e.g. ‘owing to the lack of extramural funding, other important factors such as and extra funding for travel costs to scientific meetings were not provided’
CMOCs indicated in bold highlight the three cross‐cutting themes of time, identity and relationships.
ECRs = early‐career researchers.
As Table Table1 1 shows, the same intervention can lead to positive or negative outcomes depending on the particular contexts and mechanisms triggered. This highlights greater complexity than is evident at first glance. Cross‐cutting these four interventions were three mechanisms that were regularly identified as critical to the success (or not) of a research environment: time; researcher identities, and relationships. We now present key findings for each of these cross‐cutting mechanisms and discuss how their inter‐relations lead to our modified programme theory (Fig. (Fig.3). 3 ). Note that although we have tried to separate these three mechanisms for ease of reading, they were often messily entangled. Table Table2 2 presents quotes illustrating the way in which each mechanism mediates outcomes within particular circumstances.
Modified programme theory. ECR = early‐career researcher
Time, identity and relationships as cross‐cutting mechanisms mediating successful research environments
Quote no. | Mechanism | Quote |
---|---|---|
1 | Time: efficient use of time | ‘I never say I need more time because you could use that as an excuse for anything… But I think support in terms of being quite smart at aligning research activity to other activity you're involved in is quite important’ |
2 | Identity: internal motivation | ‘[For teacher researchers] inherent satisfaction and reward from research, rather than external praise and feedback, was certainly an indication of moving towards a research identity’ |
3 | Relationships: leadership | ‘From an institutional perspective, much depends on the perceived value of research and how it is actively supported by management, for example, in terms of study leave, time allocated for research and the impact of financial savings’ |
4 | Time and identity | ‘I say personal determination and resilience is a big factor because there are people who have been given some time and have then not delivered… I mean some of them are keen, they will say they have got no time and you know that is an interesting question about whether you make time or whether you have to wait for time to be given to you’ |
5 | Identity and leadership | ‘…research leadership as a “process through which academic values and identities are constructed, promoted and maintained”. Leadership is, therefore, central to establishing a healthy and vibrant research culture’ |
6 | Time and relationships | ‘We recognise that the sense of community developed over time would not have been possible without mutual trust and respect. This has been instrumental in creating a safe environment for both academic and personal development, and has in turn made it “possible to share problems without feeling uncomfortable”. Without a sense of trust it would also have been impossible for us to become more confident both in ourselves, as emerging academics, and in our work’ |
Time was identified as an important mechanism for mobilising research outcomes across our three disciplines. Time was conceptualised severally including as: protected time; workload pressures influencing time available; efficient use of time; flexible use of time; making time, and time in career. The two most commonly considered aspects were protected time and workload implications. Protected time was largely talked about in the negative across a variety of contexts and disciplines, with lack of protected time leading to lack of researcher engagement or inactivity and reduced research productivity. 32 , 35 , 37 , 41 , 44 , 47 , 49 , 61 , 62 , 63 , 67 Also across a variety of contexts and disciplines, and acting as a positive mechanism, available protected time was found to lead to increased research productivity and active research engagement. 31 , 36 , 40 , 48 , 49 , 63 , 65 With regard to workload, limitations on the time available for research imposed by excessive other workloads led to reduced research activity, lower research productivity, poor‐quality research and reduced opportunity to attend research training. 40 , 41 , 47 , 49 , 60 , 67 Juggling of multiple responsibilities, such as clinical, teaching, administrative and leadership roles, also inhibited research productivity by diminishing the time available for research. 35 , 40 , 49 The alignment of research with other non‐research work was described as driving efficiencies in the use of time leading to greater research productivity (Table (Table2, 2 , quote 1).
Identity was also an important mechanism for mobilising research outcomes across our three disciplines. Interpretations included personal identities (e.g. gender), professional identity (e.g. as a primary practitioner or a primary researcher), and social identity (e.g. sense of belongingness). Researcher identity was often referred to in relation to first‐career practitioners (and therefore second‐career researchers). Sharp et al. 48 defined these as participants recruited into higher education not directly from doctoral study but on the basis of their extensive ‘first‐order’ knowledge and pedagogical expertise. These were also practitioners conducting research in schools or hospitals. Identities were also referenced in relation to early, mid‐career or senior researchers. Academic staff working in academic institutions needed to develop a sense of researcher identity, belongingness, self‐efficacy for research and autonomy to increase their satisfaction, competence and research activity. 39 , 40 , 44 , 46 , 51 , 67 For first‐career practitioners (i.e. teachers, doctors), the research needed to be highly relevant and aligned to their primary identity work in order to motivate them. 53 , 59 , 62 , 65 This alignment was described as having a strong research–teaching nexus. 40 , 48 Linked to this concept was the need for first‐career practitioners to see the impact of research in relation to their primary work (e.g. patient‐ or student‐oriented) to facilitate motivation and to develop a researcher identity (Table (Table2, 2 , quote 2). 36 , 37 , 41 , 49 , 53 , 54 , 67 Where research was seen as irrelevant to primary identity work (e.g. English language teaching, general practice), there was research disengagement. 37 , 48 , 52 , 59 , 67
For all researchers and across our three disciplines, relationships were important in the mediating of successful research environments. 31 , 34 , 38 , 39 , 41 , 44 , 57 , 60 , 66 , 67 Positive research relationships were characterised by mutual trust and respect, 40 , 41 , 42 , 43 , 54 , 66 , 72 whereas others described them as friendships that take time to develop. 51 Mutually supportive relationships seemed to be particularly relevant to ECRs in terms of developing confidence, self‐esteem and research capacity and making identity transitions. 35 , 43 , 48 , 58 , 67 Relationships in the form of networks were considered to improve the quality of research through multicentre research and improved collaboration. 33 , 60 Supportive leadership as a particular form of relationship was an important mechanism in promoting a successful research environment. Supportive leaders needed to monitor workloads, set the vision, raise awareness of the value of research, and provide positive role‐modelling, thereby leading to increased productivity, promoting researcher identities and creating thriving research environments (Table (Table2, 2 , quote 3). 31 , 34 , 37 , 38 , 40 , 41 , 43 , 44 , 46 , 48 , 49 , 53 , 55 , 62 Research leadership, however, could be influenced negatively by the context of compliance and counting in current university cultures damaging relationships, creating a loss of motivation, and raising feelings of devalue. Indeed, the failure of leaders to recognise researcher identities led to negative research productivity. 36 , 37 , 38 , 43 , 46 , 48 , 49
Time and identity.
Time and identity intersected in interesting ways. Firstly, time was a necessary enabler for the development of a researcher identity. 37 , 38 , 41 , 48 , 49 , 54 , 59 , 61 , 63 , 65 , 67 , 69 Secondly, those who identified as researchers (thus holding primary researcher identities) used their time efficiently to favour research activity outcomes despite a lack of protected time. 35 , 43 Conversely, for other professors who lacked personal determination and resilience for research, having protected time did not lead to better research activity. 43 This highlights the fact that time alone is insufficient to support a successful research environment, and that it is how time is utilised and prioritised by researchers that really matters (Table (Table2, 2 , quote 4).
Interventions aimed at developing researcher identity consistently focused on relationship building across the three disciplines. The interventions that supported identity transitions into research included formal research training, 44 , 48 , 52 , 68 mentoring, 41 , 48 , 57 , 65 , 72 writing groups, 72 and collaboration with peers and other researchers, 39 , 41 , 43 operating through multiple mechanisms including relationships. The mechanisms included self‐esteem/confidence, increased networks, external recognition as a researcher, belongingness, and self‐efficacy. 35 , 41 , 43 , 44 , 45 , 52 , 57 Furthermore, our data suggest that leadership can be an enabler to the development of a researcher identity. In particular, leadership enabled research autonomy, recognition and empowerment, and fostered supportive mentoring environments, leading to researcher identity development and research productivity (Table (Table2, 2 , quote 5). 34 , 38 , 46 , 48
Relationships were developed and sustained over time (Table (Table2, 2 , quote 6). Across the three disciplines, the role of leaders (managers, directors, deans) was to acknowledge and raise awareness of research, and then to prioritise time for research against competing demands, leading to effective research networks, cohesion and collaboration. 31 , 34 , 38 , 43 , 46 , 48 , 49 , 50 , 53 , 55 , 70 Second‐career PhD students who did not invest time in establishing relationships with researchers in their new disciplines (as they already had strong supportive networks in their original disciplines) found that they had limited research networks following graduation. 48
Our initial programme theory was based on previous literature reviews 1 , 4 , 5 , 6 , 7 and on the REF2014 criteria. 10 , 21 However, we were able to develop a modified programme theory on the basis of our realist synthesis, which highlights novel findings in terms of what really matters for successful research environments. Firstly, we found that key interventions led to both positive (subjective and objective) and negative (subjective and objective) outcomes in various contexts. Interestingly, we did not identify any outcomes relating to research impact despite impact nowadays being considered a prominent marker of research success, alongside quantitative metrics such as number of publications, grant income and h‐indices. 21 Secondly, we found that disciplinary contexts appeared to be less influential than individual, local and institutional contexts. Finally, our modified programme theory demonstrates a complex interplay among three cross‐cutting mechanisms (time, researcher identity and relationships) as mechanisms underpinning both successful and unsuccessful research environments.
