During the 20^th century, agricultural revolutions driven by productivity transformed farming systems worldwide. In industrial countries, crop yields primarily increased through the intensification of inorganic fertilisation, pesticide use, modern plant breeding and mechanical tillage (Hazell and Wood 2007). While this intensification of crop production led to an unprecedented abundance of food at low prices (Hazell and Wood 2007, Bommarco et al. 2013), it also imposed high environmental and socio-economical costs. The environmental costs include biodiversity loss, natural and semi-natural habitat loss, soil degradation, water eutrophication, amplified net greenhouse gas emissions from agriculture and increased exposure of living beings to toxic agrochemicals (Foley et al. 2005, Tscharntke et al. 2005, Hazell and Wood 2007, Kremen and Miles 2012, Duru et al. 2015b). Socio-economically, intensive crop production coupled with the global market for crop products has contributed to poverty and loss of land access for less productive, small-scale farmers (Hazell and Wood 2007, Kremen and Miles 2012). Additionally, the cropping systems resulting from this intensification are unsustainable in the long term. Firstly, the environmental impacts listed above degrade agricultural areas, threatening future crop productivity and stability (Hazell and Wood 2007, Bommarco et al. 2013). Secondly, intensive crop production heavily depends on off-farm inputs, many of which are depletable resources like phosphorus for fertilisation or fossil fuel, either used directly for tractors or as an energy source for synthesising inputs like mineral nitrogen (Lin et al. 2008, Woods et al. 2010, Duru et al. 2015a, Dardonville et al. 2022).

Ecosystem Services (ES) are often highlighted as a key for transitioning from intensive to sustainable crop production (Foley et al. 2005, Zhang et al. 2007, Power 2010, Bommarco et al. 2013, Duru et al. 2015b, Palomo-Campesino et al. 2018). Defined as “the contributions that ecosystems make to human well-being” (Haines-Young and Potschin 2018, page III), ES can be categorised into three sections: provisioning, regulating (as an abbreviation for regulation and maintenance) and cultural services (Haines-Young and Potschin 2018). In agroecosystems, a crop field is a sub-unit that includes a cultivated plot and its surroundings, that humans manage and that simultaneously supplies ES and benefits from ES (Potschin-Young et al. 2018). In crop fields, provisioning ES include material and energy outputs useful to humans such as plants cultivated for food, fibre, genetic resources or biofuel synthesis. Regulating ES mediate or moderate the ambient environment affecting human health, safety or comfort. In crop fields, these latter can include climate regulation, soil quality regulation, pollination or biological pest control (Zhang et al. 2007, Potschin-Young et al. 2018). Cultural ES encompass non-material outputs that affect human physical and mental states. In crop fields, these latter can include artistic inspiration, recreation or education (Haines-Young and Potschin 2018). Conversely, ecosystems may generate dis-services, defined as “negative contributions to human well-being" or "undesired negative effects resulting from the generation of other ES” (Haines-Young and Potschin 2018). Pest damage to the crop and soil erosion are examples of dis-services in crop fields. In agroecosystems, the supply of ES and dis-services is shaped by both land-use management and pedoclimatic context (Duru et al. 2015b).

Calls to regenerate, maintain or enhance ES in agroecosystems gained momentum in the 2000s (Tscharntke et al. 2005, Dale and Polasky 2007, Power 2010) and are now central in agricultural and ecological studies (Palomo-Campesino et al. 2018). Various theoretical frameworks emphasise the importance of biodiversity and ecological processes for sustainable cropping systems, including “biodiversity-based agriculture”, “ecologically intensive agriculture”, “strong ecological modernisation of agriculture” (Duru et al. 2015b), “agroecology” (Wezel et al. 2013), “agroecological intensification” (Tscharntke et al. 2012), “biologically diversified farming systems” (Kremen and Miles 2012) and “ecological intensification” (Bommarco et al. 2013). While these approaches, which we will refer to as “ecologically intensive cropping systems” throughout this article, vary in their scopes, implementations and views on the future crop yield needed, they all share the goal of reducing reliance on off-farm inputs by utilising the natural functions of well-managed agroecosystems. By doing so, ecologically intensive cropping systems aim to replace non-renewable or energy-intensive inputs with ES and to reduce negative environmental externalities, offering a sustainable alternative to intensive crop production (Duru et al. 2015b).

