PLANTdex:  An index to assess the multi-dimensional environmental impact of global crop commodities

doi:10.21203/rs.3.rs-3139724/v1

Download PDF

Article

PLANTdex: An index to assess the multi-dimensional environmental impact of global crop commodities

https://doi.org/10.21203/rs.3.rs-3139724/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

This study presents the croP environmentaL sustAiNabiliTy index (PLANTdex), one of the first spatially explicit (5 arcmin resolution) indexes to measure environmental impacts caused by the production of sixteen crop commodities globally. PLANTdex comprises five globally standardized environmental impact intensity indicators developed from robust spatially-explicit environmental models related to climate change, water stress and terrestrial and aquatic biodiversity loss.

Due to variations in PLANTdex, the index was able to differentiate production system efficiencies and environmental sensitivities, locating environmentally sustainable and unsustainable production globally. Due to PLANTdex's build, the index identified a more significant number of synergies (e.g., between biodiversity indicators) existed than trade-offs (e.g., between biodiversity-related GHG emissions and marine biodiversity loss) between indicators. The analysis identified over 32 regions that could collaborate on agricultural management strategies based on production, physical characteristics and environmental impact sensitivities that allow for the exchange of knowledge on sustainable agriculture. PLANTdex has the potential to support initiatives such as the Taskforce on Nature-related Financial Disclosures (TNFD) and E.S.G. strategies at the due diligence stage.

Earth and environmental sciences/Environmental sciences/Environmental impact

Scientific community and society/Agriculture

Due to population growth and socio-economic development, increased crop production for food, feed, fibre and fuel has led to intensive resource use, which imperils environmental integrity. Agriculture for crop and livestock production contributes to climate change by emitting 9–14% of greenhouse gas (GHG) emissions¹. Agricultural land use occupies 50% of the Earth's habitable land surface and causes up to 99% of deforestation in the tropics leading to biodiversity loss^2,3. Agriculture depletes surface and groundwater reserves, accounting for 70% of the global freshwater withdrawn⁴. Fertiliser use causes 78% of the global marine and freshwater eutrophication (the pollution of waterways with nutrient-rich pollutants impacting biodiversity)⁵. Additionally, agriculture accounts for 94% of the global mammal biomass (excluding humans), threatening 86% of species to the risk of extinction and species variability^6,7. Technological advancements do not necessarily cause excessive resource use or manifold environmental impacts. Instead, the policies and financial investments that promote the overuse of inputs that seep into the environment and the expansion of cultivation into environmentally-vulnerable regions that cannot sustain high levels of intensification are more impactful⁸.

The U.N. Sustainable Development Goals (S.D.G.)⁹ (which indicate environmental protection and better food security and nutrition as key goals), although global in scope, require sustainable commodity sourcing and informed food policy pathways at the sub-national and regional level to achieve their goals. This is particularly crucial for food production, which causes varying environmental impacts across food types and regions¹⁰. Previous global studies have made efforts to assess environmental (as well as economic and social) impacts caused by agriculture through building multi-indicator databases^{5, 11–15} (S.I Table 1). However, few are spatially-explicit below the national level^11,13,14, which is insufficient for evaluating regionally-specific environmental impacts, such as biodiversity and water stress^16–18. Additionally, many use embedded indicators that are weak and may not be scientifically evidence-based¹⁹; often relate to resource-use intensities rather than environmental impacts¹⁴; measure over different time periods¹³; relate to agriculture as a whole rather than to specific foods^{11, 13–15}; lack indicators relating to biodiversity loss (particularly eutrophication indicators¹²)⁵; or are not entirely spatially-explicit¹². Multi-indicator databases have often been aggregated into single scores to produce indexes. Whilst indexes have been popular tools to aid policy and investment decision-making in other domains, they have seldomly been devised at the global scale within agricultural studies. Béné et al., (2019) developed a global food system sustainability index at the national scale. However, their study lacked consistent data availability globally, making the correlations between their indicators inconclusive and their index a poor decision-making tool.

To achieve the U.N. SDGs, an increasing number of stakeholders outside of policy decision-making are interested in sustainability indexes and multi-indicator databases for food commodities. These stakeholders include corporations, non-profits, farmers, consumers and the general public²⁰. This interest is being driven by market lead initiatives, such as the Taskforce on Nature-related Financial Disclosures (TNFD), with a view that benchmarking and index rankings can raise awareness of nature sustainability factors whilst also helping to guide governments, banks, investors and those in the supply chain to engage with sustainable food systems and promote policy, investments and actions that reduce nature impacts. Considering this, various institutions, including the International Finance Corporation (IFC) and the UNEP World Conservation Monitoring Centre (WCMC), provide spatial tools to measure nature-related impacts. These tools are increasingly being used by banks, investors and development finance institutions who strongly advocate for using the tool. However, with the lack of globally standardized methodologies and limitations within current indicator databases, users cannot make quantitively informed and comparable decisions.

This study develops the global spatially-explicit croP environmentaL sustAiNabiliTy index (PLANTdex). PLANTdex measures at the 5 arcmin scale (10km x 10km) the environmental impacts caused by crop commodities. We used five emerging spatially-explicit environmental impact indicators for sixteen crop types circa 2000. Table 1 describes the five indicators related to Greenhouse Gas (GHG) Emissions (and estimated impact to terrestrial biodiversity), water stress and biodiversity loss (caused by land use and fertilizer use on land, freshwater and marine environments). PLANTdex is composed of intensity indicators measuring the environmental impact per unit crop (tonne of dry mass) produced. Measuring dry mass is a more reliable (relative to raw weight that includes water content) estimate of the environmental impacts caused by crops for their final use as food, feed and fibre.