Our research supports the findings of earlier reviews 1 , 5 , 6 , 7 regarding the importance of having a clear research strategy, an organisation that values research, research‐oriented leadership, access to resources (such as people, funding, research facilities and time), and meaningful relationships. However, our research extends these findings considerably by flagging up the indication that a clear linear relationship, whereby the presence of these interventions will necessarily result in a successful research environment, does not exist. For example, instituting a research strategy can have negative effects if the indicators are seen as overly narrow in focus or output‐oriented. 38 , 40 , 46 , 47 , 64 Similarly, project money can lead to the employment of more part‐time staff on fixed‐term contracts, which results in instability, turnover and lack of research team expertise. 40 , 67 , 71
Our findings indicate that the interplays among time, identity and relationships are important considerations when implementing interventions promoting research environments. Although time was identified as an important mechanism affecting research outcomes within the majority of papers, researcher identity positively affected research outcomes even in time‐poor situations. Indeed, we found that identity acted as a mechanism for research productivity that could overcome limited time through individuals efficiently finding time to prioritise research through their motivation and resilience. 35 , 43 Time was therefore more than just time spent doing research, but also included investment in developing a researcher identity and relationships with other researchers over time. 37 , 38 , 41 , 48 , 49 , 54 , 59 , 61 , 63 , 67 , 69 Relationship‐building interventions were also found to be effective in supporting difficult identity transitions into research faced by ECRs and those with first‐career practitioner backgrounds. Supportive leadership, as a particular form of relationship, could be seen as an enabler to the provision of protected time and a reasonable workload, allowing time for research and for researcher identity formation. 34 , 38 , 46 , 48 Indeed, our realist synthesis findings highlight the central importance of researcher identity and thus offer a novel explanation for why research environments may not flourish even in the presence of a research strategy, resources (e.g. time) and valuing of research.
Researcher identity is complex and intersects with other identities such as those of practitioner, teacher, leader and so on. Brew et al. 39 , 73 , 74 explored researcher identification and productivity by asking researchers if they considered themselves to be ‘research‐active’ and part of a research team. Those who identified as researchers prioritised their work differently: those who were highly productive prioritised research, whereas those in the low‐productivity group prioritised teaching. 73 Interestingly, highly productive researchers tended to view research as a social phenomenon with publications, presentations and grants being ‘traded’ in academic networks. Brew et al. 39 explain that: ‘…the trading view relates to a self‐generating researcher identity. Researcher identity develops in the act of publication, networks, collaborations and peer review. These activities support a person's identification as a researcher. They also, in turn, influence performance measures and metrics.’ Although the relationships among identity, identification and productivity are clearly complex, we explored a broader range of metrics in our realist synthesis than just productivity.
This is the first study to explore this important topic using realist synthesis to better understand the influence of context and how particular interventions lead to outcomes. We followed RAMESES 20 guidelines and adopted a rigorous team‐based approach to each analytic stage, conducting regular quality checks. The search was not exhaustive as we could have ‘exploded’ the interventions and performed a comprehensive review of each in its own right (e.g. mentoring). However, for pragmatic reasons and to answer our broad research questions, we chose not to do this, as suggested by Wong et al. 20 Although all members of the team had been involved in realist syntheses previously, the process remained messy as we dealt with complex phenomena. The messiness often lies in untangling CMOCs and identifying recurrent patterns in the large amounts of literature reviewed.
Our findings suggest that interventions related to research strategy, people, IIF and collaboration are supported under the ‘right’ conditions. We need to focus on time, identity and relationships (including leadership) in order to better mobilise the interventions to promote successful research environments.
Individuals need to reflect on how and why they identify as researchers, including their conceptions of research and their working towards the development of a researcher identity such that research is internally motivated rather than just externally driven. Those who are second‐career researchers or those with significant teaching or practitioner roles could seek to align research with their practice while they establish wider research networks.
We recommend that research leaders support individuals to develop their researcher identity, be seen to value research, recognise that research takes time, and provide access to opportunities promoting research capacity building, strong relationships and collaboration. Leaders, for example, may introduce interventions that promote researcher identities and build research relationships (e.g. collaborations, networking, mentoring, research groups etc.), paying attention to the ways in which competitive or collaborative cultures are fostered. Browne et al. 75 recently recommended discussions around four categories for promoting identity transition: reflection on self (values, experiences and expectations); consideration of the situation (circumstances, concerns); support (what is available and what is needed), and strategies (personal strategies to cope with change and thrive). With the professionalisation of medical education, 76 research units are increasingly likely to contain a mixture of first‐ and second‐career researchers, and our review suggests that discussions about conceptions of research and researcher identity would be valuable.
Finally, organisations need to value research and provide access to resources and research capacity‐building activities. Within the managerialist cultures of HEIs, compliance and counting have already become dominant discourses in terms of promotion and success. Policymakers should therefore consider ways in which HEIs recognise, incentivise and reward research in all its forms (including subjective and objective measures of quantity, quality and impact) to determine the full effects of their policies on research environments.
Future research would benefit from further exploration of the interplay among time, identities and relationships (including leadership) in different contexts using realist evaluation. 77 Specifically, as part of realist approaches, longitudinal audio‐diaries 78 could be employed to explore researcher identity transitions over time, particularly for first‐career practitioners transitioning into second‐career researchers.
RA and CER were responsible for the conception of the synthesis. All authors contributed to the protocol development. RA and PESC carried out the database searches. All authors sifted for relevance and rigour, analysed the papers and contributed to the writing of the article. All authors approved the final manuscript for publication.
Ethical approval.
not required.
Table S1. Definitions of key terms.
Table S6. Contexts, interventions, mechanisms and outcomes identified in individual studies.
we thank Andy Jackson, Learning and Teaching Librarian, University of Dundee, Dundee, UK, for his advice and help in developing our literature searches. We also thank Laura McDonald, Paul McLean and Eilidh Dear, who were medical students at the University of Dundee, for their help with database searches and with sifting papers for relevance and rigour. We would also like to thank Chau Khuong, Australian Regenerative Medicine Institute, Monash University, Melbourne, Victoria, Australia, for her work in designing Figs Figs1 1 and and3 3 .
Environmental Science and Pollution Research (ESPR) serves the international community in all broad areas of environmental science and related subjects with emphasis on chemical compounds.
Volume 31, Issue 37
Cerium-based metal–organic-frameworks with ligand tuning of the microstructures for fluoride adsorption: linear and nonlinear kinetic and isotherm adsorption models.
Meet the editors.
Learn more about the ESPR editors!
We invite you to browse through recently published special issues of Environmental Science and Pollution Research: Click here to get an overview of ESPR issues! All manuscripts will be peer reviewed as usual and following the journal's policies , with final decisions made by the Editor-in-Chief.
Journal information.
Rights and permissions
Editorial policies
© Springer-Verlag GmbH Germany, part of Springer Nature
Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Scientific Reports volume 14 , Article number: 18964 ( 2024 ) Cite this article
335 Accesses
1 Altmetric
Metrics details
Accurately and quickly estimating the soil organic carbon (SOC) content is crucial in the monitoring of global carbon. Environmental variables play a significant role in improving the accuracy of the SOC content estimation model. This study focuses on modeling methodologies and environmental variables, which significantly influence the SOC content estimation model. The modeling methods used in this research comprise multiple linear regression (MLR), partial least squares regression (PLSR), random forest, and support vector machines (SVM). The analyzed environmental variables include terrain, climate, soil, and vegetation cover factors. The original spectral reflectance (OSR) of Landsat 5 TM images and the spectral reflectivity after the derivative processing were combined with the above environmental variables to estimate SOC content. The results showed that: (1) The SOC content can be efficiently estimated using the OSR of Landsat 5 TM, however, the derived processing method cannot significantly improve the estimation accuracy. (2) Environmental variables can effectively improve the accuracy of SOC content estimation, with climate and soil factors producing the most significant improvements. (3) Machine learning modeling methods provide better estimation accuracy than MLR and PLSR, especially the SVM model which has the highest accuracy. According to our observations, the best estimation model in the study area was the “OSR + SVM” model (R 2 = 0.9590, RMSE = 13.9887, MAE = 10.8075), which considered four environmental factors. This study highlights the significance of environmental variables in monitoring SOC content, offering insights for more precise future SOC assessments. It also provides crucial data support for soil health monitoring and sustainable agricultural development in the study area.
Introduction.