The development of agroecosystem visualisations distinguishing between suppliers and beneficiaries of ES has supported the formalisation and promotion of ecologically intensive cropping systems. Zhang et al. (2007) proposed a framework that differentiates ES “to” and “from” the agroecosystem, later adapted as “input” and “output” ES by French researchers (Le Roux et al. 2008, Clermont-Dauphin et al. 2014, Duru et al. 2015b). This framework complements the three-section ES classification presented above (Haines-Young and Potschin 2018) by clarifying agroecosystem-specific dynamics. Input ES, provided by cultivated fields and their surroundings for the benefit of the farmer, support crop production through processes like pollination, soil retention, microclimate regulation and nutrient cycling, while input dis-services, such as pest damage and weed competition, impede it (Zhang et al. 2007). Farmers on one hand benefit from input ES as they can replace off-farm inputs for crop production (Dardonville et al. 2022) and, on the other hand, are affected by input dis-services. Output ES, delivered from agroecosystems for the benefit of society, include provisioning ES like food and fibre production, regulation ES like water purification, climate change mitigation or wildlife habitat and cultural ES like aesthetic landscapes. Some output ES – mostly provisioning ES - generate a direct income for the farmer, while others – mostly regulation and cultural ES – do not (Zhang et al. 2007). Output dis-services, such as greenhouse gas emissions or nutrient runoff, come from agroecosystems and negatively impact society. Using this framework, ecologically intensive cropping systems can be described as systems aiming to substitute off-farm inputs with input ES, while maximising output ES and minimising dis-services (Duru et al. 2015b). Some ES can be categorised simultaneously as input and output ES, as the farmer and the society both benefit from their supply: for example, the ES of water regulation impacts the farmer through water availability for cultivated plants on one side and the neighbourhood through the flows of water leaving the crop field on the other side. This framework of input and output ES is comprehensible and engaging for farmers and other agroecosystem stakeholders, as it highlights the mutual benefits of well-managed agroecosystems for farmers and society.

A growing body of literature supports the transition from intensive to ecologically intensive crop production (Duru et al. 2015b, Boeraeve et al. 2020). Agroecology terminology and objectives, as well as the concept of leveraging ES through agricultural management, are increasingly included in public policies (Dardonville et al. 2022, Van Ruymbeke et al. 2023). While agroecology extends far beyond ecologically intensive crop production – encompassing ecological, economic and social dimensions across entire food systems (Wezel et al. 2011) – its growing political recognition can support ecologically intensive agriculture. Specifically, two principles of agroecology at the agroecosystem level closely align with ecological intensification: (1) minimising exogenous, non-renewable inputs and (2) promoting key ecological processes and services (Altieri 1995). Furthermore, a substantial scientific groundwork has been built, with fundamental knowledge about ecological processes in agroecosystems and numerous theories on ecologically intensive cropping systems (Power 2010, Duru et al. 2015a, Duru et al. 2015b). Authors also agree on global principles for designing agroecosystems that enhance input ES, such as increasing plant diversity and soil cover, minimising mechanical and chemical soil disturbances and organising natural, semi-natural and cultivated areas to promote biological regulation (Duru et al. 2015b).

However, in industrial countries, farmers have not widely adopted ecologically intensive cropping systems (Palomo-Campesino et al. 2018). According to several authors, their practical implementation is notably hindered by the complexity and uncertainty involved in translating theoretical principles into site-specific cultivation decisions (Duru et al. 2015a, Palomo-Campesino et al. 2018). While many relationships between agricultural practices and ES are well understood at a general level, this knowledge is not always sufficient to anticipate the impact of a cropping system – and the various practices it comprises – on ES delivery in a given agroecosystem with its specific pedoclimatic conditions, erosion potential, soil fertility and pests and weeds communities (Palomo-Campesino et al. 2018). For example, while an increasing use of cover crops consistently has positive effects on the supply of pollination and erosion control across cropping systems and local conditions, its impact on crop production or climate regulation through carbon sequestration varies significantly in both magnitude and direction depending on studies (Palomo-Campesino et al. 2018, Van Ruymbeke et al. 2023). This illustrates the shift in innovation needed when moving from intensive to ecologically intensive cropping systems: more than before, general agronomic knowledge must be complemented by applied, local knowledge to support the informed adoption of context-specific agricultural practices (Duru et al. 2015b). Beyond knowledge-related challenges, a range of economic, organisational and structural barriers also contribute to the limited adoption of ecologically intensive systems – such as the cost of transition, access to adapted equipment, regulatory constraints and food chain organisation (Le Roux et al. 2008, Palomo-Campesino et al. 2018). These aspects, though beyond the scope of this paper, are important to acknowledge. Here, we focus on addressing knowledge gaps as a foundational step towards wider implementation.