Table 1

Environmental Impact Indicator, units and their description
Environmental Impact Indicator	Unit	Description
Greenhouse Gas emission (Carlson et al., 2016)¹⁷	PDF.yr.tonne^− 1	Estimates terrestrial biodiversity impacts from GHG emissions from the effects of rice paddy management, peatland draining, and nitrogen (N) fertilizer on CH₄, CO₂, N₂O emissions. GHG emissions are converted to terrestrial biodiversity impacts through LC IMPACT methodology²¹.
Freshwater Biodiversity Loss (Jwaideh et al., 2022)	PDF.yr.tonne^− 1	Estimates potentially disappearing fraction of fish species in freshwater environments for phosphorus fertilizer use.
Marine Biodiversity Loss (Jwaideh et al., 2022)	PDF.yr.tonne^− 1	Estimates potentially disappearing fraction of demersal species in marine environments for nitrogen fertilizer use
Land Biodiversity Loss (This paper in S.I, based on Chaudarry et al (2016)²²)	PDF.yr.tonne^− 1	Estimates potentially disappearing fractions of species across five taxa (mammals, birds, reptiles, amphibians, and vascular plants) caused by land occupation
Water Debt (Tuinetti et al, 2019)¹⁶	yr.tonne^− 1	Estimates the time required to replenish water resources (surface water, groundwater and evapotranspired)

2.1 Global Spatial Distributions

PLANTdex (production weighted averaged for all crops per grid cell) ranged from 0.13 to 0.90; however, most scores lay between 0.33 and 0.64(1st and 99th percentile). Due to variations in PLANTdex scores between crops at the grid cell level, average PLANTdex scores between 1st and 99th had a reduced range. Greater percentile score variability is observed for individual crops globally (SI Table 1.2 & 1.3). We identified high (worst) PLANTdex scores (> 0.52 [75th percentile]) in, but not exclusively, Central America, Southern Europe, the region between the Black and Caspian Sea, Southern Africa, South, East and South-east Asia (Fig. 1a). Regions where producing a tonne of any crop had the least impact included Mid-Western North America, parts of Central Europe, Central Russia, Western parts of South Asia and parts of Central Africa and Central South America. These regions had particularly low PLANTdex scores (< 0.43 [25th percentile]). High total impacts were found in, but not exclusively, South, South-East and East Asia, Mid-Western North America, Central South America, West Africa and Europe (Fig. 1b). Whilst some regions showed both high PLANTdex intensity impacts and high total impacts, e.g., East and South-East Asia, other regions exhibited the opposite, e.g., Western U.S.A., U.K and France.

2.2 Country-Level Crop Specific Distributions

We assessed production and impacts associated to countries for specific crops; Fig. 2 shows six example crops. The first plot identifies the total production and distribution of production at each grid cell, expanding over the highest 15 countries for production per specific crop. The percentage associated with each country for total global production tends to lie within a range of 0.1% up to 50%. The majority of crops showed intensive production in a few select countries. For example, oilpalm and soybean showed over 40% of the global production was found in Malaysia and the U.S.A., respectively. More heavily and extensively produced crops had lower maximum total production found in specific countries (e.g., Wheat). The second plot identifies the unweighted normalized environmental stressor plots, indicating dominant individual stressors within countries and variability. The third plot identifies the distribution of the PLANTdex scores across cells within countries and the total impact associated with each country. For all crops and countries, strong positive trends (r > 0.99) existed between total production and total environmental impacts (Fig. 2 plot 1 & 3). In plots 1 and 3, the boxplots identify the distribution of production and PLANTdex scores within each country. Boxplots with a large distribution indicate a high range in production and environmental impact intensities from different regions within a country. Due to variations in the PLANTdex and production distribution within countries, the ratio of global contribution between production and impacts in Fig. 2 Plot 4 was inconsistent per country and showed no clear trends for both intensities (red) and total (blue) ratios. Ratios above 1 indicate towards unsustainable production leading to excess environmental impacts (e.g., Oil Palm Production in Malaysia). Ratios below 1 indicate the environmental impacts are less than production contributions suggesting more sustainable production (e.g., Soybean production in Bolvia).

Ideally, production distributions which are negatively skewed (higher intensities of production per grid cell being achieved) along with positively skewed PLANTdex distributions (lower environmental impact intensities per grid cell being achieved) are desirable for global sustainable agricultural production systems of specific crops, e.g., soybean production in Bolivia (Fig. 2). Between countries, stronger negative correlations between total production and the PLANTdex is the ideal situation to reduce global environment impacts. A strong negative correlation suggests that heavy production of specific crops is occurring in countries with lower environmental impacts (more sustainable). Barley, maize, sorghum, soybean and sugarcane all showed negative correlations (r < -0.1 between total production and PLANTdex) with sugarcane showing the greatest negative correlations (r = -0.45). Whilst some crops had small correlations ranging between − 0.1 > r < 0.1, oilpalm, potato and rapeseed showed positive correlations (r > 0.1 between total production and the PLANTdex), which is extremely unsustainable, as most of the global production is occurring in countries with higher environmental impacts (e.g., Malaysia for Oilpalm). Oilpalm production in Malaysia showed excess contributions to global environment impacts through the ratio contribution indicator being greater than 1, whereas in D.C. Congo the indicator was less than 1. The normalized stressors plots (Fig. 2) suggested greater GHG emissions were the cause of greater oilpalm impacts found in Malaysia potentially due to higher CO₂ emissions from peatland soils found in SE Asia.

2.3 Crop-Specific Extreme Impact Analysis

Maise, wheat, rice, sugarcane and soybean represent 76% of the global total dry matter of the crops considered in this study. Hence these crops have the most impacting global PLANTdex scores (global average weighted by production), as total environmental impacts are heavily reliant on quantity of production (Fig. 3). To compare impacts across crops, we also present for every crop, the global production-weighted 75th percentile impact for each of the five environmental indicators and compare this to the global 75th (red line) and 50th (blue line) production weighted percentile of all crops (Fig. 3). This revealed extreme impacts from specific crops and environmental indicators (e.g., water stress due to Oilpalm production).

Extreme impacts for the top five most produced crops showed water stress impacts were consistently above the global 75th percentile for all of these five crops, but other environmental impacts varied when compared to the global 75th and 50th percentile (Fig. 3). Whilst rice's extreme impact to GHG emissions represented the 97th percentile compared to the global distribution for all crops, maize and sugarcane showed their extreme impacts to GHG emissions were below the global 50th percentile for all crops. With the least total production, rye was the only crop to show 75th-percentile impacts above the global 75th -percentile for all indicators. Some crops - rice, wheat, potato, cotton and rye - showed their extreme impacts verging close to or beyond the global 75th percentile. The crops with the least extreme impact, with their 75th -percentile impact below the global 50th -percentile to certain indicators included sugarcane, oilpalm and cassava for GHG emissions, sugarbeet for land biodiversity loss, and rapeseed for water stress.

