SVM) models to improve Number of times climate Quantitative Process NDVI prediction. Meanwhile, for predicting NDVI from non- scenarios have been used stationary big remote sensing time series long short-term in adaptation planning memory (LSTM) neural networks have been proposed by Reddy and Prasad (2018) and Rhif et al. (2020) and conventional LSTM (ConvLSTM) for crop forecasting by Ahmad et al. (2020b). change adaptation interventions. Although a multitude of The Elman recurrent neural network model (ERNN) has interdisciplinary indicators are essential for tracking adaptation been used for short-term NDVI index forecasting (Stepchenko as highlighted in Table 1, an SRS basedNDVImonitoring system and Chizhov, 2015). Machine learning model-based extreme of agri-biomass provides a straightforward and ex-situ method gradient boosting method has been used to predict vegetation of tracking adaptation interventions in agricultural systems to growth represented by NDVI throughout the growing season aid decision-making (Ndungu et al., 2019; Nyamekye et al., from 2001 to 2018 in China (Li et al., 2021). By assessing NDVI, 2021). leaf area index (LAI) and normalized difference water index (NDWI) derived from Landsat 8 surface reflectance, grape yield estimations were made using artificial neural network (ANN) Use of machine learning for based machine learning and regression analysis (Arab et al., 2021). spatiotemporal data fusion, Vegetation indices (NDVI and EVI) extracted from the NDVI-based vegetation prediction, 2001 to 2018 MODIS dataset have also been used to forecast and accuracy evaluation their values in 2019 using Vector Regression, Random Forest (RF), and Linear and Polynomial Regression (Roy, 2021). For The availability of simultaneous high spatiotemporal predicting maize yield from land surface temperature (LST) resolution remote sensing data is highly desirable for more and NDVI in Pakistan, Ahmad et al. (2020a) applied the effectively monitoring and predicting vegetation growth (Ning K-Nearest Neighbor clustering machine learning model. In et al., 2015; Bento et al., 2020; Maselli et al., 2020; Kloos predicting drought impacts on crop yield Mann et al. (2019) et al., 2021; Measho et al., 2021; Wang et al., 2021). It is now used a machine learning-based random forest model that takes becoming easier to improve the spatiotemporal resolution NDVI, precipitation, and evaporation as indicators. Despite of remote sensing data using machine learning to enhance the availability of numerous deep machine learning models, vegetation monitoring and prediction capacity (Ferchichi et al., their prediction accuracy may vary greatly when used in 2022). For instance, Mishra and Shahi (2021) have applied two biomass-based adaptation indexes. For instance, CNN and RF Frontiers inClimate 05 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 display good performance in vegetation growth predictions Adaptation Index (Biomass CAI) as a versatile tracking tool from NDVI (Ayhan et al., 2020; Li et al., 2021; Mishra based on SRS data to assist monitoring and decision-making and Shahi, 2021; Ferchichi et al., 2022). The performance of for agricultural resilience. For a given area of adaptation, machine learning models can be evaluated through a range the Biomass CAI metric uses inputs of both observed NDVI of approaches, including Root Mean Square Error (RMSE), directly from satellites, and predicted NDVI from deep coefficient of determinates (R2), Pearson correlation (R), and learning algorithms. As the deep learning algorithms predict structural similarity (SSIM), which have been used by Rhif NDVI accounting for shifts in climatic perturbations, this et al. (2020), Ahmad et al. (2020b), Arab et al. (2021), Htitiou represents the “counterfactual” situation that would occur if the et al. (2021), Mishra and Shahi (2021), and Roy (2021). Htitiou adaptation intervention was not implemented. By subtracting et al. (2021) use NDVI values extracted from spatial transects the deep learning predicted NDVI from observed NDVI, created across the study site to compare the performance of Very our Biomass CAI can calculate a “score” for any given Deep Super-Resolution (VDSR) against the enhanced spatial geospatial location, showing the deviation observed NDVI and temporal adaptive reflectance fusion model (ESTARFM) measurements make from the counterfactual baseline. The and the flexible spatiotemporal data fusion (FSDAF) method in Biomass CAI presents scores over time in a time-series, producing high resolution NDVI time series datasets. Based on giving the end-user a quantitative indication of the agri- the random point (RPO) sample construction method, Li et al. biomass productivity and therefore the comparative “success” (2022) have investigated the prediction capacity of four machine of the adaptation intervention over time. Figure 2A provides a learning approaches, (backpropagation neural network, decision schematic representation of a Biomass CAI time-series. Using tree, RF, and support vector machine), to predict the quality of SRS to collect NDVI data for the Biomass CAI complements cultivated land, where RF was found to be the most accurate. the comparability and comprehensiveness components of the The accuracy and performance of combined feature “4Cs” (Ford and Berrang-Ford, 2016) due to the high spatial and engineering forecasting model (SF-CNN) and CNN for temporal resolution of satellites such as MODIS (Dinan et al., forecasting NDVI have been assessed using the root mean 2015), Sentinel-2 (Drusch et al., 2012) and Landsat (Wulder square error (RMSE), mean absolute percentage error (MAPE), et al., 2019); and the capabilities of deep learning AI algorithms Nash-Sutcliffe coefficient (NSE), and mean absolute error to generate predicted data. (MAE) statistics indicators (Cui et al., 2020). Furthermore, Imageries for crop monitoring can currently be obtained to improve the performance of machine learning prediction from a variety of sources, including remote sensing (RS) by accuracy ensemble algorithms have been incorporated into a satellite, aerial, and unmanned aerial vehicles (UAVs) that collect variety of applications including crop yield prediction and data across a range of spatial, temporal, and spectral resolutions forest structure and biomass estimation (Zhang et al., 2022a). (Yao et al., 2017). Phan et al. (2021) use 1 km spatial resolution Ensemblemachine learning has been used to improve vegetation NDVI derived from MODIS for yield prediction and assessing and cropland classification accuracy (Aguilar et al., 2018; lag time between growing season and climatic variables for a Drobnjak et al., 2022). For above ground biomass estimation homogeneous farming system such as tea farming. Similar data a stacking ensemble algorithm has been used (Zhang et al., have also been used by Rigden et al. (2022) to explore the 2022b). Zhang et al. (2022a) have identified bagging, boosting, trend of drought impacts on rice, cassava, maize, and sweet and stacking as widely used ensemble techniques. The use potato yield in Madagascar. The impact of climate change on of ensemble machine learning algorithms in biomass-based farming systems in Sub-Saharan Africa has also been tracked adaptation index development will improve the performance using an 8 km resolution AVHRR-NDVI (Vrieling et al., 2011). of the index. Evaluation indices, such as Lin’s Concordance A sequence of Landsat 30-meter resolution NDVI has been used Correlation Coefficient (LCCC), can further improve prediction by Shen and Evans (2021) for estimating wheat yields in fields. accuracy. Zhao et al. (2022) uses ensemble modeling, averaging However, for detecting crop health patterns and making models using Granger–Ramanathan averaging (GRA) and appropriate interventions such as fertilizer or pesticides, LCCC, to improve prediction accuracy. They found that, multispectral and fine spatial resolution data is required. Yao though both methods improved prediction accuracy, GRA was et al. (2017), for example, were able to estimate wheat leaf area better performing. index (LAI) effectively with UAVs narrowband multispectral image (400–800 nm spectral regions, and 10 cm resolution) under varying growth conditions during five critical growth Options and challenges