Generating precise, site-specific knowledge requires carrying out field-based ES assessments that: (1) are strategically located to represent precise bio-physical conditions for soil, climate and environment (Palm et al. 2014, Duru et al. 2015a, Duru et al. 2015b); (2) consider multiple ES simultaneously to account for synergies and trade-offs between ecological processes (Power 2010, Seppelt et al. 2011, Bommarco et al. 2013, Deng et al. 2016); (3) integrate entire cropping systems rather than single cultivation practices (Kremen and Miles 2012, Palm et al. 2014, Boeraeve et al. 2020); and (4) are based on collaboration amongst scientists and stakeholders to ensure that findings are practical, applicable and support decision-making processes (Seppelt et al. 2011, Duru et al. 2015b). Assessments combining all these characteristics are emerging in industrial countries (Boeraeve et al. 2020, Chabert and Sarthou 2020).

In situ ES assessments are based on measurement methods that quantify ES supply through indicators, i.e. quantifiable biotic or abiotic variables measuring an ecosystem's capacity to provide ES (van Oudenhoven et al. 2012, Duru et al. 2015b, Vihervaara et al. 2018). Indicators vary in how closely they estimate a final ES, as conceptualised in the cascade model (Haines-Young and Potschin 2013a). In this model, the supply of ES is represented by a “cascade flow” including five levels: upstream, within the ecological system, (1) "ecosystem properties" influence (2) "ecosystem functions”, which in turn influence (3) the central level of “final ES” (TEEB 2010, van Oudenhoven et al. 2012, Potschin and Haines-Young 2016). Properties are biophysical structures of the ecosystem (Boerema et al. 2016) and functions are biological, chemical or physical changes or reactions of the ecosystem (TEEB 2010). In the centre of the cascade, "final ES" are the contributions that ecosystems make to the well-being of humans (TEEB 2010). Downstream, within the socio-economic system, (4) “benefits” and (5) “values” derive from ES supply. Benefits are the positive changes in human well-being coming from the fulfilment of needs and wants by ES and values are the economic worth of said benefits (TEEB 2010). The cascade model provides measurable entities for assessing ES (Boerema et al. 2016) and allows comprehensive and objective assessments by including both ecological and socio-economic indicators (van Oudenhoven et al. 2012, Kandziora et al. 2013).

A wide range of indicators has been used to assess ES in agroecosystems (Boerema et al. 2016, Vidaller and Dutoit 2022), several methodologies for choosing those indicators have been built (Dale and Polasky 2007, Seppelt et al. 2011, van Oudenhoven et al. 2012) and pre-made indicators sets, specific to certain agroecosystems, have been proposed (Dale and Polasky 2007, Ritz et al. 2009, Quinn et al. 2012, Dominati et al. 2014, Meyer et al. 2015, Thoumazeau et al. 2019, Tzilivakis et al. 2019). Nonetheless, comprehensive lists of indicators applicable specifically to agroecosystems remain scarce (but see Boerema et al. (2016), Paul et al. (2022), Vidaller and Dutoit (2022)) and, to our knowledge, no existing compilation focuses on indicators based on empirical data from direct crop field measurements. Yet, field-based assessments at the crop field scale are essential to understand how cultivation practices affect ES (Dale and Polasky 2007), which is necessary for advancing ecologically intensive cropping systems. Given the context-specific nature of those assessments (Duru et al. 2015b), researchers need access to an exhaustive collection of available indicators to select those most suitable for their studied agroecosystems.

This study aims to build, from a global literature review about ES indicators (Boerema et al. 2016), a clear and accessible collection of field-based indicators for assessing ES in crop fields with empirical measurements. Beyond compiling this collection, we also examine how the indicators are distributed across ES types, cascade levels and measurement methods, as well as their application in past research. Specifically, we address the following research questions:

1. In ES research, which provisioning and regulating ES are covered by field-based indicators at the crop-field scale?
2. How are these indicators distributed across the levels of the ES cascade model?
3. How have ES studies applied these indicators, particularly regarding the number of ES assessed and the integration of multiple levels of the ES cascade?

By answering these questions, we aim to provide a clearer picture of the possibilities and limitations of field-based ES assessments in crop fields, thereby supporting the development and implementation of ecologically intensive cropping systems through empirical ES studies.