Soil is the largest reservoir of organic carbon in terrestrial ecosystems 1 , 2 , storing twice the amount of organic carbon in the atmosphere and three times as much as that in vegetation 3 , 4 , 5 . Therefore, soil organic carbon (SOC) is an important component of the global carbon cycle 6 and is a fundamental basis for investigating atmospheric CO 2 content and climate change. Moreover, SOC can effectively improve soil fertility by sequestering carbon and participating in soil nutrient cycling 7 , 8 . It improved soil nutrient and water retention capacity, alongside promoting soil respiration, thereby slowing down land degradation and ensuring food production security 9 , 10 , 11 . As a result, accurate and rapid monitoring of SOC content is essential for global carbon pool monitoring, climate change research, and national agricultural management policy formulation.
Traditional methods for estimating SOC content rely on field sampling and laboratory physicochemical analyses. The results of these methods are accurate and reliable, although the process of collecting sampling points is time-consuming and labor-intensive. Besides, it can cause some damage to the surface of the area where the points were collected 12 , 13 . Furthermore, it is economically expensive and difficult to monitor the spatial distribution of SOC over a large region 14 . Therefore, the key issue to be addressed is how to estimate the SOC content of larger regions in a non-destructive, rapid, and accurate. Recently, the development of remote sensing has provided new methods for the non-destructive monitoring of SOC content in a larger region. Many scholars have constructed simple linear estimation models for the SOC content using spectral bands or their combinations based on the spectral response characteristics of soils with different SOC contents. 13 , 15 , 16 . For example, Zhou et al. 17 compared the differences in SOC response between different optical and radar sensors and used their response characteristics to predict SOC content in Spain based on RF. Allory et al. 15 discussed the differences between soil spectral profiles of urban and rural areas and predicted SOC stocks for two cities in France. However, these methods mostly use the original spectral reflectance of each band and their combinations as variables, and their prediction accuracy is not high despite their simple calculation process and strong interpretability. Therefore, it is important to research models that incorporate environmental factors (e.g., climate, terrain, soil, and vegetation cover) to improve the accuracy of SOC content estimation models based on optical remote sensing.
Many studies have demonstrated that SOC exhibits significant spatial heterogeneity 18 , 19 and its content is strongly influenced by environmental variables, including climate, terrain, and soil, particularly precipitation 8 , 20 , 21 , 22 , 23 . For example, Luo et al. 24 constructed a prediction model for the soil organic matter content in the Songnen Plain of China based on random forest and found that the model prediction accuracy was significantly affected by precipitation. Meanwhile, Luo et al. 25 examined the best variable for forecasting the soil organic matter during the bare soil period in Northeast China and concluded that climatic factors can effectively improve the predictive accuracy. These studies showed that incorporating environmental variables as auxiliary variables to the SOC content estimation model, especially those that have a more significant impact on SOC changes, can improve the accurate representation of the spatial distribution of SOC, and it is an important approach to enhance the current estimation model. However, adding more variables can easily lead to problems such as multicollinearity, the number of variables is larger than the samples, and it requires a high level of predictive model arithmetic power 26 . Among the modeling methods for estimating SOC content, PLSR resolves the problem of variable multicollinearity and works with a smaller sample size than MLR, while machine learning regression models like RF and SVM provide great computational ability for complex models 27 , 28 , 29 , 30 , 31 . At the same time, most of the current studies primarily concentrate on the impact of individual environmental factors or modeling methods on the estimation accuracy of SOC content, and there is a lack of comparing the effects of various environmental variables on the estimation accuracy using different methods. Therefore, it is crucial to ascertain the optimal environmental variables and modeling methods for models estimating the SOC content.
The terrestrial transect has already become an important approach for the International Geosphere-Biosphere Programme (IGBP) to study global change. Its aim is to understand the response of terrestrial ecosystems to global change, and to predict and evaluate the potential impacts of global change on terrestrial ecosystems 32 . The Northeast China Transect (NECT) was one of the first IGBP set up for studying global change study in 1993. It is a gradient affected mainly by precipitation 32 . The NECT follows an evident precipitation gradient from east to west, creating various vegetation types and soil characteristics that reflect the most significant and critical climate changes in the mid-latitude temperate zone of East Asia. Therefore, non-destructive monitoring of SOC content in NECT can effectively demonstrate the changes to environmental variables that affect SOC content. This is crucial for the rapid and effective estimation of SOC content over a large region.
Derivative processing is an effective method for eliminating information interference, yet it is rarely applied to Landsat 5 image. In this study, we focused on using Landsat 5 images processed in various ways as the primary data, with environmental variables serving as auxiliary factors, to develop a model for rapidly estimating soil organic carbon content in the Northeast Transect in 2001. We hypothesized that Landsat 5 images could indirectly estimate SOC content, and that derivative processing could effectively remove background interference. However, we also posited that second-order derivation might inhibit the expression of information regarding SOC content, potentially reducing the accuracy of the model's estimation. The specific objectives of this study are: (1) To compare the effects of different environmental variables on the accuracy of the SOC estimation model. (2) To evaluate the SOC estimation capability of Landsat 5 TM images before and after derivative processing. (3) To analyze the differences in optimal variables for constructing SOC estimation models using MLP, PLSR, RF, and SVM methods. (4) To identify the best variables and methods for SOC estimation and map the spatial distribution of SOC content in the study area.
The Northeast China Transect (NECT) (112°–130° 30′ E), with a total length of about 1600 km, is located in the mid-latitude semiarid region along 43° 30′ N (Fig. 1 ) and serves as a land sample zone for global change research wherein precipitation is the main driver. The NECT represents a gradual transition from an oceanic humid climate to a continental arid climate, extending from east to west, and characterized by three climatic zones, namely temperate humid zone, temperate humid semiarid zone, and temperate semiarid and arid zone. Meanwhile, the moisture gradient in the study area is extremely decreased from east to west. In 2001, the annual precipitation in the Changbai Mountains in the east was as high as 800 mm, while that in the central agricultural area was about 550 mm, that in the steppe zone in the center-west ranged from 350 to 500 mm, and that in the western desert steppe was less than 200 mm ( https://data.cma.cn/ ). The average annual temperature ranged from − 3.86 to 7.28 °C in the region ( https://data.cma.cn/ ), and the differences in heat across the region are mainly influenced by the topography with the central plains having the highest average annual temperature and the mountainous or plateau areas at the east and west ends being cooler. Besides, the land use pattern and intensity in the NECT exhibit distinct spatial variations, and there is a complete sequence and transition of forest-agriculture-pasture areas from east to west.
Climatic and geographic of the study area and spatial distribution of sampling points.
The field trip and soil samples collection took place from July 26 to August 8, 2001. The route began from the Chunhua Forest Farm in Hunchun, Jilin, China, and proceeded westward along 43° 30′ N. During this route, a total of 25 soil samples were collected. Their distribution of sampling points is shown in Fig. 1 . In the process of sampling, one soil sample was collected using the five-point sampling method at a depth of 0–20 cm within 30 × 30 m. Each soil sample weighed about 1 kg. At the same time, to locate the sampling points and determine their corresponding positions in Landsat 5 TM images, the coordinates of the sampling points were established using the Magellan GPS Field PROVTM (California, U.S.A.), which is a Global Positioning System (GPS).
In addition, during the collection of soil samples, we removed coarse stones, plant fragments, and roots, placed them in self-sealing bags, and brought them back to the laboratory for analysis. The soil samples were air-dried, ground, filtered (2 mm), and analyzed indoors. Afterward, the SOC content was determined using external heating and the potassium dichromate volumetric method 33 , while soil pH was determined using a pH meter with a soil–water ratio of 1:5.
28 Landsat 5 TM images covering the study area were collected and screened from the Geospatial Data Cloud website ( http://www.gscloud.cn/ ) based on the sampling time and cloud conditions of the soil sample sites in this study. The images were pre-processed using ENVI (The Environment for Visualizing Images, Version 5.3, http://www.harrisgeospatial.com/ ) software 34 , specifically radiometric calibration, atmospheric correction, image stitching, and clipping. During the radiometric calibration process, the thermal infrared bands were not included in the calculations due to their different resolutions. Following radiometric calibration, only six bands of images remain (bands 1–5 and 7), all of which have a spatial resolution of 30 m 35 . Moreover, the preprocessed original spectral reflectance was subjected to first-order derivation (FD) and second-order derivation (SD) through ENVI 5.3 software.
Surface reflectance results from the interaction of various surface factors 13 . The SOC content estimation model based on spectral reflectance lacks the influence of other surface factors. Consequently, its prediction fails to reflect actual scenarios, leading to significant prediction errors. In order to make the estimation results better match actual conditions, this study considered multiple environmental factors, such as climatic conditions, topographic differences, soil physicochemical properties, and vegetation cover on soil spectral reflectance.