for stages, and provide potential technical support for nitrogen developing a Biomass Climate fertilization optimization. Satellite data with a wider spectral Adaptation Index (Biomass CAI) band and multispectral imagery can help differentiate crop characteristics (i.e., leaves, area) at a stand level (Gnädinger To meet the need for an integrative tracking system for and Schmidhalter, 2017; Jin et al., 2017; Varela et al., 2018), measuring agricultural adaptation across multiple scales, within estimate crop yield (Fernandez-Ordonez and Soria-Ruiz, 2017; the TAPAS program we are developing a Biomass Climate Yadav and Geli, 2021; Rigden et al., 2022); assess crop health Frontiers inClimate 06 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 FIGURE 2 (A) Schematic showing theoretical time-series trajectories of the Biomass CAI. Positive Biomass CAI values in the “Zone of Adaptation” reflect larger observed NDVI values than predicted NDVI values, meaning the adaptation has improved agri-biomass (crop) productivity. Observed NDVI values less than predicted NDVI values will give a negative Biomass CAI “score” in the “Zone of Degradation”, meaning the adaptation intervention has not improved agri-biomass productivity and a transformational shift could be considered; (B) End-users at Global/national scale (Conde and Lonsdale, 2005; FAO, 2013; Ford et al., 2013, 2015; Ford and Berrang-Ford, 2016; Berrang-Ford et al., 2019), regional/landscape scale (Conde and Lonsdale, 2005; Galarraga et al., 2011; FAO, 2013) and local cale (Conde and Lonsdale, 2005; FAO, 2013; Sherman and Ford, 2014) of the TAPAS Biomass CAI at corresponding spatial resolutions with appropriate satellites outlined. Sentinel-2 (Drusch et al., 2012) and Landsat (Wulder et al., 2019) appropriate satellites for local level analysis, while MODIS (Dinan et al., 2015) is an appropriate satellite for analysis at landscape, national and global resolutions. such as pest pressure patterns that cannot be detected by thermal “Zone of Degradation”, it can be inferred that the adaptation imagery (Khanal et al., 2017), examine soil moisture (Hassan- has not been successful and that transformation could be Esfahani et al., 2017), and crop water stress (Maselli et al., considered and planned as part of the “preparatory” phase 2020). To overcome the spatial and spectral variations between for transformation (Rippke et al., 2016) and “transformative remote sensing data for developing the CAI, it is important to governance” (Chaffin et al., 2016). Such “pro-active” adaptation recognize that satellite data are more likely to be influenced by planning in anticipation of climate change shocks is crucial several factors, including farming system, crop type, growing to avoid agricultural systems shifting involuntarily into a new state, management objectives, and data availability. maladapted state due to climate change shocks (Folke et al., To ensure relevance to end-users and multicriteria 2010; Kates et al., 2012; Park et al., 2012; Vermeulen et al., 2018). decision-making, our proposed Biomass CAI incorporates concepts of resilience. The application of a “resilience lens” Challenges and Opportunities for to monitoring tools can provide more resilience-oriented outputs and outcomes (Douxchamps et al., 2017). The evolutionary development of a Biomass CAI encompasses the idea that resilient systems Biomass Climate Adaptation Index are dynamic (Holling, 1973; Allen et al., 2019) and can shift between incremental and transformational adaptation Throughout the development of the Biomass CAI, following monitoring and evaluation in the cycle of adaptation identification of potential barriers and solutions is important, (Park et al., 2012). In this respect, if values remain in the as outlined in Table 2. “Attribution” is a barrier faced by most Frontiers inClimate 07 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 adaptation tracking systems, where the cause of success or interfaces (Jacobs and Al-Azar, 2019; Ndungu et al., 2019), failure is difficult to attribute directly to an adaptation initiative training manuals to guide use and interpretation (FAO and (Brooks et al., 2011; FAO, 2013; Ford et al., 2013; Dinshaw UNDP, 2019) and/or collaboration with the TAPAS program to et al., 2014; Ford and Berrang-Ford, 2016; Ramasamy, 2017; facilitate knowledge transfer, adoption and scaling (FAO, 2013; Berrang-Ford et al., 2019; FAO and UNDP, 2019; Jacobs Ndungu et al., 2019). The TAPAS Biomass CAI platform is and Al-Azar, 2019; Adaptation Committee, 2021). Despite being developed with a graphical user interface (GUI) viamobile this, adaptation tracking systems with robust and flexible phone and low-bandwidth internet access, essential to bridging counterfactual scenarios (Brooks et al., 2011; Dinshaw et al., the “digital divide” represented by lack of access to bandwidth 2014), combined with large spatial and temporal datasets (Ford (Hilbert, 2016). Many regions of Africa, South America and et al., 2013) can provide a strong evidence base to attribute Asia additionally have limited capacity to implement SRS based a particular adaptation initiative as the source of success or monitoring systems through direct use of satellite data (Romijn failure. The flexibility of counterfactual baselines must focus on et al., 2012), where amore centralized “country-led” (FAO, 2013) the wider environment to help overcome attribution difficulties, Biomass CAI platform for analysis of any geospatial location accounting for changes in both climate and economic state within these regions may be more useful and sustainable, also (Brooks et al., 2011; Dinshaw et al., 2014). The Biomass CAI minimizing expense (Romijn et al., 2012). can incorporate deep learning derived counterfactual baselines Importantly, NDVI measures are “blind” to many other accounting for weather patterns, combined with detailed spatial important metrics of adaptation (including socio-economic and temporal datasets due to the nature of SRS. However, the metrics), meaning the Biomass CAI cannot be used as integration of socioeconomic data is also key to overcoming a solitary tracking mechanism for adaptation monitoring. this challenge of attribution (Brooks et al., 2011; Dinshaw et al., The Biomass CAI will be most powerful when used with 2014). This is important because, if a score decreases on the other multi-criteria indicators measuring adaptation processes Biomass CAI, this could reflect the failure of an adaptation to and outcomes in frameworks such as TAAS (Ramasamy, improve resilience to climate shocks, or it could alternatively 2017). For example, it is important that the Biomass CAI reflect an increase in the cost of inputs (e.g., fertilizer, seeds, is complemented by “gender sensitive indicators” to assess irrigation) meaning farmers have less inputs overall. Despite the impact an adaptation intervention has on equality this, as any application of the Biomass CAI will focus on a and female empowerment, in addition to crop productivity particular geospatial location combined with counterfactual (FAO, 2013; Ramasamy, 2017; FAO and UNDP, 2019). The data from the same time period, end-users can have the Biomass CAI represents a quantitative and universally scalable flexibility to integrate the Biomass CAI into their broader indicator for adaptation tracking that can be integrated with assessments of the effectiveness of the adaptation intervention additional qualitative and quantitative data of relevance to (s) at their locations of interest. The barrier of attribution will the geospatial location subject to adaptation intervention(s). be a focus during the ongoing research and development of the Such a case-by-case approach using the Biomass CAI as a Biomass CAI. reference “anchor” can allow for improved inter-comparability The TAPAS Biomass CAI allows for use of satellite derived between adaptation intervention(s) at different locations, NDVI data appropriate for different resolutions (scales) and thereby improving understanding of the underlying processes questions (Figure 2B). We acknowledge a range of remote behind “successful” adaptation. Although the Biomass CAI sensing data sources will be required for different agricultural cannot directly address the time and expense required to