2.4 Variability in impacts across different conditions

We assessed how specific crop production varies across different conditions (country income, crop production, temperature, precipitation, soil organic carbon content and slope incline) (Fig. 4). Whilst wheat and potato showed lower impacts in high production and high-income regions, the opposite was found for rice and oilpalm production. Rice produced in high production regions had greater impacts to biodiversity, in particular in marine environments from nitrogen fertilizer use. PLANTdex values for Oilpalm increased in high production regions due to high water stress and land biodiversity impacts. This is likely due to Oilpalm growing in wet, tropical regions with high biodiversity baselines (Indonesia) and regions with significant dry periods (West Africa), putting pressure on water resources. All crops, bar Oilpalm, showed strong reductions in water stress impacts in high production regions, suggesting strong water use efficiencies.

For physical factors, we identified PLANTdex scores increased with temperature, precipitation and slopes within crop-growing regions (Fig. 4). Due to varied results between environmental sterssors, soil Organic Carbon content varied PLANTdex less. Whilst marine and freshwater biodiversity impacts were found to increase with organic carbon content, water stress and land biodiversity impacts decreased with organic carbon contents for most crops, bar rice. GHG emissions showed variations between crops and attributes, with only increasing GHG impacts with increasing precipitation as a clear trend.

2.5 Trade-offs and Synergies Between Stressors

The trade-off and synergies between the environmental impacts considered in this study indicate complex relationships between environmental domains, that are not necessarily consistent amongst countries or crop types. When considering indicator pairings and countries alone, a greater number of synergies existed between indicator pairings than trade-offs. Of 3200 country samples (across all indicators and crop types), we found 1851 synergies, 937 trade-offs, 387 insignificant relationships, 25 insufficient data (Table 2).

Overall, whilst 31% of the indicator relationships were trade-offs, 57% were synergies. The strongest synergies (> 75% countries exhibiting synergies) generally involved indicators relating to biodiversity loss (Table 2 & S.I Fig. 2). Strong synergies also existed between land biodiversity loss and GHG emissions (63%) and marine biodiversity loss and water stress (71%) (Table 2). The strongest trade-offs were between GHG emissions and marine biodiversity loss (45%). Roughly equal numbers of trade-offs and synergies were identified with marine biodiversity paired with GHG emissions or land biodiversity loss. 12% of the indicator relationships showed no significant relationship and 1% had insufficient data to provide a conclusive result.

Table 2

Quantity of trade-offs, synergies, or no relationships between indicators at the country level for individual crops. Bold values represent relationships occurred over 50% count. We assessed the trade-offs and synergies among the five indicators that make the PLANTdex, through statistically significant (Spearman correlation p < 0.05) correlations between pairs of indicators over the 5 arcmin grid cell in each country and crop. For each crop, the top twenty countries in production were assessed, creating 320 samples for each indicator pairing.
Environmental Impacts	Trade-offs		Synergies		No significant relationships		Insufficient Data
	Count	%	Count	%	Count	%	Count	%
GHG vs WD	136	43	156	49	25	8	3	1
GHG vs OLBL	95	30	200	63	25	8	0	0
GHG vs NMBL	143	45	144	45	30	9	3	1
GHG vs PFBL	99	31	168	53	50	16	3	1
WD vs OLBL	120	38	135	42	58	18	7	2
WD vs NMBL	53	17	228	71	36	11	3	1
WD vs PFBL	106	33	145	45	69	22	0	0
OLBL vs NMBL	134	42	123	38	60	19	3	1
OLBL vs PFBL	47	15	242	76	31	10	0	0
NMBL vs PFBL	4	1	310	97	3	1	3	1
Total	937	31	1851	57	387	12	25	1

2.6 Policy Collaborative Framework Development

To develop collaborative frameworks, we aggregated the PLANTdex for all crops up to sub-national jurisdiction (e.g., the Kingdom of Scotland in the U.K and the state of Uttar Pradesh in India)²³. Sub-national jurisdiction frameworks were based on four important physical factors to agriculture (temperature, precipitation, soil organic carbon content and slope incline) and sub-national jurisdiction crop production levels. Economic factors were not considered, to allow the transgression of sustainable knowledge that may be found within higher income nations. Through different combinations of physical and production factors, based on the 50th percentile for all crop PLANTdex distributions, we formed 32 collaborative frameworks (Fig. 5).

PLANTdex having a high spatial resolution, allowed us to develop 32 collaborative environmental governance frameworks to encourage knowledge exchange towards sustainable agricultural practices (Fig. 5). Collaborative frameworks are based on different physical and production factors within certain income groups, based on the 50% percentile for all crop. Three characteristics of the collaborative frameworks were identified. Firstly, countries often had multiple regions within their borders that have similar physical factors suggesting potential to develop national scale frameworks (e.g., multiple states in the U.S.A). Secondly, frameworks often showed countries adjacent to one another, due to similar physical characteristics (e.g., regions in Eastern Europe). Thirdly, frameworks identified regions in different continents could be comparable. For example, for lower upper-middle income countries with good physical characteristics for agricultural production, we found that multiple regions in Central and South America showed suitable framework developments in the Middle East and East Asia.

PLANTdex having a high spatial resolution, allowed us to develop 32 collaborative environmental governance frameworks to encourage knowledge exchange towards sustainable agricultural practices (Fig. 5). Collaborative frameworks are based on different physical and production factors, based on the 50% percentile for all crops. Three characteristics of the collaborative frameworks were identified. Firstly, countries often had multiple regions within their borders with similar physical factors suggesting the potential to develop national scale frameworks (e.g., numerous states in the U.S.A). Secondly, frameworks often showed countries adjacent to one another due to similar physical characteristics (e.g., regions in Eastern Europe). Thirdly, frameworks identified areas in different continents that could be comparable. For example, for high production regions with good physical characteristics for agricultural production, the study found that multiple regions in South America showed suitable framework developments in Africa, South-east Asia and Australia (Fig. 5: collaborative framework 25).

Our results identify hotspots for environmental impacts found in regions of intense crop production. However, through PLANTdex, the study spatially differentiates crop production systems and their environmental impacts. Areas with high total impacts and PLANTdex scores indicated priority regions where agricultural intensification should be limited or practices adjusted to reduce impacts (e.g., South-east Asia and Central America). Whilst others showed low PLANTdex scores and low total impacts, indicating areas where potential intensification could occur (e.g., western U.S.A.). On the other hand, some regions showed low total impacts and high PLANTdex scores per tonne, such as South Africa and Madagascar, signifying environmentally sensitive regions where agricultural production under current practices should not expand.