Source of indicators

Filtering and structuring

From the 507 indicators identified by Boerema et al. (2016), we only kept indicators suited for in situ biophysical assessments of ES in crop fields, using four filtering criteria: ES section, applicability to crop fields, cascade level and data collection methods (Fig. 1). This process aimed to retain indicators supporting the creation of empirical, site-specific knowledge useful for ecologically intensive crop production. For the first criterion – section of ES – the review of Boerema et al. (2016) contained sufficient information for determining whether an indicator should be retained in our collection. The second, third and fourth criterion – applicability to crop field, cascade level and data collection method – required consulting source papers as the review was too synthetic. During this paper consultation, indicators were directly excluded if they failed to meet any criterion.

Figure 1.

Filtering process applied to the indicators of Ecosystem Services (ES) from Boerema et al. (2016), to retain a collection of indicators applicable to crop fields and quantifying ES in situ. Four criteria (framed) filtered the indicators: ES section, applicability of the indicator to crop fields, cascade level of the indicator and method of data collection. When necessary, the indicators retained were broken down into individual indicators.

In the first step of indicator selection, we excluded cultural ES because their provision depends on the broad landscape of a crop field to an extent that can overshadow the effects of local crop cultivation practices. Cultural ES requiring direct human interaction often rely on the crop field accessibility (Ridding et al. 2018), hence on landscape-level factors like the proximity to hiking trails or the presence of nearby schools. Additionally, these ES often arise from a configuration of multiple land covers rather than a single ecosystem (Aalders and Stanik 2019). Thus, the field scale – targeted by our collection - is not optimal for assessing cultural ES, which are typically evaluated at the landscape scale in agroecosystems (Vidaller and Dutoit 2022).

In the second step, we only retained indicators applicable to agroecosystems, specifically targeting crop fields as both an ecosystem type and a spatial scale for assessment. We focused on the crop field as it is critical for studying how agricultural practices affect ES, which is essential for creating site-specific knowledge to support ecologically intensive crop production. A crop field was defined as a cultivated plot and its immediate semi-natural surroundings, both of which influence ES provision. Concerning ecosystem type, this criterion excluded indicators irrelevant to crop fields, such as "Fish catch" for the ES of food production, which applies only to aquatic ecosystems. Within crop fields, we included indicators suitable for arable crops, permanent crops and temporary, non-grazed grasslands, without restricting to specific cultivation practices or regions, thereby ensuring broad applicability. Concerning spatial scale, we excluded indicators that could not be attributed to a single crop field. For example, we excluded "Groundwater availability" for the ES of water provision, as it depends on hydrological processes at the watershed scale and cannot be linked to individual fields. We mainly kept indicators measured within one crop field and its direct surroundings. However, we also retained indicators based on broader spatial observations if they supported the assessment of ES delivery for a specific field. For example, “Edge density of potential pest-controlling species habitat in the landscape” was retained as it refers to the landscape, but provides information for the biological control potential within the crop field, hence supporting field-level decision-making.

In the third step, we retained indicators for biophysical assessments and excluded those for monetary valuation. Biophysical indicators quantify ecosystem structures and functions, offering consistent, comparable data about the ecosystem's current state for assessments (Vihervaara et al. 2018). In contrast, monetary indicators reflect ES value within specific economic contexts, limiting comparability and reuse. For instance, a rising demand for a product can increase its perceived value and hence the monetary value of the ES providing it (van Oudenhoven et al. 2012). Consequently, we excluded indicators related to the value level of the cascade model and retained those related to the ecosystem property, ecosystem function, final ES (but there was none, as explained in "Sources of indicators") and human benefit levels, as well as dis-service indicators derived from biophysical measures. To aid future selections of indicators from our collection, each indicator’s cascade level was explicitly noted. Following Boerema et al. (2016), we categorised dis-services as distinct indicators linked to specific ES, rather than a separate, negative equivalent of an ES (as done by Zhang et al. (2007) or Power (2010)). We represented them as a special cascade level corresponding to the negative side of the benefit level (Fig. 2).

Figure 2.

Model of Ecosystem Services (ES) cascade (Haines-Young and Potschin 2013a) used to characterise the indicators of our collection. Each box represents an indicator level, in bold capital letters, with an illustrative indicator for the ES of biological control in agroecosystems, in italics. All illustrative examples come from our indicator collection, except for the “benefit to humans” level (*), as the filtration process did not retain any indicator at that level for biological control. Each upstream level, from ecosystem properties to ecosystem functions in the ecological system, directly influences the supply of ES, which, in turn, affects the downstream levels of benefits, dis-services and values within the socio-economic system. Compared to the original figure (Haines-Young and Potschin 2013a) that includes the boxes with plain lines, we added a dashed-lines box for the “dis-service level”, as a negative counterpart to the “Benefits to human” level. The black or orange outlines of the boxes respectively correpond to indicator levels retained or eliminated during our filtration process.