systems, different crops, and different geospatial locations; collect data for other indicators (FAO, 2013; Ramasamy, 2017; requiring further research to enable accurate tracking and FAO and UNDP, 2019; Jacobs and Al-Azar, 2019); the TAPAS inclusivity by TAPAS. For instance, in perennial systems such program is exploring integration of crowdsourcing approaches as coffee, an extreme rainfall event may destroy flowers of into the Biomass CAI interface, to simultaneously assess crop coffee trees meaning production would decline, but high NDVI productivity and gather participatory socio-economic data from values would be recorded due to canopy growth. For rice- local stakeholders to feed into larger indicator frameworks (Ford based systems, techniques such as Synthetic Aperture Radar et al., 2016; Rosenstock et al., 2017; FAO and UNDP, 2019). (SAR) have been used to monitor and map rice productivity (Nelson et al., 2014; Setiyono et al., 2018), an option that TAPAS is exploring. Even where appropriate satellites and Discussion data sources are used, one of the major limitations of remote sensing as a means of data collection is that data needs to be Meeting the needs of the Biomass processed into a format that can be understood and analyzed Climate Adaptation Index end-users by end-users (Jacobs and Al-Azar, 2019; Ndungu et al., 2019). End-users must also be enabled to routinely use the Biomass The Biomass CAI requires careful selection of satellites due CAI for their geo-locations of interest, requiring user friendly to varying spatial and temporal resolutions. Figure 2B provides Frontiers inClimate 08 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 TABLE 2 Summary of characteristics needed for the Biomass CAI, barriers, and approaches to overcome such barriers. Biomass CAI Potential barriers Potential approaches to overcome barriers characteristics Multi-level agri-biomass Other RS data sources may be more appropriate for different • Further research and case studies to identify the most monitoring in areas of adaptation agricultural systems than NDVI appropriate satellites and RS data sources for different agricultural systems Ability to be used and interpreted Complex terminology and concepts that are difficult to • Communication between TAPAS program and end-users by end-users around the world interpret by end-users (FAO, 2013; FAO and UNDP, 2019; to ensure understanding (FAO, 2013; Ndungu et al., 2019) Jacobs and Al-Azar, 2019; Ndungu et al., 2019) • User-friendly GUI (Jacobs and Al-Azar, 2019; Ndungu et al., 2019) • Training manuals explaining how the Biomass CAI is calculated and to guide interpretation of results and trends (FAO and UNDP, 2019) The ‘digital divide’ in developing regions (Romijn et al., • Technical support and ‘South-South’ knowledge transfer 2012; Hilbert, 2016) (Romijn et al., 2012) • GUI for low-bandwidth mobile access in developing regions (Hilbert, 2016) • Development of ‘country-led’(FAO, 2013) central Biomass CAI platforms for specific regions (Romijn et al., 2012) Ability to attribute changes in crop The Biomass CAI may not be able to attribute the adaptation • The ongoing development of the TAPAS deep learning productivity to an adaptation as the cause of agri-biomass productivity change (Brooks derived counterfactual baseline (Brooks et al., 2011; intervention et al., 2011; Ford et al., 2013; Dinshaw et al., 2014; Ford and Dinshaw et al., 2014)in relation to different crops, Berrang-Ford, 2016; Ramasamy, 2017; Berrang-Ford et al., different growth periods and varying geospatial locations 2019; FAO and UNDP, 2019; Jacobs and Al-Azar, 2019; to ensure that results can be attributed to the adaptation Adaptation Committee, 2021) intervention. • Integration of economic data into the Biomass CAI (Brooks et al., 2011; Dinshaw et al., 2014) • Large spatial and temporal datasets (Ford et al., 2013) Ability to link with other Time and expense required to gather data for other • Crowdsourcing (Ford et al., 2016; Rosenstock et al., 2017; multi-criteria indicators in wider indicators using traditional methods such as surveys (FAO, FAO and UNDP, 2019) to gather data for other indicators monitoring frameworks 2013; Ramasamy, 2017; FAO and UNDP, 2019; Jacobs and [e.g. yield per hectare, inputs used, access to climate Al-Azar, 2019) information services (Ramasamy, 2017)] within the Biomass CAI interface an overview of the satellites the TAPAS program is exploring for 30 meters (Wulder et al., 2019) and 10–60 meters (Drusch et al., integration into our Biomass CAI platform for use at different 2012), respectively. scales, by different stakeholders and for different systems. Time-series biomass maps and current crop yield estimates Satellite data from satellites such as MODIS (Dinan et al., 2015) generated using SRS and NDVI have the potential to help are being integrated into the Biomass CAI platform to track various level stakeholders by providing information on NDVI at global, national and landscape-level. MODIS has a how yield varies over time and space and optimizing minimum spatial resolution of 250 meters, and is advantageous crop management (Yao et al., 2017; Shen and Evans, as it can calculate a 16 day NDVI composite to minimize 2021). Furthermore, it contributes to advanced crop interference from aspects such as clouds (Dinan et al., 2015). To and environmental analytical tools that assist farmers in measure adaptation progress at local resolution on smallholder implementing the appropriate management practices at the farmswith a typical size of two hectares (141 x 141meters) or less appropriate rates, times, and locations, hence, meeting both (Lowder et al., 2016), high spatial resolution satellites are needed economic and environmental targets (Khanal et al., 2017). to infer NDVI and enhance “pixel purity” without influence The use of RS and NDVI has the potential to improve the from other features such as non-agricultural vegetation, roads agricultural system by allowing stakeholders to conveniently and buildings (Duveiller and Defourny, 2010). Landsat (Wulder and cost-effectively collect, visualize, and evaluate crop status et al., 2019) and Sentinel- 2 (Drusch et al., 2012) would be and associated factors at various stages of production, as well appropriate satellites for this purpose, with spatial resolutions of as address problems quickly (Jung et al., 2021; Xu, 2021). Frontiers inClimate 09 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 Integration by the TAPAS program of multiple satellites to the and Lonsdale, 2005; Galarraga et al., 2011; FAO, 2013) are Biomass CAI platform will help recognize and mitigate any important actors involved in the implementation of adaptation bias and uncertainties arising from different satellite sensors plans and programmes as they have more interaction with and algorithms (Yang et al., 2013). The TAPAS project aims people than national stakeholders (Galarraga et al., 2011). to identify the satellites most applicable for different spatial This level of governance can enable the development of resolutions during the development of the Biomass CAI. policies informed by local needs, such as better allocation and management of resources at a landscape level (Rama Rao et al., 2022). Targets set by central government, such as in National and global-level stakeholders National Adaptation Plans (NAPs), can be developed by local National and global-level stakeholders such as the Parties government to take into account the specific socioeconomic, (governments) to the United Nations Framework Convention political, and environmental factors of a particular landscape (Ji on Climate Change (UNFCCC), climate and agri-financing et al., 2022). As farms in an agro-ecosystem are often connected entities, ministries, private sector companies and NGOs are (Veldkamp et al., 2001), discussion has focused on “Climate- central to the development of country-wide adaptation policies, Smart Landscapes” to ultimately align goals and create synergies plans and investments in the agriculture sector (Conde and between adaptation, mitigation and food security (Scherr et al., Lonsdale, 2005; FAO, 2013; Ford et al., 2013, 2015; Ford 2012). There is a need for tracking systems at the landscape and Berrang-Ford, 2016; Berrang-Ford et al., 2019). Enabling