At the country scale (both within and between), stronger negative correlations between production and PLANTdex is ideal for reducing global environmental impacts. Although all countries should constantly be adopting sustainable agricultural techniques, the global production vs impact ratio contribution indicator identified countries where sustainable intensification of crop production could potentially occur (countries with ratios below 1) and where countries with values greater than 1 may consider policies and techniques to reduce their global contribution to environmental impacts (Fig. 2). However, it may not be practical to vary production in particular countries or regions. Whilst dietary and cultural requirements may mean demand for particular crops is less in certain countries, other countries may not have the optimal environmental characteristics to intensify production. Additionally, producing crops further away from where they are consumed increases environmental impacts further down supply chain of the product's life cycle, through increasing transportation and distribution impacts. On the other hand, decreasing production may impact localized food security and the livelihoods of those who rely on specific crops for employment and income. The global production vs impact ratio contribution indicator provides an easy tool to support decision makers to assess crop commodity impacts in specific regions and can be used at the 5 arcmin scale to support investment & policy decision making.

Crop-specific extreme impact analysis showed no single crop had similar extreme impacts to specific indicators (Fig. 3). Some less heavily produced crops exhibited high extreme impacts across all indicators. However, as demands shift due to consumption and supply (e.g. for wheat alternatives due to recent conflict events²⁴), global policy efforts may use extreme impact analysis to channel their legislation towards reducing extreme impacts associated to specific crops as their demand increases (e.g. rye) (Fig. 3).

Trade-offs and synergies were exhibited by this research between impact indicators (Figs. 3 & 4). Stronger synergies existed between environmental indicators than trade-offs, particularly between biodiversity indicators; indicating high species densities and vulnerability can often exist in the same region across habitats (land, freshwater, marine). Water stress and biodiversity impact synergies may exist due to the increased vulnerability of species in environments with low surface water volumes, increasing the species densities over reduced areas and volumes for species to survive, which are often optimal agricultural growing regions too. Trade-offs between marine biodiversity loss and GHG emissions may be due to nitrogen impacting both GHG emissions and marine biodiversity loss. Hence, nitrogen removed from fields and entering marine receptors via hydrological pathways reduces the nitrogen available for denitrification and N2O (a GHG) emissions. Similarly, trade-offs between GHG emissions and freshwater biodiversity loss should be higher than indicated in this study, as nitrogen may also impact freshwater species. However, due to global eutrophication modelling limitations, only phosphorus is considered to impact freshwater environments²⁵. Additionally, synergies between GHG emissions and aquatic biodiversity can also exist due to excessive fertilizer use lending to high emissions to the atmosphere and aquatic environments. Other trade-offs may exist due to crop production in certain countries (e.g., rice farming in countries with a monsoon season, such as China having lower water stress impacts but still having high GHG emissions due to anaerobic decomposition in rice paddy fields- S.I. Figure 2), or variations in environmental impact spatial extents with minimal overlap (e.g., land and marine biodiversity loss in Uganda for cassava and sorghum - S.I. Figure 2).

Most indicator pairings did not show consistent trade-offs or synergies, suggesting policies used to reduce impacts in one indicator may impact another indicator adversely depending on location and crop type. One example is between GHG emissions and water stress indicators, where wheat and oil palm showed an average synergistic correlation of 0.33 (synergizing) and − 0.37 (trade off), respectively (S.I. Figure 2). This suggests policy used to reduce the impacts of water stress via irrigation strategies may also reduce GHG emissions for crops such as wheat, however, it may increase GHG impacts for oil palm production. Sustainable irrigation strategies have the potential to decrease GHG emissions. However, variability in the effect of GHG emissions exists between crops, irrigation strategies, gases emitted, and cultivation practices (e.g. whether fields are flooded)²⁶. Additionally, for some indicator pairings, countries had relatively consistent correlations across the various crop types, indicating the strong influence of the environmental sensitivity in those countries (S.I. Figure 2). In India, land and freshwater biodiversity loss consistently showed strong synergistic correlations (~ 0.43). Other indicator pairings showed varying correlations according to crop type for the same country, suggesting a strong influence of the cropping system resource use efficiencies due to excessive resource use intensities. For example, in China, correlations between GHG and water stress showed moderately strong trade-offs in rapeseed, groundnut and cotton production and moderately strong positive synergies in wheat, maize and potato.

Different economic and physical attributes may explain some of these relationships between indicators (Fig. 4). Lower impacts in both high-income and high-production regions suggest these regions benefit from more efficient farming practices, reducing environmental resource losses through economies of scale and investment. Low impacts from lower-income and production regions may be due to more suitable physical attributes for the growth of specific crops (e.g., high temperatures found in lower-middle-income countries are more suitable for oil palm and cassava production). Or the lack of agricultural inputs (e.g., water and fertilizers) reduces environmental losses. Considering physical factors, higher organic carbon contents improve soil structures for terrestrial biodiversity (e.g., plants and microorganisms) and increase soil drainage, replenishing groundwater reserves. However, better drainage can cause more significant nutrient leaching, particularly nitrogen, impacting marine biodiversity loss and reducing denitrification for GHG emissions. Where higher temperatures, rainfall and slopes exist, greater runoff pathways may exist within biodiverse sensitive regions, particularly around the tropics, leading to more significant impacts across biodiversity indicator

PLANTdex is evaluated here as a baseline impact year, circa 2000, for environmental impacts without considering the policy regulations, incentive programmes or sustainable agricultural practices that may alter the environmental impact of crop production in specific regions since 2000. This is due to the current quality and availability of agricultural data and environmental indicators for more recent years being inadequate²⁷. Hence, the novelty of the PLANTdex, having a high spatial resolution, allowed us to develop 32 collaborative environmental governance frameworks to encourage knowledge exchange for sustainable agricultural practices. The results identify transregional frameworks can be produced at the sub-national, national, and international levels with countries that are not necessarily within the same continent. The results also determined that a single policy framework may not be suitably applicable across all regions in one country and would need to be tailored for each crop type using the analysis within this study. Such spatially-explicit frameworks based on physical and economic attributes are critical to exchange knowledge to reduce environmental impacts and work towards the UN SDGs²⁷. Sustainable practices within these framework regions can be replicated in other regions within the same framework due to similar production, economic levels, physical characteristics and environmental sensitivities. Developing collaborative frameworks with similar attributes have previously been motivated in global agricultural research by grouping regions with similar climates²⁸

The high uncertainties identified through different PLANTdex builds (varying normalization, weighting and aggregation - Section 4.5), relayed the importance of a clear and transparent methodology that fits the distribution of the data and targeted extreme impact interests. The sensitivity analysis indicated the normalization and aggregation processes are dominant input factors to an index. As such, index development should occur with expert opinions who understand the indicator datasets, their potential correlations and good index formations depending on the ethos of the final index. The ethos of the index should be developed in collaboration with stakeholders and represents what the index is meant to achieve. In sustainability, this would refer to either soft or hard sustainability outcomes, along with understanding which environmental issues are of particular importance to stakeholders for weighting considerations²⁹.