In the fourth step, we excluded indicators relying solely on non-direct data, such as modelling, expert judgement, theoretical study or data extraction. Non-direct data are useful for exploratory purposes, such as comparing ES across cropping system scenarios (Vihervaara et al. 2018). However, the lack of site-specific knowledge about how cropping systems affect ES limits their applicability. We prioritised field-based indicators that rely on empirical data, as these are essential for accurately quantifying ES and building this foundational knowledge. Hence, we retained indicators linked to methods collecting empirical data from the field via remote sensing or field observation - in situ analysis or sampling with ex situ analysis (Vidaller and Dutoit 2022).

After the filtering process, we simplified the retained indicators’ names from Boerema et al. (2016), synthesising them for clarity. For example, for biological control, "Vegetation - indicator: percentage of cover of weed species" from Boerema et al. (2016) was renamed "Weed species cover". We simultaneously stored descriptions of the protocols and units used to obtain indicators in each reference paper to avoid losing information through the name simplification. We also identified and decomposed assemblies of multiple indicators, presented as single indicators, by consulting the original protocols. These assemblies, often composite indicators summarising ES supply, were individualised, as presenting individual indicators allows researchers to select those most relevant to their study context. For instance, “Multiple soil properties: Once-off measurement: chemical properties: SOM, P, N, C, Al, base cations” from Boerema et al. (2016) was split into six indicators, one per soil component. Then, we eliminated redundancies amongst the newly-created individual indicators and the rest of the retained list.

We summarised this final indicator collection in a dataset detailing for each indicator the ES section, ES category in the framework “input and output ES”, ES name in Boerema classification, ES group, class and code in the CICES V.5.1, indicator level in the cascade model, original and simplified indicator name, DOI of reference paper(s) and main authors of reference paper(s). For each combination of ES indicator and reference paper, we also included the method of data collection, the unit of the indicator and a brief description of the protocol.

Analysis of the final collection

The indicator collection underpinning the analysis reported in this paper is available in different formats:

• A unique table including the whole dataset, with one line by combination of ES indicator and reference paper, deposited as an Excel file on the Zenodo repository (Leveau 2025, https://doi.org/10.5281/zenodo.15234008);
• A relational dataset that can be explored interactively through a Shiny application, accessible either online (https://plantmodelling.shinyapps.io/crop_field_es_indicators) or via local deployment on the R software of the scripts available on GitHub and Zenodo (Lobet and Leveau 2025);
• A unique table that is a simplified version of the dataset, excluding original indicator names, ES group, DOI of reference papers and descriptions of the protocols, available as Supplementary material (Suppl. material 1). An extract from this table, including only the "food production" ES, is presented in the Results section to illustrate the data collected on the indicators.

Out of the 507 indicators from Boerema et al. (2016), our filtering process yielded a collection of 59 indicators for field-based assessments of ES in crop fields. The first filtering step excluded 104 cultural ES indicators, while the subsequent steps – evaluating applicability to crop fields, cascade level and methods of data collection – excluded together an additional 344 indicators.

Out of the 59 retained indicators, 20 assemblies of multiple indicators were disaggregated, primarily in the ES of soil quality regulation, climate regulation and pollination (details in Leveau (2025)). After disaggregation and removal of redundancies, the final collection comprised 128 single indicators (Table 2).

Out of the 128 indicators, 121 were unique to one ES, while seven appeared in two ES. All duplicates were ecosystem property or ecosystem function indicators that related to soil quality regulation and either soil retention, water purification or climate regulation (Suppl. material 1). For example, soil available phosphorus concentration and soil nitrogen mineralisation rate appeared in both soil quality regulation and water purification. Interestingly, all the duplicates appeared after the disaggregation of multiple indicators into single ones, so they were not apparent in the original indicator list of Boerema et al. (2016). The existence of duplicates in ES indicators shows that the same properties or functions of an agroecosystem can influence the supply of different ES, which is well documented in agricultural ES research (Tibi and Therond 2024).

The 128 indicators in our collection showed an uneven distribution, whether across ES sections (provisioning or regulating), input and output ES categories, specific ES, cascade levels (Fig. 3) or methods of data collection (Suppl. material 1).

Figure 3.