Average annual precipitation and average annual temperature were chosen to characterise the climate of the study area. The two variables have a crucial contribution to the input and decomposition of SOC, and they significantly affect SOC storage 25 , 31 . The climate factors in this study were obtained from the National Earth System Science Data Centre ( http://www.geodata.cn/ ), and its spatial distribution in the study area was mapped by ArcGIS (Geographic Information System, Version 10.2, http://www.esri.com/software/arcgis ) software 36 , with a sample resolution of up to 30 m using the resampling method (Fig. 2 a,b).
The environmental variables selected for this study, 2001. [( a ) is the spatial distribution of average annual precipitation. ( b ) shows average annual temperature. ( c ) shows the spatial distribution of elevation. ( d ) shows the spatial distribution of slope. ( e ) is the spatial distribution of slope orientation. ( f ) shows soil pH. ( g ) is the spatial distribution of NDVI. While ( h ) shows EVI)].
Terrain factors have a significant impact on hydrological and ecological mechanisms, like surface runoff, plant growth, and distribution, which in turn directly impacts the SOC content 37 , 38 . Among them, elevation, slope, and slope orientation are frequently utilized to indicate the topographical condition of a region. Therefore, these factors were selected to accurately represent the topographic features of the study area. The terrain factors of the study area were obtained by using the spatial analysis tool of ENVI 5.3 software to extract data from ASTERGDEM ( http://earthexplorer.usgs.gov/ ), and the results are shown in Fig. 2 c,e.
The SOC content is affected at the local scale by soil physicochemical properties 37 . Soil pH is one of these properties and influences the rate of the SOC decomposition process by shaping the behavior of soil microorganisms, thus modifying its speed 39 . In acidic soils, microbial species and numbers are constrained, leading to a slower SOC decomposition rate, particularly when the pH is out of the optimal range, that is, below 5.5 or above 8.5, since these values are highly detrimental to microbial growth and activity 40 , 41 . The spatial distribution of soil pH in the study area was mapped using the Inverse Distance Weight interpolation (IDW) method of ArcGIS 10.2 software that was based on the sampling points. The spatial resolution was sampled up to 30 m using the resampling method. The processed spatial distribution of soil pH is shown in Fig. 2 f.
The Normalised Difference Vegetation Index (NDVI) is the most commonly used vegetation index to reflect vegetation cover. The concept of ratio calculation can significantly reduce noise, which effectively represents the plant growth status and the degree of vegetation cover 42 . The greater the plant cover, the more abundant the apoplastic material and leaves, while the residual roots and leaves remain in the soil to be decomposed by microorganisms, thereby increasing the SOC content 43 . The Enhanced Vegetation Index (EVI) is an improvement of NDVI by adding blue light to correct for atmospheric aerosol scattering and soil background. It reduces the effects of background noise and stably characterises the vegetation condition in high biomass areas 44 . We calculated the NDVI and EVI of the study area using ENVI 5.3 software, based on the preprocessed Landsat 5 TM images (Fig. 2 g,h). The specific calculation formulas are described below.
where Band1, Band3, and Band4 represent the original reflectance of the corresponding bands in the preprocessed Landsat 5 TM images, respectively.
The multiple linear regression model is widely used in remote sensing inversion studies. In this study, we constructed the SOC content estimation model by employing MLR and PLSR models. Additionally, to address the saturation issue in non-linear fitting, we employed random forest and support vector machines models. The details are presented below.
Multiple linear regression (MLR): MLR is a traditional method for exploring the relationship between dependent and independent variables while effectively determining the influence of multiple independent variables on the dependent variable. In particular, by entering distinct arrangements of independent variables, examining and discerning the deviation between the anticipated and actual values of the dependent variables, and choosing the combination of independent variables with the least prognostication error, the most outstanding combination of regression independent variables under the approach is determined. In this study, the MLR modeling process was performed based on the “stats” package of R 4.2.3 45 .
Partial least squares regression (PLSR): PLSR is a commonly used multivariate linear regression modeling method that analyzes the interdependence of predictor and response variables 46 . PLSR concentrates on principal component analysis, linear regression analysis, and typical correlation analysis, and is suitable for studies with multiple sets of predictor variables and small sample sizes. In this study, the parameter selection and modeling process of PLSR were performed based on the “pls” package of R 4.2.3 45 , 47 .
Random forest (RF): RF is a machine learning model in which the set of predictor variables is randomly constrained in each split point, which effectively avoids the problem of tree-to-tree correlation in the computation, and thus achieves highly accurate predictions 11 , 48 . Compared with traditional decision tree construction methods, RF can effectively deal with the problem of insensitivity to the multicollinearity of predictor variables. Furthermore, the ease of parameter tuning is another important advantage of RF, in which the number of spanning trees (ntree), the minimum number of leaves (nodesize), and the number of predictor variables used to split the nodes at each node (mtry) are the key parameters to improve the accuracy of the model. In this study, the estimated model ntree was taken as 500, nodesize was set to 5, and mtry was 3. The parameter selection and modeling process were carried out using the “randomForest” package of R 4.2.3 45 , 49 .
Support Vector Machines (SVM): In the early days, SVM analysis was mainly applied to classification problems. The algorithm is flexible and efficient by inputting data from multiple variables and creating multiple planes to classify the variables into different categories 50 . Recently, many scholars demonstrated that SVM exhibits good modeling stability and high prediction accuracy 28 , 51 , 52 . Therefore, this study constructed the SOC estimation model based on the “e1071” package of R 4.2.3 45 , 53 .
To assess the accuracy of the model, we used the model decision coefficients (R 2 ), root mean square error (RMSE), and mean absolute error (MAE). The formulas for R 2 , RMSE, and MAE are described in Eqs. ( 3 – 5 ). Among these, the range of R 2 is between 0 and 1, and a higher value of R 2 indicates a stronger correlation between the measured and the predicted values. In contrast, RMSE and MAE are both in the range of [0, + ∞]. The lower the values of RMSE and MAE, the closer the measured values are to the predicted values. On the other hand, larger values of RMSE and MAE indicate higher prediction errors. In this study, the three indicators were calculated based on the “modelr” package of R 4.2.3 45 , 54 .
where y i and \({\text{y}}_{{\text{i}}}^{\prime }\) represent the measured and predicted SOC content, respectively, \({\bar{\text{y}}}\) is the average of the measured SOC content, and n is the number of samples.
SOC primarily originates from animal, plant, and microbial residues and their secretions, and undergoes a constant process of decomposition and formation. The SOC content represents the current dynamic equilibrium state of the ecosystem to a certain extent 55 , 56 , 57 . According to Table 2 , the overall SOC content was high in the study area, attaining a mean value of about 22.90 g kg −1 . However, The spatial variation of the SOC content was considerably different based on various environmental factors such as climate, vegetation, and soil, ranging from 2.06 to 107.04 g kg −1 . The standard deviation of 27.04 g kg −1 and the coefficient of variation (CV) was 118.10%. The regions towards the east of the study area were identified by dense vegetation, high accumulation of apomixis, and gradual breakdown of SOC, leading to an increase in SOC content. Meanwhile, the regions towards the west had reduced SOC content, which could be due to the hastened rate of SOC decomposition caused by the climatic circumstances (Fig. 3 ). In addition, 25 samples were randomly divided into a training dataset and a validation dataset in this study, with the former including 20 samples and the latter including 5 samples. The statistical characteristics of both datasets were similar (Table 2 ). It was therefore feasible to apply the estimation model in this study.
Distribution of SOC content in soil sample sites collected in the study area.
The spatial distribution of SOC content was impacted by environmental variables, however, the effects varied among different classes of environmental variables. Table 1 shows that the terrain in the study area significantly varies. Specifically, the sampling points had an elevation range of 120.00–1580.00 m, with a mean of 722.56 m and a standard deviation of 410.70 m. The slope ranged from 1.01 to 43.39°, with a CV of 104.76%, while the standard deviation of the slope orientation was 107.16%. Furthermore, the NDVI ranged from 0.08 to 0.88, with a mean value of 0.46 and CV of 59.31%, while EVI ranged from 0.13 to 1.00 with a mean value of 0.67 and CV of 47.54%. These observations indicate that the vegetation cover in the study area was high and unevenly distributed. Climatic resources varied significantly, especially the variation of average annual precipitation, ranging from 101.04 to 906.59 mm, with an average of 388.48 mm and CV of 67.34%. The average temperature was about 3.67 °C, with a CV of 51.60%.
Defining the spectral reflectance characteristics of SOC is an important basis for estimating SOC content. Figure 4 shows the spectral reflectance characteristics of soils at five different SOC levels. In OSR, the spectral characteristics of different SOC contents showed significant variance, with spectral differences mainly occurring in Band1-3, 5, and 7, especially in the lower SOC levels. In the lower level range, the increase in SOC contents led to a significant decrease in reflection across the five bands, consistent with previous research 16 , 58 , 59 . With higher SOC content, the spectral curve differences primarily occurred in Band4, with fewer variations in the other bands. In FD, the spectral differences of lower SOC contents mainly occurred in Band4, Band5, and Band7, while the differences of spectral curves in the bands of higher SOC contents were small. In addition, the differences in different SOC contents are mainly observed in Band7 in SD.