PLANTdex has limitations (section 4.6), which include lack of stakeholder opinion and target thresholds, as well as data limitations embedded within the indicators, their interconnectivity and quantity of indicators. All of these can be improved within subsequent formations of PLANTdex and its embedded indicators. Nonetheless, PLANTdex's novelty lies in its high spatial resolution due to its spatial globally standardized indicators, developed from strong evidence-based environmental and biodiversity models. PLANTdex shortens the gap between global and local environmental impact profiling, providing the initial global analysis to assess region- and crop-specific environmental impact hotspots needed to measure biodiversity impacts. This enables crop commodity stakeholders to benchmark their environmental impact against others and supports corporate decision making for sustainable crop commodity trading and sourcing decisions, encouraging score improvement via comparable regional drivers for behaviour and mitigation solutions³⁰, whilst supporting initiatives such as the TNFD³¹.

We recommend further research into developing evidence-based indicators for good agricultural practices or policies to allow leveraging against PLANTdex scores, offset environmental impacts and support good agricultural practices in decision-making processes. We also recommend that livestock is also considered, as 65% of all agricultural land is used for pastoral farming.

4.1 Index Structure

Conceptually, indexes are based on sub-indicators, x, that may not have a common unit of measurement. An index, I, is built as follows:

$$I= \sum _{k=1}^{m}{w}_{k}{x}_{k}$$

Where, x is the normalized variable and $w$ is the weight attached to $x$. As such, the formula indicates that normalization and weighted aggregation of the variables are important steps for developing a index.

The reliability of an index strongly depends not only on the meaningfulness and reliability of the indicators embedded within, but also on the underlying methods used to construct the index³². The terminology of a "meaningful" indicator originated from Ebert and Welsch (2004)³³ and implies that the value of the index does not vary with different scale or underlying indicators. Significant scientific debates surround index methods; Sharpe (2004)³⁴ argued that producing an index is not appropriate to explain and compare complex observed phenomenon to capture the relative importance of separate components. Mayer (2008)³⁵ identified bias due to method choices for index construction, whilst Barclay (2019)³⁶ found studies lack clarity and transparency in their methodologies. Hence index formations may be misused to support desired policies³⁷. The various methods of index construction also present large uncertainties which should be reflected within a study's results³⁶. Therefore, to build a meaningful index, we highlight the significance of relevant methods for normalization, weighting and aggregation that fits the distribution of our data and targeted interests (e.g., extreme values). We also conduct an uncertainty and sensitivity analysis to understand the reliability of our results against other potential PLANTdex formations.

4.2 Normalization

Normalization is the process of transforming indicators with different units of measurements into dimensionless variables. It is an essential step before weighting and aggregation, however it increases the probability of uncertainty in measurement error due to manipulation of the data³⁸. Hence the choice of the normalization technique is an important factor. There are several normalization techniques which can be broadly categorized into linear, log and power normalizations. Although linear normalization is most commonly used, a modeller may choose a log or power normalization technique depending upon whether changes are more significant at lower end (log) or higher end (exponential or power) of the distribution. Here our interests lied within the extreme value of our dataset distributions to identify hotspots for environmental impacts. Hence, we applied Box-Cox transformations (power transformation) to accentuate extreme values of interests.

4.3 Weighting

The incorporation of subjective weighting, from expert and/or stakeholder opinion, is often sought after for places with specific issues, concerns, or goals. However, at the global scale, weighting is not possible due to multiple conflicting interests. Hence, we used an objective method: the CRITIC method (Criteria Importance through Intercriteria Correlation); a widely used statistical weighting³⁹. Statistical weighting has been widely used for decision making in recent years due to the advantage of avoidance of human interface; particularly when decision-makers have conflicting perspectives for the values of weights. In addition, statistical weighting reduces the impact of double counting of equal weighting by giving more precedence to indicators with low degrees of mutual correlation⁴⁰. Equal weighting is often applied in the absence of comprehensive understanding for sustainability. With the improvement of data collection techniques and the extensive research on sustainability, equal weightings have gradually been abandoned³².

The CRITIC method is a correlation method proposed by Diakoulaki et al. (1995)⁴¹, which uses standard deviation of the normalized indicator values and the correlation coefficients of all pairs of indicators to determine the indicator weights required, as its technique relies on multi-selection³⁹.

$${C}_{x}= { \sigma }_{x}\sum _{k=1}^{m}(1-{r}_{xk})$$

Where${C}_{x}$ is the CRITIC weighting coefficient, ${ \sigma }_{x}$ is the standard deviation of indicator variable x; m is the number of indicators; ${r}_{xk}$ is the correlation coefficient; $k$is the variable. The objective weight, w_x, is formed through normalization by the sum of weighting coefficients C_x:

$${w}_{x}=\frac{{C}_{x}}{\sum _{k=1}^{m}{C}_{k}}$$

	GHG	WD	LD	N	P
${\varvec{w}}_{\varvec{x}}$	0.1830	0.1615	0.2536	0.2889	0.1131

4.4 Aggregation technique

Aggregation techniques can be divided into compensatory and non-compensatory methods. According to Munda, "compensability refers to the existence of trade-offs, i.e. the possibility of offsetting a disadvantage on some criteria by a sufficiently large advantage on another criterion, whereas smaller advantages would not do the same" ⁴². Multiple sources of literature suggest the use of non-compensatory techniques are preferred to build sustainability indexs^43,44. The main argument for non-compensatory technique, is that the concept of sustainability is to avoid compensation (i.e. trade-offs) in environmental dimensions and implement holistic drivers towards improvement in environmental performance⁴⁴; This is also termed "strong sustainability" in some literature⁴⁵. However, often achieving optimal sustainability is not feasible in all cases to avoid trade-offs due to the interconnectivity between environmental systems. Nevertheless, a large majority of indexes use compensatory methods for their aggregation process to simplify and increase transparency for stakeholders and endpoint users. Non-compensatory outranking methods were initially considered for this study. However, high computational efforts were required to build the outranking matrix for the data at 5 arcmin resolutions. Keeping our data at the 5 arcmin resolution rather than aggregating to coarse resolutions is a novelty of this paper, therefore we choose compensatory methods which require less computational requirements. We include non-compensatory aggregation in our sensitivity analysis conducted at the national scale.