Distribution of the 128 Ecosystem Services (ES) indicators retained from Boerema et al. (2016) after our filtering process. The left and right plots respectively display indicators from the sections of provisioning ES and regulation and maintenance ES. In each plot, ES are presented on the y-axis while the cascade level to which each indicator belongs is presented on the x-axis. For each ES, its category in the framework “input and output ES” is specified in parentheses: (I) = Input ES; (O+i) = Output ES with direct income for the farmer; (O-i) = Output ES without direct income for the farmer. The “value” level of the cascade is not represented on the x-axis, as it was filtered out during our selection process. For each combination of ES and cascade level, a green diamond represents the number of indicators available, indicated by its size and central number.

In terms of sections, 8% of the indicators related to provisioning ES, while 92% related to regulating ES. This disparity cannot be solely attributed to the number of ES in each section, as provisioning ES accounted for 44% (seven ES) of the total, compared to 56% (nine ES) for regulating ES.

In terms of input and output ES categories, the range of indicators per ES varies notably. Output ES providing direct income to farmers were represented by 1 - 4 indicators, while output ES not providing direct income were represented by 0 (for three ES) to 12 indicators. Input ES exhibited the widest range, as they were represented by 5 - 56 indicators.

The number of indicators per ES ranged from 0 - 56, with soil quality regulation (56) and pollination (18) having the most. Notably, no indicator met our filtering criteria for three ES – water provision, air quality regulation and life cycle maintenance – indicating that these ES have not been assessed empirically in crop fields in the studies reviewed by Boerema et al. (2016). Interestingly, all three are output ES without direct income for the farmer (Zhang et al. 2007, Le Roux et al. 2008).

The distribution of indicators across cascade levels varied considerably by ES section. In the provisioning section, only 20% of the indicators were related to ecosystem properties, while 80% were related to human benefits. Only two provisioning ES – food production and genetic resources – were measurable with indicators from the ecological system (e.g. properties and functions), whereas others only included indicators from the social and economic system (e.g. benefits and dis-services). This pattern was reversed for the regulation section, where 63% of indicators related to properties, 31% to functions and only 6% to human benefits or dis-services. Out of nine regulating ES, only four – biological control, pollination, soil retention and water purification – included indicators from the social and economic system. Besides, pollination was the only one including benefit indicators.

The retained indicators were derived from 66 articles reviewed by Boerema et al. (2016), published between 2006 and 2014. Half of these articles were published in four journals: Agriculture, Ecosystem & Environment; Applied Soil Ecology; Biological Conservation; and Ecological Indicators. Analysing cascade levels in the 66 reference papers provided insight into the methodologies of in situ biophysical assessments of ES in crop fields. Most articles (64%) assessed one ES, 22% assessed two and 14% assessed three to six. Remarkably, only 14% measured both provisioning and regulating ES. Only four papers used both ecological and socio-economic indicators: one evaluated pollination (function and benefit), one assessed soil retention (function and dis-service) and two evaluated water purification (property and dis-service or property, function and dis-service). Additionally, 24 papers assessed climate regulation, pollination, soil retention or soil quality regulation using a combination of property and function indicators. Most studies focused on a single cascade level per ES.

Finally, in terms of methods for data collection, each indicator was either only obtained through field observation or through remote sensing, even when mutliple reference papers measured it. Globally, 91% of indicators were based on field observation and 9% on remote sensing. Encouragingly, all ES, except water provision, air quality regulation and life cycle maintenance, could be quantified in crop fields through field observations, which globally covered all cascade levels. In contrast, remote sensing was only used for property indicators and was limited to food production, biological control, climate regulation, pollination and soil retention. Remote sensing data were either land-cover classes (arable crops, permanent crops, forests etc.), vegetation types combined with their spatial configurations or soil slope.

This study aimed to identify field-based indicators for assessing ecosystem services (ES) in crop fields and to highlight gaps in their distribution across ES types and cascade levels. In this discussion, we first examine the causes of disparities in indicator availability between ES, considering both methodological filters and broader research dynamics. We then explore and discuss the theoretical possibilities and real applications of integrating multiple levels of the ES cascade – particularly ecological and socio-economic aspects – within an ES assessment.

Causes of disparities in indicators availability amongst Ecosystem Services

Possibilities and realities of the integration of multiple cascade levels

The methodology for constructing our indicator collection provided a good foundation for designing field-based ES assessments, but has its biases and is open to improvements. This section examines three key methodological choices and their implications for future users.

Firstly, incorporating dis-services indicators as a distinct level within the cascade, while unconventional (Zhang et al. 2007, Power 2010), proved relevant and useful for in situ assessments at the field scale. Indeed, all our dis-service indicators – such as pest damages or soil loss from water erosion – represent negative consequences that can be mitigated by the corresponding ES – such as biological control or soil retention. This framework may enhance stakeholder engagement by highlighting the benefits of an ES through the assessment of its corresponding dis-services.