Spectral characteristics of Landsat 5 TM with different SOC contents.
Furthermore, association analysis was used to investigate the relationship between SOC content and spectral reflectance. As shown in Table 2 , except for Band1, which was processed by the first-order derivation and had a weak correlation with SOC content, all the other bands had high correlations. Specifically, Band1-3, Band5, and Band7 showed a significant negative correlation with SOC content in OSR, especially Band7 with a correlation coefficient as high as − 0.523, while Band4 was significantly positively correlated with SOC content (0.409). After the first-order derivation transformation, a significant increase in the correlation between the SOC content and Band2, Band3, and Band5 was noted. Specifically, Band3 depicted a significant positive correlation (0.594), whereas Band5 showed a highly significant negative correlation (− 0.625). After the second-order derivation transformation, a significant decrease was observed in the correlation of Band5 with SOC content. Moreover, the correlation between SOC content and the remaining bands was comparatively weaker, especially Band5. In summary, although there was a high correlation between reflectance and SOC content in the original bands, the first-order derivation improved this correlation to some extent. However, the second-order derivation did not demonstrate this ability, which may be related to the lack of information caused by the amplified noise effect after the second-order derivation.
Derivative processing.
The training and validation results of the 72 estimation models for SOC content are shown in Table 3 . The results showed that derivative processing did not significantly improve the estimation accuracy. Moreover, the second-order derivative led to a significant decrease in the estimation ability of the model. In particular, the derivative processing reduced the estimation accuracy for the model constructed using a training dataset. Taking the model without environmental variables as an example, the accuracy of “FD + MLR” decreased by 0.1395 compared to “OSR + MLR”. Also, the accuracy of “SD + MLR” decreased by 0.1708 compared to “OSR + MLR”. Similarly, machine learning methods exhibited the same trend, with “FD + Band + SVM” decreasing the accuracy by 0.0069 compared to “OSR + Band + SVM”, and “SD + Band + SVM” decreasing the estimation accuracy by 0.0828 compared to “OSR + Band + SVM”. Moreover, environmental variables alleviated the negative influence of derivative processing to some extent, while this phenomenon still existed. Consider the model with four environmental variables, where the accuracy of “FD + EV + SVM” decreased by 0.0328 compared to “OSR + EV + SVM”, while “SD + EV + SVM” was 0.0112 less accurate than “OSR + EV + SVM”. It is worth mentioning that terrain factors and derivative processing improved the estimation accuracy of the MLR method. Regarding the “MLR + TF” context, derivative processing improved by 0.0204 (FD) and 0.0882 (SD) respectively.
The modeling accuracy was significantly higher for the model which takes environmental variables into account in comparison to the spectral model that only considers spectral bands. The addition of VF, CF, TF, and SF individually can improve the accuracy of the model, especially of CF and SF (Table 3 ). Specifically, the estimation models constructed using the training dataset exhibited better simulation accuracies under various derivative treatments when they incorporated CF, TF, and SF independently. For example, under different derivative processing, the SVM method improved the model with only CF added by 0.1499 (OSR), 0.0839 (FD), and 0.1854 (SD) respectively, when compared to the model with no environmental variables added. In contrast, the accuracy of the model with only TF added improved by 0.1451 (OSR), 0.1305 (FD), and 0.1829 (SD) respectively. By adding only SF, the model accuracy improved by 0.1504 (OSR), 0.0846 (FD), and 0.1888 (SD) respectively. Furthermore, although adding only VF improved the model accuracy, it had less influence compared to the three other types of variables. For example, in OSR, under different modeling methods, adding only VF improved the model accuracy by 0.1202 (PLSR), 0.0211 (RF), and 0.0054 (SVM) compared to the model with no environmental variables, respectively. In addition, there was no significant difference in the accuracy precision of the MLSR method. It is worth mentioning that adding only the TF in the model constructed based on PLSR led to a significant decrease of 0.6123 (OSR), 0.3174 (FD), and 0.3364 (SD) in the accuracy of model estimation in the context of derivative processing. However, this phenomenon was not present in the other modeling methods. Furthermore, adding environmental variables, especially VF and CF, significantly improves the accuracy of the model constructed from the validation dataset, compared to the model without these variables. However, including only TF in “OSR + PLSR” lowers its accuracy (R 2 = 0.0812), which is notably inferior to that of the other models.
Although these environmental variables impacted the estimation model differently, adding them at the same time can significantly improve its fitting accuracy. Taking the model constructed based on the training dataset as an example, the accuracy of “OSR + EV + SVM” was 0.1702 higher than that of “OSR + Band + SVM”, 0.1756 higher than that of “OSR + VF + SVM”, 0.0203 higher than that of “OSR + CF + SVM”, 0.0250 higher than that of “OSR + TF + SVM”, and 0.0198 higher than that of “OSR + SF + SVM”. In summary, the combination of VF, CF, TF, and SF can significantly enhance SOC estimation accuracy within the confines of the study area. This was especially evident when the study covered a vast region and displayed substantial spatial variations in diverse environmental variables.
Comparing the four modeling methods, it can be found that the conclusions of the models constructed based on the training and validation datasets were consistent. The results showed that under different derivative processing, both RF and SVM methods have higher accuracy in fitting and estimating SOC content compared with traditional regression analysis methods (MLR and PLSR). Specifically, the model based on the SVM method, with the original spectral bands and the four types of environmental variables achieved the highest accuracy (Training dataset: R 2 = 0.9590, RMSE = 13.9887, MAE = 10.8075; Validation dataset: R 2 = 0.9220, RMSE = 11.6165, MAE = 10.8075), followed by the “OSR + EV + RF” model. Specifically, considering the model developed based on the training dataset under identical derivative processing conditions, the accuracy of the “OSR + EV + SVM” model was 0.0856 greater than that of the “OSR + EV + RF” model. It exceeds the “OSR + EV + MLR” model by 0.1050 (RMSE = 11.0773, MAE = 9.2049) and the “OSR + EV + PLSR” model (RMSE = 21.6330, MAE = 15.0098) by 0.5531. Therefore, including environmental variables can effectively improve SOC accuracy, and the SVM method presents the best estimation accuracy, while the PLSR method is unsuitable for the study area. Besides, this study plotted the plots of the estimated and measured values of the sample points based on the SVM method (Fig. 5 ). The above results showed that including environmental variables significantly improved the estimation of SOC content, and the predicted values were in good agreement with measured values. Moreover, the “OSR + EV + SVM” model (R 2 = 0.9387) showed the best fit between the predicted and measured values (Fig. 5 d). These results confirm that the “OSR + EV + SVM” model was the optimal SOC estimation model for the study area.
Comparison of measured and predicted SOC content among different SVM models.
In general, environmental variables significantly increased the accuracy of the SOC content estimation model, and the best model was the “OSR + SVM” model combined with environmental variables, which had the best estimation performance. Derivative processing did not yield better estimation results. Among them, the effects of FD and OSR processing were similar, while the prediction effect of SD was weaker and significantly worse than that of the case without derivative processing. As for the choice of modeling methods, machine learning prediction has a higher accuracy than traditional methods. Among them, the SVM model has the most prediction ability, followed by RF, while the PLSR model is not suitable for fitting and predicting.
Since the “OSR + EV + SVM” model had a higher prediction accuracy than other models, we used this model to estimate the SOC content in the study area (Fig. 6 ). Spatially, the SOC content in the study area in 2001 was highly spatially variable, with a mean value of 31.74 g kg −1 and a standard deviation of 28.31 g kg −1 . The central and western parts of the NECT had low SOC content while the eastern part had high SOC content. This trend was consistent with environmental variables, particularly elevation, slope, and precipitation. Statistical analysis of the SOC content indicated high SOC content in the entire study area. About 57.50% of the area had a SOC content higher than 20.00 g kg −1 . 23.7% of the area had SOC content higher than 50.00 g kg −1 , and 33.80% of the covered area had a SOC content between 20.00 and 50.00 g kg −1 . In addition, areas with SOC content above 100.00 g kg −1 and below 5.00 g kg −1 accounted for 2.30% and 19.90% of the total area, respectively.
The SOC content in the study area based on the best model, 2001.