In this study, we use the TOPSIS (Technique for Order Performance by Similarity to Ideal Solution) aggregation, a compensatory technique based on the distance-to-target presented in Chen and Hwany(1992)⁴⁶. TOPSIS is based on the principle that the best performance (best ranked) of indicator sets should have the shortest geometric distance from the positive ideal solution (zero environmental impact) and the farthest geometric distance from the negative ideal solution (worst case environmental impact). According to Kim et al., (1997)⁴⁷ and our comparisons for suitable aggregation methods for this study, four TOPSIS advantages are identified:

A sound logical representation that mirrors the rationale of human choice (i.e., closest to the ideal solution);

A scalar value that accounts for both the best (positive ideal solutions: PLANTdex = 0) and worst solutions (negative ideal solution: PLANTdex = 1) simultaneously;

Provides a method that is easily programmed and requires moderate computational demands for large datasets

The aggregation goal for PLANTdex was for a strongest sustainability solution possible. Given non-compensatory aggregation methods require enormous computational powers at the 5 arcmin resolution. The TOPSIS aggregational methodology is less compensatory than standard extreme compensatory aggregation practices such as the sum and product rules.

We multiplied and normalized the PLANTdex index scores by production in each cell, to assess the total impact caused by production of all crops.

4.5 Index Uncertainty and Sensitivity Analysis

Uncertainty arises in indexes from the various steps in the indicator's construction:

Selection of sub indicators
Data selection
Data normalization
Weighting scheme
Aggregation technique

Here we assess the uncertainty of 9 normalization approaches, 6 weighting schemes and 9 aggregation techniques on wheat production at the country scale (assessing only countries which product over 0.5% of the global production for Wheat). This constructs a total of 486 different index designs which impact the final PLANTdex score. We used a Monte Carlo approach, which is based on multiple evaluations of the model with three randomly selected index model designs. Wheat was chosen due to it being one of the most heavily produced crops and production is more evenly spread out globally.

For the sensitivity analysis, we used variance-based techniques, which are most suitable for non-linear models to create the index⁴⁸. Variance based techniques should be used when several layers of uncertainty could be present simultaneously, making an index non-linear and non-additive⁴⁹. We used the first order sensitivity index to measure the fraction contribution of model output variance that is caused by uncertainty in each model input factors.

For non-linear aggregation index models, higher order sensitivity indices, which evaluate the interaction effects among multiple input factors, need to be calculated. However, higher order indices are computationally demanding, therefore total effect sensitivity indexes are formulated, which concentrate in one single term all the interactions involving a given input factor. The full sensitivity method is presented by Saltelli et al(2008)⁴⁹.

From the uncertainty analyze (Fig. 6.a) we can conclude the width of the 5th -95th percentile bounds and the ordering of the median rank for alternative indicator design combinations often differ from those of the original PLANT index rank. For some countries (e.g. Ethiopia and Turkey), the original PLANT index rank overlaps or is near to the median rank for different variations of the PLANT index. For half of the countries, our PLANT index rank lies within the interquartile range arbitrated by the PLANT index rank variations. Although the difference between the countries that are leaders and laggards can be relatively identified by the interquartile range rank across combinations, there are considerable differences between the median rank and original PLANT index rank (particularly for middle-of-the-plot countries) and large uncertainties in rank position when considering 5th -95th percentile rank bounds. Such high uncertainties within the index system solidifies two requirements within the development of an index:

1. A clear and transparent index methodology that not only fits the distribution of the data, but also targets the data of interest (i.e., extreme values).

2. the strong need for collaboration with stakeholders and expert opinion in the development to provide a useful tool for policy-decision makers and end users.

The first-order sensitivity analysis (Fig. 6b.) showed that the normalization method followed by aggregation and weighting methods are the most dominant input factors. However, we also identify a high degree of non-additive, non-linearity in the composite model due to several layers of uncertainty presented simultaneously. The non-additive that cannot be described through the first-order indices ranges from 31.7% in Belarus to 100% in Bulgaria. For most countries the non-additive part of the variance is over 50%. This underlines the requirement to compute higher-order indices that compute the simultaneous interaction effect among the input factors.

The total effect sensitivity indices (Fig. 6c) concentrate on one single term for all the interactions involving each input factor. As such the indices add up to a number greater than 1 due to the interactions between dominant input factors. If there were no interactions between the input factors (normalization, weighting and aggregation) the first order and total effect sensitivity indices would have summed to 1. Given most countries total effect sensitivity indices are greater than 1, the sensitivity indices show a high degree of non-linearity for the PLANTdex methodology and the importance of the interactions between different builds of the index.

As there are significant uncertainties in the value of the PLANTdex, one could assume that a country's ranking is irrelevant. However, here we have presented an exaggerated case to the methodology as our literature review has identified specific normalization and aggregation techniques which would not be suitable for this study (e.g., linear and log normalizations, equal weighting and additive aggregation).

Removing these methods from our sensitivity analysis would reduce the uncertainty found within the indicator. Therefore, a clearly defined, with expertise reasoning, on methodological choices is crucial to develop a meaning indicator.

4.6 Index Method Limitations

The design of the current PLANTdex is subject to limitations; which should be addressed in subsequent index formations. Firstly, the five environmental impact indicator datasets, used in this study, have been developed for crop production data from 1997–2003 (circa 2000) and were derived from studies available on earthstat.org^28,50,51. More recent crop production datasets, with temporal aspects up to 2014, are available from MapSPAM⁵² which covers 42 crops. Though, MapSPAM does not provide phosphorus and nitrogen fertilizer application rates for specific crops, which is essential for modelling fertilizer biodiversity impacts in fresh and marine waters. Nevertheless, updating the index with more current datasets (once crop-specific global fertilizer-use data becomes available) would improve this study by providing a temporal aspect and make it more current to the environmental situation arising today.