scale to monitor these synergies and to foster the development decision-making by such stakeholders is a fundamental utility of “Climate-Smart Landscapes” over time (Scherr et al., 2012). of the Biomass CAI. The UNFCCC’s signatory parties enacted By tracking the spatial progress of adaptation in a particular the Paris Agreement in 2015, representing a major step forward region or landscape, the Biomass CAI can aid regional-level in tackling climate change with both mitigation and adaptation stakeholders to identify and evaluate areas where adaptation recognized as key goals (UNFCCC, 2015). Indeed, Article 7 interventions are underperforming or subject to diminishing of the Paris Agreement indicates the need for monitoring resilience to climatic shocks. Landscape-level stakeholders, and evaluation systems to facilitate national reporting to the such as local government, can often be in a better position UNFCCC in the form of National Adaptation Plans (NAPs) to understand the local factors which may be affecting the and Nationally Determined Contributions (NDCs) (UNFCCC, performance of adaptation interventions. Based on a robust use 2015). NAPs aim to build on short term National Adaptation of Biomass CAI results with other indicators, revised landscape- Programmes of Action (NAPAs) by planning for longer level plans can be adjusted according to the specific needs of a term climate impacts, acknowledging that transformational region. If the Biomass CAI indicates a significant decrease in adaptation may be necessary (UNFCCC, 2012). The Biomass a region or values remain in the “Zone of Degradation”, the CAI can be integrated into monitoring and evaluation stakeholders can move forward with an “evidence base” that frameworks in NAPs and NDCs, tracking the progress of can aid the development and implementation of more tailored climate change adaptation interventions, identifying areas transformational adaptation interventions (Kates et al., 2012; where transformational adaptation needs are arising, and Park et al., 2012; Rickards and Howden, 2012; Chaffin et al., providing the UNFCCC process with a universally comparable 2016; Rippke et al., 2016; Vermeulen et al., 2018). “anchored” metric for measuring adaptation progress regarding the resilience of the photosynthetic biomass that humanity is dependent upon. The Biomass CAI can also be used by Farm-level stakeholders the climate investment community (e.g., Adaptation Fund, Farm-level stakeholders such as farm households, farmers Green Climate Fund, Development Banks, etc.) alongside organizations and value chain actors are vital stakeholders that existing project specific monitoring frameworks (Jacobs and Al- can work from the “bottom up” as part of a participatory Azar, 2019) to allow more appropriate targeting of adaptation approach to develop, implement and sustain adaptation projects investments (Ford et al., 2013). Subsequently, the Biomass CAI (Conde and Lonsdale, 2005; FAO, 2013; Sherman and Ford, will help foster transparent communication between donors 2014). Indeed, indigenous stakeholders have been noted as key and recipients (Dinshaw et al., 2014). There is potential for actors in the development and use of existing remote sensing synergies with the IFADAdaptation Framework Tool, where the decision support tools (Ndungu et al., 2019). Differences in Biomass CAI could augment the evidence base for both national farm sizes, exposures to climate stresses and access to adaptive adaptation planning and climate investments (IFAD, 2021). capacity all contribute to the challenge of effective measurement of adaptation interventions at farm-level. As it can measure across different geographic scales (including across farm-scales), Landscape-level stakeholders the Biomass CAI can be used by farm-level stakeholders for Landscape-level stakeholders such as regional governments, more specific “precision adaptation” than regional and national private sector companies, NGOs and financing entities (Conde stakeholders. We define precision adaptation as “climate change Frontiers inClimate 10 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 adaptation interventions which are spatially and temporally Conclusions targeted to have the most impact on climate change adaptation for any biophysical and/or socio-economic system with a Tracking climate change adaptation is imperative for defined geospatial boundary”. We consider that such “precision end-users in the agriculture sector to monitor adaptation adaptation” will be most powerful when focused on 10–60 meter progress and recognize priorities (Ford et al., 2013; Ford spatial resolution units (Drusch et al., 2012; Wulder et al., and Berrang-Ford, 2016) to inform evidence-based planning 2019) across the geospatial footprint of each farm. Biomass and investment regarding incremental and transformational CAI data can be used to classify individual farms according adaptation (Kates et al., 2012; Park et al., 2012; Rickards to their resilience to climate change, and hence their value and Howden, 2012; Chaffin et al., 2016; Rippke et al., 2016; locally and within agri-value chains. The integration of farmers Vermeulen et al., 2018) in response to climate projections that (including women and young people) in adaptation planning could decimate crop yields and reduce food security (Morton, and implementation processes is particularly important, with 2007; Wheeler and Von Braun, 2013; Lipper et al., 2014; Mbow discussions at the COP23 Koronivia Joint Work on Agriculture et al., 2019; Wing et al., 2021). Although different adaptation (KJWA) calling for the integration of such local stakeholders tracking approaches are challenged to secure agreement (and into adaptation discussion, planning and monitoring (FAO, adoption) by multiple stakeholders regarding their validity and 2019). Using ubiquitous mobile phone technology, farmers can effectiveness (Adaptation Committee, 2021), a recent report by also be enabled as key Biomass CAI end-users to show where the UNFCCC’s Adaptation Committee highlights the need for certain adaptations are performing best or worst on their farms, flexible tracking mechanisms and frameworks that can adapt to build capacity and knowledge for planning both incremental to more innovative approaches to data collection (Adaptation and transformational adaptations (Kates et al., 2012; Park et al., Committee, 2021). The Biomass CAI we are developing 2012; Rickards and Howden, 2012; Chaffin et al., 2016; Rippke presents a revolutionary and innovative approach for tracking et al., 2016; Vermeulen et al., 2018). biomass productivity in locations undergoing adaptation, utilizing remote sensing capabilities to deliver worldwide monitoring at farm, landscape, national and global-levels and Broadening the Biomass CAI end-users encompassing a deep learning predicted counterfactual baseline. Broadening the Biomass CAI end-users beyond the The Biomass CAI can support agricultural communities agriculture sector can also support wider adaptation and in sustaining resilience through appropriate incremental or conservation interventions (e.g., biodiversity, ocean and ecosystem services conservation). Natural vegetation and transformational adaptations in anticipation of adverse climate marine ecosystems are two of the most important carbon sinks change projections, linking to crop yield and economic data for effective mitigation (IPCC, 2018). Due to the capability of where available.Working with high-resolution satellites (Drusch SRS based vegetation indices to measure biomass of different et al., 2012; Wulder et al., 2019), the Biomass CAI can ecosystems, the Biomass CAI can be adapted to monitor also facilitate “precision adaptation”, allowing end-users to biomass productivity and conservation in forests (Zhu and Liu, channel context-specific adaptation interventions to particular 2015) and marine photosynthetic organisms [i.e., macroalgae areas. The potential for the Biomass CAI is evident and (Garcia et al., 2013) and microphytobenthos (Daggers et al., development of the Biomass CAI is currently underway in 2018)]. For example, many plant and tree species will need to our TAPAS program. We anticipate that future research and migrate with climate change to avoid a “migration lag” that partnerships