Accurate quantification of soil organic carbon content is crucial for soil safety assessment, carbon cycling, and climate change research. Currently, constructing models to estimate SOC content based on optical remote sensing has become a pivotal research approach. However, the choice of different image processing methods, modeling techniques, and selection of environmental variables significantly influences model accuracy. Derivative processing aids in smoothing background information and reducing noise interference 27 , 29 , 30 , widely applied in hyperspectral studies. We hypothesized that integrating Landsat 5 with environmental factors can effectively estimate SOC content, while derivative processing suppresses information expression in Landsat 5 images. Model comparison results validate our hypothesis that Landsat 5 images effectively convey SOC information, with derivative processing showing no significant improvement in SOC content estimation accuracy, consistent with previous studies 13 , 60 . This may be attributed to the large coverage area, high variability in soil physicochemical properties, significant climate differences, and complex terrain and vegetation cover variations in the study area. Regarding modeling methods, studies by Bao et al. 61 and Sabetizade et al. 62 demonstrate that machine learning methods better handle the non-linear relationship between SOC content and spectral features, enhancing prediction reliability and predictability. Our findings support these conclusions. Additionally, our results indicate that Random Forest (RF) and Support Vector Machine (SVM) exhibit significant modeling advantages in estimating SOC content based on small sample datasets, particularly with SVM demonstrating stable modeling capabilities.
Our study reveals that adding environmental factors can effectively enhance SOC content estimation accuracy. Among these factors, mean annual precipitation and soil pH are the primary determinants influencing SOC content estimation accuracy, followed by elevation, slope, and aspect. This aligns with existing research 4 , 28 , 31 . Temperature influences SOC content significantly through its effects on soil microbial activity and soil respiration 63 , 64 . Soil pH affects soil chemical reactions and ion exchange properties, with acidic soils typically limiting microbial activity and organic matter decomposition, thereby inhibiting SOC accumulation 65 . Precipitation indirectly affects SOC content by influencing soil moisture status. These findings confirm the significant impact of environmental factors on the spatial distribution of SOC content, with varied effects observed among different environmental variables.
In our study, integrating soil organic carbon data with Landsat 5 elucidates the influence of environmental variables on SOC content estimation. While our conclusions are insightful, they come with limitations. Firstly, the factors influencing spatial distribution of soil organic carbon content are complex, and this study only discusses common environmental variables. Future research will incorporate additional factors influencing spatial distribution of soil organic carbon content and analyze the differential importance of various environmental variables. For instance, soil clay particles adsorb and protect soil organic matter, preventing microbial decomposition or hydraulic erosion, often correlating with higher soil organic carbon content, a factor worth considering in future studies. Secondly, anthropogenic management practices also significantly influence soil organic carbon content. Subsequent studies should adequately characterize factors such as land use types and management practices (e.g., land use intensity) to enhance SOC content estimation capability. Lastly, our study covers a broad area with diverse land cover types, necessitating the collection of more sample points and the use of machine learning algorithms and feature selection suitable for big data to further analyze the impact of different environmental factors on soil organic carbon content across different regions, thereby improving SOC mapping. In the next step, we will integrate the latest satellite data with the results of a new round of large-scale sampling. This approach will enable us to include additional environmental variables and spectral bands for SOC mapping, thereby refining our methods and enhancing the accuracy of SOC content detection.
Through comparative analysis of derivative processing and four different modeling methods, our study generated soil organic carbon content maps for the surface layer of the Northeast Transect at a spatial resolution of 30 m. Results demonstrate that adding environmental variables enhances the accuracy of SOC content estimation across various models. Concerning derivative processing, the performance of OR and FD modeling methods showed minimal difference, while SD modeling exhibited poorer capabilities. In terms of modeling method selection, SVM models constructed with the inclusion of four categories of environmental factors consistently displayed strong predictive abilities, whereas Partial Least Squares Regression (PLSR) performed less effectively. Therefore, “OR + environmental variables + SVM” represents the optimal model for estimating SOC content in the study area (R 2 = 0.9590, RMSE = 13.9887, MAE = 10.8075). Our research highlights the importance of key environmental variables in monitoring SOC content. This study provides theoretical reference for the rapid estimation of large-area soil organic carbon content using remote sensing images combined with environmental factors. It also offers technical support for more precise SOC content detection in the future, which is crucial for soil conservation and sustainable development in Northeast China.
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Lal, R. Soil carbon sequestration to mitigate climate change. Geoderma 123 (1–2), 1–22. https://doi.org/10.1016/j.geoderma.2004.01.032 (2004).
Article ADS CAS Google Scholar
Zeraatpisheh, M. et al. Improving the spatial prediction of soil organic carbon using environmental covariates selection: A comparison of a group of environmental covariates. CATENA 208 , 105723. https://doi.org/10.1016/j.catena.2021.105723 (2022).
Article CAS Google Scholar
Liang, Z. et al. National digital soil map of organic matter in topsoil and its associated uncertainty in 1980’s China. Geoderma 335 , 47–56. https://doi.org/10.1016/j.geoderma.2018.08.011 (2019).
Article ADS Google Scholar
Zhang, H. et al. Prediction of soil organic carbon in an intensively managed reclamation zone of eastern China: A comparison ofmultiple linear regressions and the random forest model. Sci. Tot. Environ. 592 , 704–713. https://doi.org/10.1016/j.scitotenv.2017.02.146 (2017).
Zhang, S., Tian, J., Lu, X. & Tian, Q. Temporal and spatial dynamics distribution of organic carbon content of surface soil in coastal wetlands of Yancheng, China from 2000 to 2022 based on Landsat images. CATENA 223 , 106961. https://doi.org/10.1016/j.catena.2023.106961 (2023).
Yang, R., Huang, L., Zhang, X., Zhu, C. & Xu, L. Mapping the distribution, trends, and drivers of soil organic carbon in China from 1982 to 2019. Geoderma 429 , 116232. https://doi.org/10.1016/j.geoderma.2022.116232 (2023).
Venter, Z. S., Hawkins, H. J., Cramer, M. D. & Mills, A. J. Mapping soil organic carbon stocks and trends with satellite-driven high resolution maps over South Africa. Sci. Tot. Environ. 771 , 145384. https://doi.org/10.1016/j.scitotenv.2021.145384 (2021).
Wang, X. et al. Remote estimates of soil organic carbon using multi-temporal synthetic images and the probability hybrid model. Geoderma 425 , 116066. https://doi.org/10.1016/j.geoderma.2022.116066 (2022).
Li, Z., Liu, F., Peng, X., Hu, B. & Song, X. Synergetic use of DEM derivatives, Sentinel-1 and Sentinel-2 data for mapping soil properties of a sloped cropland based on a two-step ensemble learning method. Sci. Tot. Environ. 866 , 161421. https://doi.org/10.1016/j.scitotenv.2023.161421 (2023).
Lu, M. Y., Liu, Y. & Liu, G. J. Precise prediction of soil organic matter in soils planted with a variety of crops through hybrid methods. Comput. Electron. Agric. 200 , 107246. https://doi.org/10.1016/j.compag.2022.107246 (2022).
Article Google Scholar
Nguyen, T. T. et al. A novel intelligence approach based active and ensemble learning for agricultural soil organic carbon prediction using multispectral and SAR data fusion. Sci. Tot. Environ. 804 , 150187. https://doi.org/10.1016/j.scitotenv.2021.150187 (2022).
Florinsky, I. V., Eilers, R. G., Manning, G. R. & Fuller, L. G. Prediction of soil properties by digital terrain modeling. Environ. Environ. Model. Softw. 17 (3), 295–311. https://doi.org/10.1016/S1364-8152(01)00067-6 (2002).
Yu, Q. et al. Improving estimation of soil organic matter content by combining Landsat 8 OLI images and environmental data: A case study in the river valley of the southern Qinghai-Tibet Plateau. Comput. Electron. Agric. 185 , 106144. https://doi.org/10.1016/j.compag.2021.106144 (2021).
Chen, D., Chang, N., Xiao, J., Zhou, Q. & Wu, W. Mapping dynamics of soil organic matter in croplands with MODIS data and machine learning algorithms. Sci. Tot. Environ. 669 , 844–855. https://doi.org/10.1016/j.scitotenv.2019.03.151 (2019).
Allory, V. et al. Quantification of soil organic carbon stock in urban soils using visible and near infrared reflectance spectroscopy (VNIRS) in situ or in laboratory conditions. Sci. Total Environ. 686 , 764–773. https://doi.org/10.1016/j.scitotenv.2019.05.192 (2019).
Article ADS CAS PubMed Google Scholar
Tiruneh, G. A. et al. Geospatial modeling and mapping of soil organic carbon and texture from spectroradiometric data in Nile basin. Remote Sens. Appl. Soc. Environ. 29 , 100879. https://doi.org/10.1016/j.rsase.2022.100879 (2023).
Zhou, T. et al. Effects of optical and radar satellite observations within Google Earth Engine on soil organic carbon prediction models in Spain. J. Environ. Manag. 338 , 117810. https://doi.org/10.1016/j.jenvman.2023.117810 (2023).