Secondly, an index should incorporate expert and stakeholder opinion at all stages of the development, to ensure index transparency and usefulness. Here we used a literature review to enhance and develop the PLANTdex, but a wider engagement of experts/stakeholders (out of scope for this study) could improve the PLANTdex applicability. In the absence of expert and stakeholder opinion, we used our literature review to determine the most suitable normalization and aggregation process for our data inputs and interests of outcomes. For the weighting, we used objective statistical weighting which have gained significant attention in recent years, particularly when conflicting stakeholder/expert perspectives occur between various environmental impacts.

Thirdly, an index heavily relies on the quality of data inputs. Although we used some of the most newly emerged environmental impact datasets, modelled a the 5 arcmin scale, there are underlying limitations and biases to each of the environmental models in this study, which are addressed in their respective papers.

Fourthly, the correlations between the environmental indicators is an important attribute which has not been fully modelled in this paper. Here we show the trade-off and synergy relationships through regression analysis. However, relationships exist between the environmental dimensions that can be inter-linked and modelled to show actual measurable trade-offs and synergies. One example of the interconnectivity not modelled in this study is between GHG emissions and nitrogen fertilizer-use impacts on marine environments. Nitrogen transportation can through its gaseous, liquid or solid phase and consequently impact GHG emissions through nitrous oxide (a greenhouse gas) release and cause eutrophication through runoff and atmospheric deposition. The connectivity of nitrogen in its gaseous, liquid and solid phase has not been shown in this study and is something that can be addressed in subsequent index formations.

Furthermore, here we used five indicators to describe five environmental impacts caused by crop production. However, there are multiple other environmental impacts that have not been included such as acidification, land degradation, deforestation and desertification. Additional meaningful indicators will provide a more substantial understanding of the overall complex environmental impacts of crop production.

The PLANTdex signifies baseline scenarios for environmental impacts without considering the policy regulations, incentive programmes or sustainable agricultural practices that may alter the environmental impact from crop production in specific regions. This is due to current quality and availability of such data being inadequate²⁷. Hence, the novelty of the PLANTdex having a high spatial resolution, allows us to develop collaborative environmental governance frameworks. Such frameworks allow for the exchange of knowledge on policy, incentive programmes and sustainable agricultural practices within regions of similar physical and economic attributes as well as environmental complexities^53,54.

4.7 Index Analysis Methodology

To understand how specific crop production varies across different conditions, we assessed the indicator distributions for four crops (wheat, rice, oilpalm and potato) over four important physical factors to agriculture (temperature, precipitation, soil organic carbon content and slope incline) and two economic factors (country income and grid cell crop production levels) (Fig. 3). Here we define each attribute into high, medium or low based on the attributes 33rd and 67th percentile (e.g., high production, low precipitation or medium slope) within their specific crop growing regions. For income, we split via four income groups (Low, Lower Middle, Upper Middle, High). We then assessed the percentage of PLANTdex and stressor values that represent high, average, or low impacts based on the 0.33 and 0.67 percentile of the respective stressor (e.g., GHG, W.D.).