will be needed to address knowledge gaps regarding can ultimately lead to their extinction (Corlett and Westcott, appropriate RS data sources for different crops, growth periods 2013). In anticipation of such ecological state shifts, there are and geospatial locations; the barrier of attribution (Brooks proposals that “assisted migration” can be implemented (Corlett et al., 2011; FAO, 2013; Ford et al., 2013; Dinshaw et al., and Westcott, 2013; Williams and Dumroese, 2013), where 2014; Ford and Berrang-Ford, 2016; Ramasamy, 2017; Berrang- monitoring is a critical aspect of tracking plant movements Ford et al., 2019; FAO and UNDP, 2019; Jacobs and Al- and assisted migration efforts (Corlett and Westcott, 2013; Azar, 2019; Adaptation Committee, 2021), the “digital divide” Williams and Dumroese, 2013). Indeed, SRS is suggested as an (Romijn et al., 2012; Hilbert, 2016), and the integration of important tool for biodiversity monitoring (Pereira et al., 2013) multidisciplinary indicator frameworks (Brooks et al., 2011; with major potential for the integration of the Biomass CAI into Ramasamy, 2017; Jacobs and Al-Azar, 2019) to ultimately biodiversity conservation initiatives and decision-making at provide a comprehensive overview of adaptation interventions multiple scales. that deliver resilience outcomes. Frontiers inClimate 11 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 Data availability statement (Ireland) for the Tracking Adaptation Progress in Agriculture (TAPAS) project (SFI Grant No: 19/FIP/AI/7515). This paper The original contributions presented in the study are arose from a collaborative research project between the Master’s included in the article/supplementary material, further inquiries degree program in Climate Change, Agriculture and Food can be directed to the corresponding author. Security of the University of Galway, and the TAPAS project. Part of this work was carried out under Real Time Monitoring of Food Systems Work Package of the Digital Innovation and Author contributions Transformation Initiative with the financial support of the CGIAR Systems Transformation Research Program | Digital The concept of a Biomass CAI was originated by AJ Innovation and Transformation Initiative under the project and developed by CS, AJ, AG, AL, and LR during the Tracking Climate Adaptation Progress in Agriculture. The Concept and Seed Phase of the Science Foundation Ireland– Initiative Digital Innovation and Transformation is carried out Department of Foreign Affairs and Trade (Ireland) funded with support from CGIAR Trust Fund Donors and through Tracking Adaptation Progress in Agriculture (TAPAS) project bilateral funding agreements. We would like to thank all funders (SFI Grant No: 19/FIP/AI/7515). Under supervision of CS and who supported this research through their contributions to the PM, AF developed a range of drafts of the manuscript, which CGIAR Trust Fund. were revised by CS and PM. Additional contributions to the manuscript were made by the TAPAS team members/authors. The manuscript was finalized by CS prior to submission. Conflict of interest All authors contributed to the article and approved the submitted version. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Funding This study was funded by CIAT Bioversity Alliance Publisher’s note Bioversity International - CIAT and Science Foundation Ireland. All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated Acknowledgments organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or CS, AG, and AJ acknowledge funding from Science claim that may be made by its manufacturer, is not guaranteed Foundation Ireland—Department of Foreign Affairs and Trade or endorsed by the publisher. References Adaptation Committee (2021). Considering approaches to reviewing the overall Anchang, J. Y., Prihodko, L., Kaptu,é, A. T., Ross, C. W., Ji, W., Kumar, S. S., progress made in achieving the global goal on adaptation. Bonn: AC/2021/TP/GGA, et al. (2019). Trends inWoody andHerbaceous Vegetation in the Savannas ofWest UNFCCC. Africa. Remote Sens. 11, 576. doi: 10.3390/rs11050576 Agard, J., Schipper, L. F., Birkmass, J., Campos, M., Dubex, C., Nojiri, Y., et al. Arab, S. T., Noguchi, R., Matsushita, S., and Ahamed, T. (2021). Prediction of (2014). “Annex II: Glossary”. IPCC. grape yields from time-series vegetation indices using satellite remote sens and a machine-learning approach. Remote Sens. Applicat: Society Env. 22, 100485. Aguilar, R., Zurita-Milla, R., Izquierdo-Verdiguier, E., and De By, doi: 10.1016/j.rsase.2021.100485 R. (2018). A cloud-based multi-temporal ensemble classifier to map smallholder farming systems. Remote Sens. 10, 729. doi: 10.3390/rs1005 Asrar, G. Q., Fuchs, M., Kanemasu, E. T., and Hatfield, J. L. (1984). Estimating 0729 absorbed photosynthetic radiation and leaf area index from spectral reflectance in wheat1. Agron. J. 76, 300–306. doi: 10.2134/agronj1984.00021962007600020029x Ahmad, I., Singh, A., Fahad, M., and Waqas, M. M. (2020a). Remote Sens.-based framework to predict and assess the interannual Atzberger, C. (2013). Advances in remote sens of agriculture: context description, variability of maize yields in Pakistan using Landsat imagery. existing operational monitoring systems and major information needs. Remote Comput. Electron. Agric. 178, 105732. doi: 10.1016/j.compag.2020.10 Sens. 5, 949–981. doi: 10.3390/rs5020949 5732 Ayhan, B., Kwan, C., Budavari, B., Kwan, L., Lu, Y., Perez, D., et al. (2020). Ahmad, R., Yang, B., Ettlin, G., Berger, A., and Rodríguez-Bocca, P. (2020b). A Vegetation detection using deep learning and conventional methods. Remote Sens. machine-learning based ConvLSTM architecture for NDVI forecasting. Int. Trans. 12, 2502. doi: 10.3390/rs12152502 Oper. Res. doi: 10.1111/itor.12887 Bento, V. A., Gouveia, C. M., Dacamara, C. C., Libonati, R., and Trigo, Allen, C. R., Angeler, D. G., Chaffin, B. C., Twidwell, D., and I. F. (2020). The roles of NDVI and land surface temperature when using Garmestani, A. (2019). Resilience reconciled. Nat. Sustain. 2, 898–900. the vegetation health index over dry regions. Glob. Planet. Change 190. doi: 10.1038/s41893-019-0401-4 doi: 10.1016/j.gloplacha.2020.103198 Frontiers inClimate 12 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 Berrang-Ford, L., Biesbroek, R., Ford, J. D., Lesnikowski, A., Tanabe, A.,Wang, F. adaptability and transformability. Ecol. Soc. 15, 20. doi: 10.5751/ES-03610-1 M., et al. (2019). Tracking global climate change adaptation among governments. 50420 Nat. Clim. Chang. 9, 440–449. doi: 10.1038/s41558-019-0490-0 Ford, J., Berrang-Ford, L., Biesbroek, R., Araos, M., Austin, S., and Lesnikowski, Brooks, N., Anderson, S., Ayers, J., Burton, I., and Tellam, I. (2011). “Tracking A. (2015). Adaptation tracking for a post-2015 climate agreement. Nat. Clim. adaptation and measuring development”, in IIED Climate Change Working Paper Chang. 5, 967–969. doi: 10.1038/nclimate2744 No. 1. IIED (London: IIED). Ford, J. D., and Berrang-Ford, L. (2016). The 4Cs of adaptation tracking: Chaffin, B. C., Garmestani, A. S., Gunderson, L. H., Benson, M. H., Angeler, D. consistency, comparability, comprehensiveness, coherency. Mitig. Adapt. Strateg. G., Arnold, C. A., et al. (2016). Transformative environmental governance. Annu. Glob. Chang. 21, 839–859. doi: 10.1007/s11027-014-9627-7 Rev. Environ. Resour. 41, 399–423. doi: 10.1146/annurev-environ-110615-085817 Ford, J. D., Berrang-Ford, L., Lesnikowski, A., Barrera, M., and Heymann, S. J. Conde, C., and Lonsdale, K. (2005). “Engaging stakeholders in the adaptation (2013). How to track adaptation to climate change: a typology of approaches for process,” in Adaptation Policy Frameworks for Climate Change. (Cambridge and national-level application. Ecol. Soc. 18, 40. doi: 10.5751/ES-05732-180340 New York: Cambridge University Press) p. 47–66. Ford, J. D., Tilleard, S. E., Berrang-Ford, L., Araos, M., Biesbroek, R., Corlett, R. T., and Westcott, D. A. (2013). Will plant movements keep up with Lesnikowski, A. C., et al. (2016). Opinion: Big data has big potential for climate change? Trends Ecol. Evol. 28, 482–488. doi: 10.1016/j.tree.2013.04.003 applications to climate change adaptation. Proc. Nat. Acad. Sci. 113, 10729–10732. doi: 10.1073/pnas.1614023113 Cui, C., Zhang,W., Hong, Z., andMeng, L. (2020). Forecasting NDVI inmultiple complex areas using neural network techniques combined feature engineering. Int. Galarraga, I., Gonzalez-Eguino, M., and Markandya, A. (2011). The role of J. Digit. Earth. 