Caires, S. A. D., Keshavarzi, A., Bottega, E. L. & Kaya, F. Towards site-specific management of soil organic carbon: Comparing support vector machine and ordinary kriging approaches based on pedo-geomorphometric factors. Comput. Electron. Agric. 216 , 108545. https://doi.org/10.1016/j.compag.2023.108545 (2024).
Kaya, F. et al. Assessing machine learning-based prediction under different agricultural practices for digital mapping of soil organic carbon and available phosphorus. Agriculture 12 (7), 1062. https://doi.org/10.3390/agriculture12071062 (2022).
Castaldi, F. et al. Evaluating the capability of the Sentinel 2 data for soil organic carbon prediction in croplands. J. Photogramm. Remote Sens. 147 , 267–282. https://doi.org/10.1016/j.isprsjprs.2018.11.026 (2019).
Liu, F. et al. Mapping high resolution national soil information grids of China. Sci. Bull. 67 (3), 328–340. https://doi.org/10.1016/j.scib.2021.10.013 (2022).
Vasenev, V. I., Stoorvogel, J. J., Leemans, R., Valentini, R. & Hajiaghayeva, R. A. Projection of urban expansion and related changes in soil carbon stocks in the Moscow Region. J. Clean. Product. 170 , 902–914. https://doi.org/10.1016/j.jclepro.2017.09.161 (2018).
Zhang, P. et al. Spatial variation and distribution of soil organic carbon in an urban ecosystem from high-density sampling. Catena 204 , 105364. https://doi.org/10.1016/j.catena.2021.105364 (2021).
Luo, C., Wang, Y., Zhang, X., Zhang, W. & Liu, H. Spatial prediction of soil organic matter content using multiyear synthetic images and partitioning algorithms. Catena 211 , 106023. https://doi.org/10.1016/j.catena.2022.106023 (2022).
Luo, C., Zhang, W., Zhang, X. & Liu, H. Mapping soil organic matter content using Sentinel-2 synthetic images at different time intervals in Northeast China. Int. J. Digit. Earth 16 (1), 1094–1107. https://doi.org/10.1080/17538947.2023.2192005 (2023).
Zhang, Z., Ding, J., Wang, J. & Ge, X. Prediction of soil organic matter in northwestern China using fractional-order derivative spectroscopy and modified normalized difference indices. Catena. 185 , 104257. https://doi.org/10.1016/j.catena.2019.104257 (2020).
Hong, Y. et al. Exploring the potential of airborne hyperspectral image for estimating topsoil organic carbon: Effects of fractional-order derivative and optimal band combination algorithm. Geoderma 365 , 114228. https://doi.org/10.1016/j.geoderma.2020.114228 (2020).
Mahmoudzadeh, H., Matinfar, H. R., Mehrjardi, R. T. & Kerry, R. Spatial prediction of soil organic carbon using machine learning techniques in western Iran. Geoderma Regional 21 , e00260. https://doi.org/10.1016/j.geodrs.2020.e00260 (2020).
Wang, Y. et al. A comparison of multiple deep learning methods for predicting soil organic carbon in Southern Xinjiang, China. Comput. Electron. Agric. 212 , 108067. https://doi.org/10.1016/j.compag.2023.108067 (2023).
Wang, Y. et al. Evaluation of data pre-processing and regression models for precise estimation of soil organic carbon using Vis-NIR spectroscopy. J. Soils Sediments 23 , 634–645. https://doi.org/10.1007/s11368-022-03337-2 (2023).
Zhou, T. et al. High-resolution digital mapping of soil organic carbon and soil total nitrogen using DEM derivatives, Sentinel-1 and Sentinel-2 data based on machine learning algorithms. Sci. Tot. Environ. 729 , 138244. https://doi.org/10.1016/j.scitotenv.2020.138244 (2020).
Zhang, X. et al. A gradient analysis and prediction on the northeast China transect (NECT) for global change study. Acta Botanica Sinica. 9 , 785–799+889-890 (1997) ( in Chinese ).
Google Scholar
Nelson, D. W. & Sommers, L. E. A rapid and accurate method for estimating organic carbon in soil. Proc. Indiana Acad. Sci. 84 , 456–462 (1975).
Boulder, C. Exclis Visual Information Solutions (2015). http://www.harrisgeospatial.com
NASA. Landsat 5 Satellite Handbook (1984). http://www.usgs.gov/land-resources/nli/landsat/landsat-5
Esri. ArcGIS Desktop: Release 10.2. Redlands (Environmental Systems Research Institute, 2013). http://www.esri.com
Lamichhane, S., Kumar, L. & Wilson, B. Digital soil mapping algorithms and covariates for soil organic carbon mapping and their implications: A review. Geoderma 352 , 395–413. https://doi.org/10.1016/j.geoderma.2019.05.031 (2019).
Minasny, B., McBratney, A. B., Malone, B. P. & Wheeler, I. Chapter one - Digital mapping of soil carbon. Adv. Agron. 118 , 1–47. https://doi.org/10.1016/B978-0-12-405942-9.00001-3 (2013).
Cao, H. et al. Soil pH, total phosphorus, climate and distance are the major factors influencing microbial activity at a regional spatial scale. Sci. Rep. 6 (1), 25815. https://doi.org/10.1038/srep25815 (2016).
Article ADS CAS PubMed PubMed Central Google Scholar
Bhattacharyya, S. S., Ros, G. H., Furtak, K., Iqbal, H. M. N. & Parra-Saldívar, R. Soil carbon sequestration-An interplay between soil microbial community and soil organic matter dynamics. Sci. Total Environ. 815 , 152928. https://doi.org/10.1016/j.scitotenv.2022.152928 (2022).
Article CAS PubMed Google Scholar
Chen, S. et al. A high-resolution map of soil pH in China made by hybrid modeling of sparse soil data and environmental covariates and its implications for pollution. Sci. Tot. Environ. 655 , 273–283. https://doi.org/10.1016/j.scitotenv.2018.11.230 (2019).
Huete, A. et al. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens. Environ. 83 (1–2), 195–213. https://doi.org/10.1016/S0034-4257(02)00096-2 (2002).
Jobbágy, E. G. & Jackson, R. B. The vertical distribution of soil organic carbon and its relation to climate and vegetation. Ecol. Appl. 10 (2), 423–424. https://doi.org/10.1890/1051-0761(2000)010[0423:TVDOSO]2.0.CO;2 (2000).
Jiang, Z., Huete, A. R., Didan, K. & Miura, T. Development of a two-band enhanced vegetation index without a blue band. Remote Sens. Environ. 112 (10), 3833–3845. https://doi.org/10.1016/j.rse.2008.06.006 (2008).
R Core Team. R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, 2023). https://www.R-project.org/
Khaledian, Y. et al. Modeling soil cation exchange capacity in multiple countries. Catena 158 , 194–200. https://doi.org/10.1016/j.catena.2017.07.002 (2017).
Liland, K. H., Mevik, B. H., Wehrens, R., Hiemstra, P. pls: Partial Least Squares and Principal Component Regression (2023). https://CRAN.R-project.org/package=pls
Breiman, L. Random forests. Mach. Learn. 45 (1), 5–32 (2001).
Liaw, A. & Wiener, M. Classification and regression by random forest. R News 2 (3), 18–22 (2002).
Cortes, C. & Vapnik, V. Support-vector networks. Mach. Learn. 20 , 273–297. https://doi.org/10.1023/A:1022627411411 (1995).
Maynard, J. J. & Levi, M. R. Hyper-temporal remote sensing for digital soil mapping: Characterizing soil-vegetation response to climatic variability. Geoderma 285 , 94–109. https://doi.org/10.1016/j.geoderma.2016.09.024 (2017).
Welikhe, P., Quansah, J. E., Fall, S. & McElhenney, W. Estimation of soil moisture percentage using LANDSAT-based moisture stress index. J. Remote Sens. GIS 6 (1), 1000200. https://doi.org/10.4172/2469-4134.1000200 (2017).
Meyer, D., Dimitriadou, E., Hornik, K., Weingessel, A., Leisch, F., Chang, C.C., Lin, C. C. e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien (2023). https://CRAN.R-project.org/package=e1071
Wickham, H. Software, P. modelr: Modeling Functions that Work with the Pipe (2023). https://CRAN.R-project.org/package=modelr
Martinez, C., Hancock, G. R. & Kalma, J. D. Relationships between 137Cs and soil organic carbon (SOC) in cultivated and never-cultivated soils: An Australian example. Geoderma 158 (3–4), 137–147. https://doi.org/10.1016/j.geoderma.2010.04.019 (2010).
Wang, S. et al. Patterns of SOC and soil 13 C and their relations to climatic factors and soil characteristics on the Qinghai-Tibetan Plateau. Plant Soil 363 , 243–255. https://doi.org/10.1007/s11104-012-1304-6 (2013).
Wei, S. et al. Impact of soil water erosion processes on catchment export of soil aggregates and associated SOC. Geoderma 294 , 63–69. https://doi.org/10.1016/j.geoderma.2017.01.021 (2017).