Masson-Delmotte, V., Intergovernmental Panel on Climate Change, WMO, & United Nations Environment Programme. Climate change and land: an IPCC special report on climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems : summary for policymakers. (Intergovernmental Panel on Climate Change, 2019).
Ritchie, H. & Roser, M. Environmental impacts of food production. Our World in Data (2020).
Pendrill, F. et al. Disentangling the numbers behind agriculture-driven tropical deforestation. Science 377, eabm9267 (2022).
of the United Nations, F. and A. O. The State of the World’s Land and Water Resources for Food and Agriculture. (Routledge, 2013). doi:10.4324/9780203142837.
Poore, J. & Nemecek, T. Reducing food’s environmental impacts through producers and consumers. Science 360, 987–992 (2018).
Benton, T. G., Bieg, C., Harwatt, H., Pudasaini, R. & Wellesley, L. Food system impacts on biodiversity loss. 75.
Bar-On, Y. M., Phillips, R. & Milo, R. The biomass distribution on Earth. PNAS 115, 6506–6511 (2018).
Pretty, J. & Bharucha, Z. P. Sustainable intensification in agricultural systems. Annals of Botany 114, 1571–1596 (2014).
The State of Food Security and Nutrition in the World 2020. (FAO, IFAD, UNICEF, WFP and WHO, 2020). doi:10.4060/ca9692en.
Kanter, D. R. et al. Translating the Sustainable Development Goals into action: A participatory backcasting approach for developing national agricultural transformation pathways. Global Food Security 10, 71–79 (2016).
Chaudhary, A., Gustafson, D. & Mathys, A. Multi-indicator sustainability assessment of global food systems. Nature Communications 9, 848 (2018).
Gerten, D. et al. Feeding ten billion people is possible within four terrestrial planetary boundaries. Nat Sustain 3, 200–208 (2020).
Béné, C. et al. Global map and indicators of food system sustainability. Sci Data 6, 279 (2019).
Zhang, X., Yao, G., Vishwakarma, S. & Davidson, E. Measuring sustainable agriculture on a national scale. Earth and Space Science Open Archive http://www.essoar.org/doi/10.1002/essoar.10502709.1 (2020) doi:10.1002/essoar.10502709.1.
Springmann, M. et al. Options for keeping the food system within environmental limits. Nature 562, 519–525 (2018).
Tuninetti, M., Tamea, S. & Dalin, C. Water Debt Indicator Reveals Where Agricultural Water Use Exceeds Sustainable Levels. Water Resources Research 55, 2464–2477 (2019).
Carlson, K. M. et al. Greenhouse gas emissions intensity of global croplands. Nature Climate Change 7, 63–68 (2017).
Payen, S., Cosme, N. & Elliott, A. H. Freshwater eutrophication: spatially explicit fate factors for nitrogen and phosphorus emissions at the global scale. Int J Life Cycle Assess (2021) doi:10.1007/s11367-020-01847-0.
Global Map of Environmental and Social Risks in Agro-Commodity Product. https://www.ifc.org/wps/wcm/connect/Topics_Ext_Content/IFC_External_Corporate_Site/Sustainability-At-IFC/Company-Resources/GMAP.
Terazono, E. Food industry starts to measure sustainability. Financial Times (2021).
Verones, F. et al. LC-IMPACT: A regionalized life cycle damage assessment method. Journal of Industrial Ecology 24, 1201–1219 (2020).
Chaudhary, A., Verones, F., de Baan, L., Pfister, S. & Hellweg, S. 11. Land stress: Potential species loss from land use (global; PSSRg). 12.
Global Administrative Areas. GADM database of Global Administrative Areas, version 2.0. [online] URL: www.gadm.org. http://globe.umbc.edu/documentation-overview/global-administrative-areas-gadm/ (2012).
Wang, Y. & Jian, C. Sustainable plant-based ingredients as wheat flour substitutes in bread making. npj Sci Food 6, 49 (2022).
Jwaideh, M. A. A., Sutanudjaja, E. H. & Dalin, C. Global impacts of nitrogen and phosphorus fertiliser use for major crops on aquatic biodiversity. Int J Life Cycle Assess 27, 1058–1080 (2022).
Sapkota, A., Haghverdi, A., Avila, C. C. E. & Ying, S. C. Irrigation and Greenhouse Gas Emissions: A Review of Field-Based Studies. Soil Systems 4, 20 (2020).
Piñeiro, V. et al. A scoping review on incentives for adoption of sustainable agricultural practices and their outcomes. Nat Sustain 3, 809–820 (2020).
Mueller, N. D. et al. Closing yield gaps through nutrient and water management. Nature 490, 254–257 (2012).
Daly, H. E. Elements of environmental macroeconomics. Ecological economics: The science and management of sustainability 32–46 (1991).
Kelley, J. G. & Simmons, B. A. Politics by Number: Indicators as Social Pressure in International Relations. American Journal of Political Science 59, 55–70 (2015).
TNFD – Taskforce on Nature-related Financial Disclosures. TNFD https://tnfd.global/.
Zhou, P. & Zhang, L. P. Composite Indicators for Sustainability Assessment: Methodological Developments. in Energy, Environment and Transitional Green Growth in China (eds. Pang, R., Bai, X. & Lovell, K.) 15–36 (Springer, 2018). doi:10.1007/978-981-10-7919-1_2.
Ebert, U. & Welsch, H. Meaningful environmental indices: A social choice approach. Journal of Environmental Economics and Management 270–283.
Sharpe, A. Literature review of frameworks for macro-indicators. Literature review of frameworks for macro-indicators. 79 (2004).
Mayer, A. L. Strengths and weaknesses of common sustainability indices for multidimensional systems. Environment International 34, 277–291 (2008).
The problem with composite indicators | BMJ Quality & Safety. https://qualitysafety.bmj.com/content/28/4/338.abstract.
Nardo, M. et al. Handbook on Constructing Composite Indicators and User Guide. OECD Statistics Working Paper vol. 2005 (2008).
Hudrlíková L., Kramulová J. & Zeman J. Measuring Sustainable Development at the Lower Regional Level in the Czech Republic based on Composite Indicators : Measuring Sustainable Development in Czech LAU 1 Regions using Composite Indicators. Regional Statistics : journal of the Hungarian Central Statistical Office 3, 117–140 (2013).
Žižović, M., Miljković, B. & Marinković, D. Objective methods for determining criteria weight coefficients: A modification of the CRITIC method. Decision Making: Applications in Management and Engineering 3, 149–161 (2020).
Becker, W., Saisana, M., Paruolo, P. & Vandecasteele, I. Weights and importance in composite indicators: Closing the gap. Ecological Indicators 80, 12–22 (2017).
Diakoulaki, D., Mavrotas, G. & Papayannakis, L. Determining objective weights in multiple criteria problems: The critic method. Computers & Operations Research 22, 763–770 (1995).
Munda, G. “Measuring Sustainability”: A Multi-Criterion Framework. Environ Dev Sustain 7, 117–134 (2005).
El Gibari, S., Gómez, T. & Ruiz, F. Building composite indicators using multicriteria methods: a review. J Bus Econ 89, 1–24 (2019).
Greco, S., Ishizaka, A., Tasiou, M. & Torrisi, G. On the Methodological Framework of Composite Indices: A Review of the Issues of Weighting, Aggregation, and Robustness. Soc Indic Res 141, 61–94 (2019).
Barbier, E. & Burgess, J. Natural Resource Economics, Planetary Boundaries and Strong Sustainability. (2017) doi:10.3390/SU9101858.
Chen, S.-J. & Hwang, C.-L. Fuzzy Multiple Attribute Decision Making Methods. in Fuzzy Multiple Attribute Decision Making: Methods and Applications (eds. Chen, S.-J. & Hwang, C.-L.) 289–486 (Springer, 1992). doi:10.1007/978-3-642-46768-4_5.
Kim, G., Park, C. S. & Yoon, K. P. Identifying investment opportunities for advanced manufacturing systems with comparative-integrated performance measurement. International Journal of Production Economics 50, 23–33 (1997).
Saltelli, A., Tarantola, S. & Campolongo, F. Sensitivity Analysis as an Ingredient of Modeling. Statistical Science 15, 377–395 (2000).
Global sensitivity analysis: the primer. (John Wiley, 2008).
West, P. C. et al. Leverage points for improving global food security and the environment. Science 345, 325–328 (2014).
Monfreda, C., Ramankutty, N. & Foley, J. A. Farming the planet: 2. Geographic distribution of crop areas, yields, physiological types, and net primary production in the year 2000. Global Biogeochemical Cycles 22, (2008).
International Food Policy Research Institute. Global Spatially-Disaggregated Crop Production Statistics Data for 2010 Version 2.0. (2019) doi:10.7910/DVN/PRFF8V.
Bodin, Ö. Collaborative environmental governance: Achieving collective action in social-ecological systems. Science (2017) doi:10.1126/science.aan1114.
Collaborative Framework for Food Systems Transformation. A multi-stakeholder pathway for sustainable food systems. One Planet network https://www.oneplanetnetwork.org/knowledge-centre/resources/collaborative-framework-food-systems-transformation-multi-stakeholder (2019).

There is NO Competing Interest.

SupportingInformation.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

PLANTdex: An index to assess the multi-dimensional environmental impact of global crop commodities

Status:

Version 1

Abstract

Figures

1 Introduction

2 Results

2.1 Global Spatial Distributions

2.2 Country-Level Crop Specific Distributions

2.3 Crop-Specific Extreme Impact Analysis

2.4 Variability in impacts across different conditions

2.5 Trade-offs and Synergies Between Stressors

2.6 Policy Collaborative Framework Development

3 Discussion

4 Index Methodology

4.1 Index Structure

4.2 Normalization

4.3 Weighting

4.4 Aggregation technique

4.5 Index Uncertainty and Sensitivity Analysis

4.6 Index Method Limitations

4.7 Index Analysis Methodology

References

Additional Declarations

Supplementary Files

Status:

Version 1