Secondly, we retained all indicators applicable to in situ, biophysical assessments in crop fields without filtering for quality, ensuring a broad range of options for future users, as each assessment has specific objectives and resources. However, not all indicators are equally suitable for assessing ES in a given context. Several guidelines propose quality criteria for the design of ES assessments that can support indicator selection for agroecosystems (Dale and Polasky 2007, Seppelt et al. 2011, van Oudenhoven et al. 2012, Duru et al. 2015b, Meyer et al. 2015, Boerema et al. 2016, Maes et al. 2016). We encourage users of our collection to consult this literature when selecting indicators. Although we did not formally assess indicator quality, our selection process implicitly addressed two quality criteria present in the literature: representativeness and comprehensiveness. The representativeness of an ES indicator is its capacity to accurately reflect the reality it approximates, by measuring states and processes that effectively translate into ES supply (Seppelt et al. 2011, Duru et al. 2015b, Boerema et al. 2016). Since ecosystem functions better reflect ES supply than ecosystem properties, the attribution of a cascade level to each indicator of our collection can help choose the most representative indicators for assessing the ecological aspects of an ES, for example, function indicators where available. Comprehensiveness, for its part, refers to the capacity of an indicator or, more frequently, a set of indicators, to represent all the key structural and compositional aspects of the ES under study (Dale and Polasky 2007, van Oudenhoven et al. 2012, Boerema et al. 2016). Once again, the presence of cascade levels in our dataset can help design assessments that integrate both the ecological aspects – via properties and function indicators – and the socio-economic aspects – via benefits and/or dis-services – of an ES. Such multidimensional assessments are more comprehensive than those focusing on one of the cascade aspects only (van Oudenhoven et al. 2012, Kandziora et al. 2013, Boerema et al. 2016).

Thirdly, our collection is inherently shaped by the methodology used by Boerema et al. (2016) to select reference papers and extract ES indicators. One key factor influencing our results is the specific list of ES categories and terminology they adopted. Some ES were deliberately excluded from their review because they were considered "supporting ES", a debated fourth ES section often omitted from assessments to avoid double counting. This concerns, for example, biodiversity, habitat and nutrient cycling. The argument of avoiding double counting appears justified when we consider that many indicators present in our collection would also fall under these excluded categories. For instance, all the indicators concerning auxiliaries habitats in pollination and biological control would be related to habitat and many indicators for water purification would be listed for nutrient cycling. Another bias of their ES terminology lies in the relatively narrow list of search terms used to identify reference papers for some ES. For example, biological control was only associated with "biological control" and "pest", whereas supplementary terms like "disease", "pathogen" and "regulation" could have expanded the reference corpus and resulted in a larger and more representative reference corpus and, thereby, in a more comprehensive list of indicators. The same limitation is, of course, applicable to our own corpus of 66 articles, as it is directly derived from the Boerema et al. (2016) review.

A second set of key factors of the Boerema et al. (2016) methodology that influences our results is that it limits references to literature published until 2014 and only includes papers explicitly stating that they measured an ES. We believe that the pronounced disparities in indicator distribution across ES and cascade levels are unlikely to change significantly with the inclusion of more recent studies – 2015 to 2025 – or broader literature – beyond explicit ES research. Many underlying causes – such as the scale of ES functioning, high costs of comprehensive field assessments, limited ES provision by crop fields and challenges in representing both ecological and socio-economic aspects – are not time- or discipline-dependent. Thus, while expanding the scope of this work to include recent studies and adjacent disciplines might add a few more indicators due to constant knowledge and technologies improvements, the overall trends identified in our collection are expected to persist. We identify three exceptions that could cause major changes in our collection: rapid innovations for assessing ES that has newly gained interest, measures coming from recent remote sensing development and scientific disciplines systematically using indicators relevant for ES measurement without labelling it as ES research.

For example, over the past decade, research on climate regulation and soil health has grown substantially in agroecosystems, driven by the urgency of addressing climate change and sustainable agriculture (El Chami et al. 2020, Christel et al. 2021, van Wesemael et al. 2023). The new insights that resulted from it led to the development of new indicators, such as soil enzymes production (Gao et al. 2020) or the ratio between actual and expected soil organic carbon (Poeplau and Don 2023) and to the refinement of existing indicators by measurement standardisation or new reference ranges.