13, 1733–1749. doi: 10.1080/17538947.2020.1808718 regional governments in climate change policy. Environ. Policy Gov. 21, 164–182. doi: 10.1002/eet.572 Daggers, T. D., Kromkamp, J. C., Herman, P. M., and Van Der Wal, D. (2018). A model to assess microphytobenthic primary production in tidal Garcia, R. A., Fearns, P., Keesing, J. K., and Liu, D. (2013). Quantification of systems using satellite remote sens. Remote Sens. Environ. 211, 129–145. floating macroalgae blooms using the scaled algae index. J. Geophys. Res. 118, doi: 10.1016/j.rse.2018.03.037 26–42. doi: 10.1029/2012JC008292 Damian, J. M., Pias, O. H. D. C., Cherubin, M. R., Fonseca, A. Z. and Gnädinger, F., and Schmidhalter, U. (2017). Digital counts of maize plants by D., Fornari, E. Z., and Santi, A. L. (2020). Applying the NDVI from satellite unmanned aerial vehicles (UAVs). Remote Sens. 9, 544. doi: 10.3390/rs9060544 images in delimiting management zones for annual crops. Sci. Agric. 77. Grafton, R. Q., Doyen, L., Bén,é, C., Borgomeo, E., Brooks, K., Chu, L., doi: 10.1590/1678-992x-2018-0055 et al. (2019). Realizing resilience for decision-making. Nat. Sustain 2, 907–913. Dinan, K., Barreto Munoz, A., Solano, R., and Huete, A. (2015). “MODIS doi: 10.1038/s41893-019-0376-1 Vegetation Index User’s Guide (MOD13 Series)”. Arizona: The University Hassan-Esfahani, L., Torres-Rua, A., Jensen, A., and Mckee, M. (2017). Spatial of Arizona. root zone soil water content estimation in agricultural lands using bayesian-based Dinshaw, A., Fisher, S., Mcgray, H., Rai, N., and Schaar, J. (2014). “Monitoring artificial neural networks and high- resolution visual, NIR, and thermal imagery. and evaluation of climate change adaptation: methodological approaches”, in Irrig. Drain. Syst. 66, 273–288. doi: 10.1002/ird.2098 OECD Environment Working Papers (Paris: OECD). Hilbert, M. (2016). The bad news is that the digital access divide is here to stay: Douxchamps, S., Debevec, L., Giordano, M., and Barron, J. (2017). Domestically installed bandwidths among 172 countries for 1986–2014. Telecomm. Monitoring and evaluation of climate resilience for agricultural development– Policy 40, 567–581. doi: 10.1016/j.telpol.2016.01.006 a review of currently available tools. World Dev. Perspect. 5, 10–23. Holling, C. S. (1973). Resilience and stability of ecological systems. Annu. Rev. doi: 10.1016/j.wdp.2017.02.001 Ecol. Syst. 4, 1–23. doi: 10.1146/annurev.es.04.110173.000245 Drobnjak, S., Stojanovi,ć, M., Djordjevi,ć, D., Bakra,č, S., Jovanovi,ć, Howden, S., Crimp, S., and Nelson, R. (2010). “Australian agriculture in a J., and Djordjevi,ć, A. (2022). Testing a new ensemble vegetation climate of change”, in: Managing climate change: papers from the Greenhouse classification method based on deep learning and machine learning 2009 conference: Commonwealth Scientific and Industrial Research Organization methods using aerial photogrammetric images. Front. Environ. Sci. Eng. 10. (Canberra: CSIRO), p. 101–111. doi: 10.3389/fenvs.2022.896158 Htitiou, A., Boudhar, A., and Benabdelouahab, T. (2021). Deep learning-based Drusch, M., Del Bello, U., Carlier, S., Colin, O., Fernandez, V., Gascon, F., et al. spatiotemporal fusion approach for producing high-resolution NDVI time-series (2012). Sentinel-2: ESA’s optical high-resolution mission for GMES operational datasets. Can. J. Remote Sens. 47, 182–197. doi: 10.1080/07038992.2020.1865141 services. Remote Sens. Environ. 120, 25–36. doi: 10.1016/j.rse.2011.11.026 Huang, J., Wang, X., Li, X., Tian, H., and Pan, Z. (2013). Remotely sensed rice Duveiller, G., and Defourny, P. (2010). A conceptual framework to define the yield prediction using multi-temporal NDVI data derived from NOAA’s-AVHRR. spatial resolution requirements for agricultural monitoring using remote sens. PLoS ONE. 8, e70816. doi: 10.1371/journal.pone.0070816 Remote Sens. Environ. 114, 2637–2650. doi: 10.1016/j.rse.2010.06.001 Huang, S., Ming, B., Huang, Q., Leng, G., and Hou, B. (2017). A case study on a European Commision. (2018). “Commission Implementing Regulation (EU) combination NDVI forecasting model based on the entropy weight method.Water 2018/746 of 18 May 2018 amending Implementing Regulation (EU) No 809/2014 Resour. Manag. 31, 3667–3681. doi: 10.1007/s11269-017-1692-8 as regards modification of single applications and payment claims and checks”. European Commission (Brussels: EC). Huete, A., Justice, C., and Liu, H. (1994). Development of vegetation and soil indices for MODIS-EOS. Remote Sens. Environ. 49, 224–234. Evans, J., and Geerken, R. (2004). Discrimination between climate doi: 10.1016/0034-4257(94)90018-3 and human-induced dryland degradation. J. Arid Environ. 57, 535–554. doi: 10.1016/S0140-1963(03)00121-6 Ibrahim, Y. Z., Balzter, H., Kaduk, J., and Tucker, C. J. (2015). Land degradation assessment using residual trend analysis of GIMMS NDVI3g, soil moisture and FAO (2013). “Climate-Smart Agriculture Sourcebook”. Rome (Rome: Food and rainfall in Sub-SaharanWest Africa from 1982 to 2012. Remote Sens. 7, 5471–5494. Agricultural Organization of the United Nations). doi: 10.3390/rs70505471 FAO (2019). Koronivia Joint Work on Agriuclture: Analysis of submissions on IFAD (2021). Adaptation Framework Tool. Rome: International Fund for topics 2(b) and 2(c). Rome: Food and Agricultural Organization of the United Agricultural Development. Nations. IPCC (2018). “Global Warming of 1.5◦C. An IPCC Special Report on the impacts FAO and UNDP. (2019). Strengthening monitoring and evaluation for adaptation of global warming of 1.5◦C above pre-industrial levels and related global greenhouse planning in agriculture sectors. Rome. gas emission pathways, in the context of strengthening the global response to the Ferchichi, A., Abbes, A. B., Barra, V., and Farah, I. R. (2022). Forecasting threat of climate change, sustainable development, and efforts to eradicate poverty”, vegetation indices from spatio-temporal remotely sensed data using deep learning- Masson-Delmotte, V., Zhai, P., P?Rtner, H-O., Roberts, D., Skea, J., Shukla, P. R., based approaches: a systematic literature review. Ecol. Inform. 68, 101552. et al. (eds). doi: 10.1016/j.ecoinf.2022.101552 IPCC (2020). Climate Change and Land: An IPCC Special Report on climate Fernandez-Ordonez, Y. M., and Soria-Ruiz, J. (2017). “Maize crop yield change, desertification, land degradation, sustainable land management, food estimation with Remote Sens. and empirical models”, in 2017 IEEE International security, and greenhouse gas fluxes in terrestrial ecosystems. Geoscience and Remote Sens. Symposium (IGARSS) (New York, NY: IEEE). IPCC (2021). “Climate Change 2021: The Physical Science Basis. Contribution of doi: 10.1109/IGARSS.2017.8127638 Working Group I to the Sixth Assessment Report of the Intergovernmental Panel Folke, C., Carpenter, S. R., Walker, B., Scheffer, M., Chapin, T., on Climate Change”, (eds.) Masson-Delmotte, V., Zhai, P., Pirani, A., Connors, S., and Rockström, J. (2010). Resilience thinking: integrating resilience, Péan, L. C., Berger, S., et al. (eds). Cambridge University Press. Frontiers inClimate 13 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 Jacobs, H., and Al-Azar, R. (2019). “Dare to Understand and Measure (DaTUM). Measho, S., Chen, B., Pellikka, P., Guo, L., Zhang, H., Cai, D., et al. A literature review of Monitoring and Evaluation (M&E) frameworks for Climate- (2021). Assessment of vegetation dynamics and ecosystem resilience in the Smart Agriculture”. Rome: FAO. context of climate change and drought in the horn of Africa. Remote Sens. 13. doi: 10.3390/rs13091668 Ji, Q., Managi, S., and Zhang, D. (2022). Introduction to the special feature on managing climate risks for a sustainable future: adaptation strategies and Mishra, B., and Shahi, T. B. (2021). Deep learning-based framework for resilience-building. Sustain. Sci. 17, 1717–1721. doi: 10.1007/s11625-022-01229-5 spatiotemporal data fusion: an instance of Landsat 8 and Sentinel 2 NDVI. J. Appl. Remote Sens. 15. doi: 10.1117/1.JRS.15.034520 Jia, L., Li, Z. B., Xu, G. C., Ren, Z. P., Li, P., Cheng, Y. T., et al. (2020). Dynamic change of vegetation and its response to climate and topographic factors Moriondo, M., Maselli, F., and Bindi, M. (2007). A simple model of in the Xijiang River basin, China. Environ. Sci. Pollut. Res. Int. 27, 11637–11648. regional wheat yield based on NDVI data. Eur. J. Agron. 26, 266–274. doi: 10.1007/s11356-020-07692-w doi: 10.1016/j.eja.2006.10.007 Jin, X., Liu, S., Baret, F., Hemerl,é, M., and Comar, A. (2017). Estimates of plant Morton, J. F. (2007). The impact of climate change on smallholder density of wheat crops at emergence from very low altitude UAV imagery. Remote and subsistence agriculture. Proc. Nat. Acad. Sci. 104, 19680–19685. Sens. Environ. 198, 105–114. doi: 10.1016/j.rse.2017.06.007 doi: 10.1073/pnas.0701855104 Jones, L., Constas, M. A., Matthews, N., and Verkaart, S. (2021). Myneni, R. B., Hall, F. G., Sellers, P. J., and Marshak, A. L. (1995). The Advancing resilience measurement. Nat. Sustain. 4, 288–289. interpretation of spectral vegetation indexes. IEEE Transact. Geosci. Remote Sens. doi: 10.1038/s41893-020-00642-x 33, 481–486. doi: 10.1109/TGRS.1995.8746029 Jung, J., Maeda, M., Chang, A., Bhandari, M., Ashapure, A., and Landivar- Ndungu, L., Oware, M., Omondi, S., Wahome, A., Mugo, R., and Bowles, J. (2021). The potential of Remote Sens. and artificial intelligence as tools Adams, E. (2019). Application of MODIS NDVI for monitoring Kenyan to improve the resilience of agriculture production systems. Curr. Opin. Biotechnol. rangelands through a web based decision support tool. Front. Env. Sci. 7, 187. 70, 15–22. doi: 10.1016/j.copbio.2020.09.003 doi: 10.3389/fenvs.2019.00187 Kates, R. W., Travis, W. R., and Wilbanks, T. J. (2012). Transformational Nelson, A., Setiyono, T., Rala, A. B., Quicho, E. D., Raviz, J. V., Abonete, P. J., adaptation when incremental adaptations to climate change are insufficient. Proc. et al. (2014). Towards an operational SAR-based rice monitoring system in Asia: Nat. Acad. Sci. 109, 7156–7161. doi: 10.1073/pnas.1115521109 examples from 13 demonstration sites across Asia in the RIICE project. Remote Sens. 6, 10773–10812. doi: 10.3390/rs61110773 Khanal, S., Fulton, J., and Shearer, S. (2017). An overview of current and potential applications of thermal remote sens in precision agriculture. Comput. Electron. Ning, T., Liu, W., Lin, W., and Song, X. (2015). NDVI variation and its responses Agric. 139, 22–32. doi: 10.1016/j.compag.2017.05.001 to climate change on the northern loess plateau of China from 1998 to 2012. Adv. Meteorol. 2015, 1–10. doi: 10.1155/2015/725427 Kloos, S., Yuan, Y., Castelli, M., and Menzel, A. (2021). Agricultural drought detection with MODIS based vegetation health indices in southeast Germany. Nyamekye, C., Schönbrodt-Stitt, S., Amekudzi, L. K., Zoungrana, B. J. B., and Remote Sens. 13. doi: 10.3390/rs13193907 Thiel, M. (2021). Usage of MODIS NDVI to evaluate the effect of soil and water conservation measures on vegetation in Burkina Faso. Land Degradat. Dev. 32, Lamchin, M., Lee, W. K., Jeon, S. W., Wang, S. W., Lim, C. H., Song, C., 7–19. doi: 10.1002/ldr.3654 et al. (2018). Long-term trend of and correlation between vegetation greenness and climate variables in Asia based on satellite data. MethodsX. 5, 803–807. Park, S. E., Marshall, N. A., Jakku, E., Dowd, A. M., Howden, S. M., doi: 10.1016/j.mex.2018.07.006 Mendham, E., et al. (2012). Informing adaptation responses to climate change through theories of transformation. Global Env. Change. 22, 115–126. Lebrini, Y., Boudhar, A., Htitiou, A., Hadria, R., Lionboui, H., Bounoua, L., et al. doi: 10.1016/j.gloenvcha.2011.10.003 (2020). Remote monitoring of agricultural systems using NDVI time series and machine learning methods: a tool for an adaptive agricultural policy. Arabian J. Peng, D. L., Huang, J. F., Cai, C. X., Deng, R., and Xu, J. F. (2008). Assessing Geosci. 13, 1–14. doi: 10.1007/s12517-020-05789-7 the response of seasonal variation of net primary productivity to climate using Remote Sens. data and geographic information system techniques in Xinjiang. J. Li, C., Wang, J., Ge, L., Zhou, Y., and Zhou, S. (2022). Optimization of sample Integr. Plant Biol. 50, 1580–1588. doi: 10.1111/j.1744-7909.2008.00696.x construction based on NDVI for cultivated land quality prediction. Int. J. Environ. Res. Public Health 19, 7781. doi: 10.3390/ijerph19137781 Pereira, H. M., Ferrier, S., Walters, M., Geller, G. N., Jongman, R., Scholes, R. J., et al. (2013). Essential biodiversity variables. Science . 339, 277–278. Li, X., Yuan, W., and Dong, W. (2021). A machine learning method doi: 10.1126/science.1229931 for predicting vegetation indices in China. Remote Sens. 13, 1147. doi: 10.3390/rs13061147 Pettorelli, N., Vik, J. O., Mysterud, A., Gaillard, J.-M., Tucker, C. J., and Stenseth, N. C. (2005). Using the satellite-derived NDVI to assess Liao, L., Song, J., Wang, J., Xiao, Z., and Wang, J. (2016). Bayesian method for ecological responses to environmental change. Trends Ecol. Evol. 20, 503–510. building frequent landsat-like NDVI datasets by integrating MODIS and landsat doi: 10.1016/j.tree.2005.05.011 NDVI. Remote Sens. 8, 452. doi: 10.3390/rs8060452 Phan, P., Chen, N., Xu, L., Dao, D. M., and Dang, D. (2021). NDVI variation Lipper, L., Thornton, P., Campbell, B. M., Baedeker, T., Braimoh, A., Bwalya, and yield prediction in growing season: a case study with tea in Tanuyen Vietnam. M., et al. (2014). Climate-smart agriculture for food security. Nat. Clim. Chang. 4, Atmosphere. 12, 962. doi: 10.3390/atmos12080962 1068–1072. doi: 10.1038/nclimate2437 Piedallu, C., Chéret, V., Denux, J.-P., Perez, V., Azcona, J. S., Seynave, Liu, Y., Li, Y., Li, S., and Motesharrei, S. (2015). Spatial and temporal patterns of I., et al. (2019). Soil and climate differently impact NDVI patterns global NDVI trends: correlations with climate and human factors. Remote Sens. 7, according to the season and the stand type. Sci. Total Env. 651, 2874–2885. 13233–13250. doi: 10.3390/rs71013233 doi: 10.1016/j.scitotenv.2018.10.052 Lopresti, M. F., Di Bella, C.M., andDegioanni, A. J. (2015). Relationship between Rama Rao, C. A., Raju, B. M. K., Samuel, J., Rao, A. V. M. S., Kumar, R. MODIS-NDVI data and wheat yield: a case study in Northern Buenos Aires N., Srinivasa Rao, M., et al. (2022). Impact of climate change on productivity province, Argentina. Inf. Process. Agric. 2, 73–84. doi: 10.1016/j.inpa.2015.06.001 of food crops: a sub-national level assessment for India. Environ. Res. Commun. Lowder, S. K., Skoet, J., and Raney, T. (2016). The number, size, and distribution doi: 10.1088/2515-7620/ac8b68 of farms, smallholder farms, and family farms worldwide. World Dev. 87, 16–29. Ramasamy, S. (2017). “Tracking adaptation in agricultural sectors: climate doi: 10.1016/j.worlddev.2015.10.041 change adaptation indicators”. Food and Agriculture Organisation of the United Mann, M. L., Warner, J. M., and Malik, A. S. (2019). Predicting high- Nations (Rome: FAO). magnitude, low-frequency crop losses using machine learning: an application to Reddy, D. S., and Prasad, P. R. C. (2018). Prediction of vegetation dynamics cereal crops in Ethiopia. Clim. Change 154, 211–227. doi: 10.1007/s10584-019- using NDVI time series data and LSTM. Model. Earth Syst. Environ. 4, 409–419. 02432-7 doi: 10.1007/s40808-018-0431-3 Maselli, F., Chiesi, M., Angeli, L., Fibbi, L., Rapi, B., Romani, M., Reed, B. C., Brown, J. F., Van Der Zee, D., Loveland, T. R., Merchant, J. W., and et al. (2020). An improved NDVI-based method to predict actual Ohlen, D. O. (1994). Measuring phenological variability from satellite imagery. J. evapotranspiration of irrigated grasses and crops. Agri. Water Manag. 233, Vegetat. Sci. 5, 703–714. doi: 10.2307/3235884 106077. doi: 10.1016/j.agwat.2020.106077 Rhif, M., Ben Abbes, A., Martinez, B., and Farah, I. R. (2020). A deep learning Mbow, C., C., Rosenzweig, L. G., Barioni, T. G., Benton, M., Herrero, M., et al. approach for forecasting non-stationary big remote sens. time series. Arabian J. (2019). “Food Security. In: Climate Change and Land: an IPCC special report on Geosci. 