Besides, remote sensing technologies have advanced in recent years and are increasingly used to estimate ES in agricultural landscapes (Defourny et al. 2019, Masenyama et al. 2022, Paula et al. 2023). Several new empirical measures at the crop-field scale have emerged: tree stratum mappings and various vegetation status indexes were used as property indicators for ES of plant production and climate regulation. Starting dates of plant greening, wilting and drying served for their part as property indicators for the ES of water regulation (as above-ground carbon stock) (West et al. 2019, Weiss et al. 2020, Masenyama et al. 2022, Sharma et al. 2022, Paula et al. 2023, Pouladi et al. 2023, van Wesemael et al. 2023).

Finally, for some ES lacking indicators in our collection, well-established biophysical indicators from adjacent disciplines may exist, even if these studies do not explicitly frame their work within the ES context. For instance, pollutant emissions and air quality variables used in research on agricultural impacts (Borghi et al. 2023) could serve as property or dis-service indicators for air quality regulation. However, many of these indicators are likely already known to ES researchers, but are deemed "too resource-intensive" for use in ES assessments, as extensive measures would be required to ensure representativity.

Considering these three exceptions, we advise future users to:

• Explore new remote sensing applications, especially for plant production, water regulation and climate-related ES. More globally, explore emergent technologies that empirically assess ES;
• Engage with researchers studying agroecosystems via adjacent disciplines when indicators are lacking in our collection, as it is an opportunity to uncover innovative perspectives and methodologies for ES assessments.

In summary, while our methodology for constructing the indicator collection has its limitations, it offers a robust starting point for field-based ES assessments. By acknowledging the potential biases and gaps in our approach, we encourage researchers to critically evaluate and complement our collection when necessary.

This study provides a comprehensive collection of field-based indicators for assessing ecosystem services (ES) in crop fields, derived through a rigorous filtering process applied to the global indicator list compiled by Boerema et al. (2016). By retaining only indicators linked to empirical, in situ data collection at the crop field scale – and focusing on ecosystem properties, functions, benefits and dis-services – we sought to bridge the gap between broad ES frameworks and the site-specific knowledge needed to implement ecologically intensive cropping systems.

Our final collection of 128 indicators, drawn from 66 peer-reviewed articles, supports assessments of both provisioning and regulating ES. However, it also reveals marked disparities in indicator availability across ES, cascade levels and data collection methods and gaps in current ES assessments. Regulating ES are well represented at the property and function levels, while provisioning ES are mostly associated with benefit and dis-service indicators. This imbalance highlights the ongoing challenge of integrating both ecological and socio-economic dimensions within comprehensive field-based assessments. The presence of property and function levels indicators for most input ES shows, however, that integration of levels within the ecological side of the ES cascade is theorically feasible, although rarely applied in the studies we reviewed.

Moreover, some output ES without direct market value were absent from our collection. These gaps can reflect methodological constraints, practical challenges or limitations in the original review’s scope. For water provision, the field scale appeared to be too small compared to the spatial scale at which water flows take place, which is rather the water basin in agricultural landscapes. For air quality regulation, crop fields are not considered as a hotspot for ES delivery, in contrast to ecosystems including perennial vegetation like forests or urban parks. Besides, carrying out empirical measures representative of a whole year or rotation of pollutants emissions and filtration is highly resource-intensive so modelling was often preferred to empirical assessments. Finally, the initial search terms used to obtain references for the ES of life cycle maintenance were probably too narrow to represent the real use of this ES in field-based assessments.

The findings of this research should encourage future users to not only rely on ES literature, but also to explore well-established indicators from adjacent biophysical disciplines. Such interdisciplinary exploration will likely yield additional insights into how cropping systems influence ES, ultimately enriching the framework for assessing and improving sustainable agricultural practices. Furthermore, it is important to remain vigilant for the emergence of new indicators in the ES literature, particularly in areas of growing research interest over the past decade, such as climate regulation or soil health. Advances in remote sensing technology, alongside increasing attention to ES like carbon sequestration and water regulation, may have led to the development of new indicators.

Although our collection is comprehensive within the constraints of its methodological scope, it remains open to future refinement as new indicators and methodologies emerge. By providing a flexible, yet robust set of indicators, we hope this work will serve as a valuable resource for researchers and practitioners aiming to enhance the ecological sustainability of cropping systems.

The authors would like to thank the reviewers for their detailed and constructive feedbacks on the manuscript.

L. Leveau : conceptualisation, data curation, formal analysis, investigation, methodology, project administration, resources, visualisation, writing - original drat, writing - review and editing ; N. Biot : writing - review and editing ; G. Lobet : data curation, software, visualisation ; P. Bertin : conceptualisation, funding aquisition, methodology, supervision, validation, writing - review and editing.