13, 561–571. doi: 10.1007/s12517-020-06140-w climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems”, Shukla, J. S., Rickards, L., and Howden, S. M. (2012). Transformational adaptation: Calvo Buendia, E., Masson-Delmotte, V., Pörtner, H. O., Roberts, D. C., Zhai, P., agriculture and climate change. Crop Pasture Sci. 63, 240–250. et al. IPCC. doi: 10.1071/CP11172 Frontiers inClimate 14 frontiersin.org Ferguson et al. 10.3389/fclim.2022.938975 Rigden, A. J., Golden, C., and Huybers, P. (2022). Retrospective predictions of the landscape level. Soil Tillage Res. 58, 129–140. doi: 10.1016/S0167-1987(00)0 rice and other crop production in madagascar using soil moisture and an NDVI- 0163-X based calendar from 2010–2017. Remote Sens. 14, 1223. doi: 10.3390/rs14051223 Vermeulen, S. J., Dinesh, D., Howden, S. M., Cramer, L., and Thornton, P. K. Rippke, U., Ramirez-Villegas, J., Jarvis, A., Vermeulen, S. J., Parker, (2018). Transformation in practice: a review of empirical cases of transformational L., Mer, F., et al. (2016). Timescales of transformational climate change adaptation in agriculture under climate change. Front. Sustainable Food Syst. 2, 65. adaptation in sub-Saharan African agriculture. Nat. Clim. Chang. 6, 605–609. doi: 10.3389/fsufs.2018.00065 doi: 10.1038/nclimate2947 Von Keyserlingk, J., De Hoop, M., Mayor, A., Dekker, S., Rietkerk, M., and Romijn, E., Herold, M., Kooistra, L., Murdiyarso, D., and Verchot, L. Förster, S. (2021). Resilience of vegetation to drought: Studying the effect of grazing (2012). Assessing capacities of non-Annex I countries for national forest in aMediterranean rangeland using satellite time series. Remote Sens. Environ. 255, monitoring in the context of REDD+. Environ. Sci. Policy 19, 33–48. 112270. doi: 10.1016/j.rse.2020.112270 doi: 10.1016/j.envsci.2012.01.005 Vrieling, A., De Beurs, K. M., and Brown, M. E. (2011). Variability of African Rosenstock, T. S., Lamanna, C., Chesterman, S., Hammond, J., Kadiyala, farming systems from phenological analysis of NDVI time series. Clim. Change S., Luedeling, E., et al. (2017). When less is more: Innovations for tracking 109, 455–477. doi: 10.1007/s10584-011-0049-1 progress toward global targets. Curr. Opinion Env. Sustainab. 26–27, 54–61. Walker, B., Holling, C. S., Carpenter, S. R., and Kinzig, A. (2004). Resilience, doi: 10.1016/j.cosust.2017.02.010 adaptability and transformability in social–ecological systems. Ecol. Soc. 9. Rosenstock, T. S., Lamanna, C., Namoi, N., Arslan, A., and Richards, M. doi: 10.5751/ES-00650-090205 (2019). “What is the evidence base for climate-smart agriculture in East Wang, H., Li, Z., Cao, L., Feng, R., and Pan, Y. (2021). Response of NDVI and Southern Africa? A systematic map,” in The Climate-Smart Agriculture of natural vegetation to climate changes and drought in China. Land. 10, 966. Papers, eds T. Rosenstock, A. Nowak, and E. Girvetz (Cham: Springer). doi: 10.3390/land10090966 doi: 10.1007/978-3-319-92798-5_12 Washington-Allen, R. A., Ramsey, R., West, N. E., and Norton, B. E. (2008). Roy, B. (2021). Optimum machine learning algorithm selection for forecasting Quantification of the ecological resilience of drylands using digital remote sens. vegetation indices: MODIS NDVI and EVI. Remote Sens. Applications: Society and Ecol. Soc. 13. doi: 10.5751/ES-02489-130133 Environment. 23, 100582. doi: 10.1016/j.rsase.2021.100582 Wei, W., Wu, W., Li, Z., Yang, P., and Zhou, Q. (2015). Selecting Scherr, S. J., Shames, S., and Friedman, R. (2012). From climate-smart the optimal NDVI time-series reconstruction technique for crop agriculture to climate-smart landscapes. Agriculture Food Security. 1, 1–15. phenology detection. Intelligent Automat. Soft Comput. 22, 237–247. doi: 10.1186/2048-7010-1-12 doi: 10.1080/10798587.2015.1095482 Schiavon, E., Taramelli, A., Tornato, A., and Pierangeli, F. (2021). Monitoring Wheeler, T., and Von Braun, J. (2013). Climate change impacts on global food environmental and climate goals for European agriculture: user perspectives on the security. Science. 341, 508–13. doi: 10.1126/science.1239402 optimization of the Copernicus evolution offer. J. Environ. Manage. 296, 113121. doi: 10.1016/j.jenvman.2021.113121 Williams, M. I., and Dumroese, R. K. (2013). Preparing for climate change: forestry and assisted migration. J. Forest. 111, 287–297. doi: 10.5849/jof.13-016 Setiyono, T. D., Quicho, E. D., Gatti, L., Campos-Taberner, M., Busetto, L., Collivignarelli, F., et al. (2018). Spatial rice yield estimation based on MODIS Wing, I. S., De Cian, E., and Mistry, M. N. (2021). Global vulnerability and Sentinel-1 SAR data and ORYZA crop growth model. Remote Sens. 10, 293. of crop yields to climate change. J. Environ. Econ. Manage. 109, 102462. doi: 10.3390/rs10020293 doi: 10.1016/j.jeem.2021.102462 Shen, J., and Evans, F. H. (2021). The potential of landsat NDVI sequences to Wulder, M. A., Loveland, T. R., Roy, D. P., Crawford, C. J., Masek, J. G., explain wheat yield variation in fields in Western Australia. Remote Sens. 13, 2202. Woodcock, C. E., et al. (2019). Current status of Landsat program, science, and doi: 10.3390/rs13112202 applications. Remote Sens. Environ. 225, 127–147. doi: 10.1016/j.rse.2019.02.015 Sherman, M. H., and Ford, J. (2014). Stakeholder engagement in adaptation Xu, D. (2021). RETRACTED ARTICLE: Agricultural climate change interventions: an evaluation of projects in developing nations. Climate Policy. 14, based on remote sens image and emergency material supply management 417–441. doi: 10.1080/14693062.2014.859501 of agriculture, rural areas and farmers. Arabian J. Geosci. 14, 2076. doi: 10.1007/s12517-021-07221-0 Stepchenko, A., and Chizhov, J. (2015). NDVI short-term forecasting using recurrent neural networks. Int. J. Sci. Environ. Technol. 3. Yadav, K., and Geli, H. M. E. (2021). Prediction of crop yield for new mexico doi: 10.17770/etr2015vol3.167 based on climate and remote sens data for the 1920–2019 period. Land. 10, 1389. doi: 10.3390/land10121389 Strohmaier, R., Rioux, J., Seggel, A., Meybeck, A., Bernoux, M., Salvatore, M., et al. (2016). “The agriculture sectors in the Intended Nationally Determined Yang, J., Gong, P., Fu, R., Zhang, M., Chen, J., Liang, S., et al. (2013). The role Contributions: analysis”, in: Environment and Natural Resources Management of satellite Remote Sens. in climate change studies. Nat. Clim. Chang. 3, 875–883. Working Paper 62. doi: 10.1038/nclimate1908 The World Bank Group (2021). “Climate Change Action Plan 2021-2025, Yao, X., Wang, N., Liu, Y., Cheng, T., Tian, Y., Chen, Q., et al. (2017). Estimation Supporting Green, Resilient and Inclusive Development”. Washington DC. of wheat LAI at middle to high levels using unmanned aerial vehicle narrowband multispectral imagery. Remote Sens. 9, 1304. doi: 10.3390/rs9121304 Tucker, C. J. (1979). Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ. 8, 127–150. Zhang, Y., Liu, J., and Shen, W. (2022a). A review of ensemble doi: 10.1016/0034-4257(79)90013-0 learning algorithms used in remote sens. applications. Applied Sci. 12. doi: 10.3390/app12178654 UNFCCC (2012). “National Adaptation Plans - Technical guidelines for the national adaptation plan process”. Least Developed Countries Expert Group. Zhang, Y., Ma, J., Liang, S., Li, X., and Liu, J. (2022b). A stacking ensemble algorithm for improving the biases of forest aboveground biomass estimations UNFCCC (2015). “Adoption of the Paris Agreement, 21st Conference of from multiple remotely sensed datasets. GIScience Remote Sens. 59, 234–249. Parties”. Paris. doi: 10.1080/15481603.2021.2023842 UNFCCC (2021). Glossary of climate change acronyms and terms. Zhao, D., Wang, J., Zhao, X., and Triantafilis, J. (2022). Clay content mapping Varela, S., Dhodda, P., Hsu, W., Prasad, P. V., Assefa, Y., Peralta, N., et al. (2018). and uncertainty estimation using weighted model averaging. Catena. 209, 105791. Early-season stand count determination in corn via integration of imagery from doi: 10.1016/j.catena.2021.105791 unmanned aerial systems (UAS) and supervised learning techniques. Remote Sens. Zhu, X., and Liu, D. (2015). Improving forest aboveground biomass 10, 343. doi: 10.3390/rs10020343 estimation using seasonal Landsat NDVI time-series. ISPRS Journal of Veldkamp, A., Kok, K., De Koning, G., Schoorl, J., Sonneveld, M., and Photogrammetry and Remote Sens. 102, 222–231. doi: 10.1016/j.isprsjprs.2014. Verburg, P. H. (2001). Multi-scale system approaches in agronomic research at 08.014 Frontiers inClimate 15 frontiersin.org