Dou, X. et al. Prediction of soil organic matter using multi-temporal satellite images in the Songnen Plain, China. Geoderma 356 , 113896. https://doi.org/10.1016/j.geoderma.2019.113896 (2019).
Stoner, E. R. & Baumgardner, M. F. Characteristic variations in reflectance of surface soils. Soil Sci. Soc. Am. J. 45 (6), 1161–1165. https://doi.org/10.2136/sssaj1981.03615995004500060031x (1981).
Kooistra, L., Wehrens, R., Leuven, R. S. E. W. & Buydens, L. M. C. Possibilities of visible-near-infrared spectroscopy for the assessment of soil contamination in river floodplains. Anal. Chim. Acta 446 , 97–105. https://doi.org/10.1016/S0003-2670(01)01265-X (2001).
Bao, N., Liu, S. & Zhou, Y. Predicting particle-size distribution using thermal infrared spectroscopy from reclaimed mine land in the semi-arid grassland of North China. CATENA 183 , 104190. https://doi.org/10.1016/j.catena.2019.104190 (2019).
Sabetizade, M., Gorji, M., Roudier, P., Zolfaghari, A. A. & Keshavarzi, A. Combination of MIR spectroscopy and environmental covariates to predict soil organic carbon in a semi-arid region. Catena 196 , 104844. https://doi.org/10.1016/j.catena.2020.104844 (2021).
Akpa, S. I. C., Odeh, I. O. A., Bishop, T. F. A., Hartemink, A. E. & Amapu, I. Y. Total soil organic carbon and carbon sequestration potential in Nigeria. Geoderma 271 , 202–215. https://doi.org/10.1016/j.geoderma.2016.02.021 (2016).
Moura-Bueno, J. M., Dalmolin, R. S. D., Caten, A. T., Dotto, A. C. & Demattê, J. A. M. Stratification of a local VIS-NIR-SWIR spectral library by homogeneity criteria yields more accurate soil organic carbon predictions. Geoderma 337 , 565–581. https://doi.org/10.1016/j.geoderma.2018.10.015 (2019).
Orgill, S. E. et al. Soil with high organic carbon concentration continues to sequester carbon with increasing carbon inputs. Geoderma 285 , 151–163. https://doi.org/10.1016/j.geoderma.2016.09.033 (2017).
Download references
Acknowledgement for the data support from “National Earth System Science Data Center, National Science & Technology Infrastructure of China ( http://www.geodata.cn )”.
This research was funded by the National Natural Science Foundation of China (No. 42141007).
Authors and affiliations.
College of Resources and Environmental Sciences, China Agricultural University, Beijing, 100193, China
Xiao Xiao, Qijin He, Selimai Ma, Jiahong Liu, Weiwei Sun, Yujing Lin & Rui Yi
Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters (CIC-FEMD), Nanjing University of Information Science and Technology, Nanjing, 210044, China
You can also search for this author in PubMed Google Scholar
X.X.: Formal analysis, Investigation, Methodology, Writing-original draft, Visualization. Q.H.: Conceptualization, Resources, Writing-review and editing, Validation, Supervision, Project administration, Funding acquisition. S.M.: Data collection, Data curation, Validation. J.L.: Software, Writing-review and editing. W.S.: Software. Y.L.: Data collection, Data curation. R.Y.: Data collection, Data curation. All authors have read and agreed to the published version of the manuscript. All the authors approved the final article.
Correspondence to Qijin He .
Competing interests.
The authors declare no competing interests.
Publisher's note.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ .
Reprints and permissions
Cite this article.
Xiao, X., He, Q., Ma, S. et al. Environmental variables improve the accuracy of remote sensing estimation of soil organic carbon content. Sci Rep 14 , 18964 (2024). https://doi.org/10.1038/s41598-024-68424-5
Download citation
Received : 08 March 2024
Accepted : 23 July 2024
Published : 16 August 2024
DOI : https://doi.org/10.1038/s41598-024-68424-5
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.
Provided by the Springer Nature SharedIt content-sharing initiative
By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.
Intended for healthcare professionals
New concordat is necessary but no match for the scale of the challenge
The newly released Concordat for the Environmental Sustainability of Research and Innovation Practice, co-developed by the UK research and innovation community—including universities, research organisations, funders, and their partners—represents a broad ambition for the sector to transition to a sustainable future. 1 Concordat signatories are asked to prioritise action on leadership and system change, sustainable infrastructure, sustainable procurement, emissions from business and academic travel, collaborations and partnerships, and reporting data on the environmental impact of their activities.
We commend the developers of the concordat for encouraging the advancement of environmental sustainability in research and innovation across all sectors, including healthcare, but it does not go far enough given the scale of the challenge. Important limitations include the voluntary nature of the concordat, lack of clarity in reporting standards and guidance, lack of verification, and a limited capacity to address the environmental impact of supply chains.
First, action stronger than voluntary participation is needed to achieve the scale and speed of the changes required. The clock is ticking, and the global carbon budget is dwindling. 2 The concordat emphasises the “need to act now” with measures in the next 5-10 years that include “deep, rapid, and sustained reductions in greenhouse gas emissions” and “actions to address unsustainable resource consumption.” But to really drive change, public and private research funders, including the National Institute for Health and Care Research (NIHR) and Wellcome, should consider making funding contingent on a commitment to the concordat. This would create the strongest possible incentive for all stakeholders.
Second, although the concordat calls on its signatories to publicly disclose their commitments and report progress, there is no verification requirement and the consequences of failure to follow through are unclear. Verification is essential and could be achieved through the EU Corporate Sustainability Reporting Directive (CSRD) or a similar authority. The directive requires large and listed companies, including some independent research organisations, to measure, track, and disclose direct and indirect emissions, along with their efforts to operationalise sustainability. 3 Assurances are assessed through transparent third party verification, to avert greenwashing and reduce the risk of conflicts of interest.
Clear reporting standards are also important for ensuring accountability, but the concordat provides limited guidance on emissions accounting. Consistent with guidance from the Environmental Association for Universities and Colleges, 4 the concordat recommends including direct greenhouse gas emissions, indirect energy emissions, and other indirect emissions that are material to an organisation’s activities. The first two are well defined and quantifiable but usually contribute only a minority (15-35%) of an organisation’s emissions. Other indirect emissions, such as those arising from supply chains, business travel, and waste management, comprise a much larger proportion of an organisation’s carbon footprint but are harder to measure. Accuracy is particularly important when comparing emissions across different organisations and when tracking emissions over time.
Measuring greenhouse gas emissions associated with procurement and supply chains is critical since purchased goods and services often comprise at least 50% of an organisation’s emissions. The concordat encourages life cycle assessment (LCA) where possible, as this is the most accurate way to quantify these emissions. However, LCAs require expertise and resources unavailable to many organisations. Furthermore, each analysis has a specific scope and goals, limiting generalisability, and the quality of existing assessments is heterogeneous.
In response to the mixed quality of existing LCAs, proposed guidelines for assessing the environmental consequences of healthcare (Ecohealth) aim to provide a reporting standard for analyses relevant to healthcare, including research and innovation. 5 This will improve both the quality and the comparability of sustainability reports.
Ultimately, though, industry partners in research and innovation will have to use their own knowledge of materials, production methods, and energy sources to report product level emissions in a standardised and verifiable way that enables downstream organisations to report supply chain emissions accurately. This would be better than the spend based models suggested by the concordat.
Finally, the concordat asks institutions to establish “sustainable procurement policy … that prioritise[s] more environmentally sustainable options” but does not indicate how. Manufacturers make numerous claims about the environmental credentials of their products and services, but without evidence based on standardised product level LCAs and independent verification of claims, purchasers may make incorrect choices based on inaccurate or misleading environmental information.
Consistent with the EU’s CSRD, NHS England is phasing in a requirement for vendors of healthcare products and services to report emissions in a standardised, transparent, verified manner along with decarbonisation plans consistent with the Paris agreement. Product level disclosures will be required by 2028. 6 The concordat could help create collective purchasing power to drive down embodied emissions by directing signatories to require standardised and verified product level environmental disclosures in all their purchasing processes. This would also improve environmental accounting and procurement decisions.
We support the concordat’s vision and aspiration but call for more, to enable the research and innovation sector to truly lead change. This means mandatory, verified reporting of emissions by all stakeholders using accurate, comparable methods to help organisations make better environmental choices.
Competing interests: The BMJ has judged that there are no disqualifying financial ties to commercial companies. The authors declare the following other interests: JES has received consulting fees from Teleflex, AstraZeneca, and AlphaSights; honorariums for speaking on healthcare sustainability from the University of New Mexico, Columbia University, the University of Colorado, Harvard University, and the Institute for Healthcare Improvement; and travel reimbursements to speak on healthcare sustainability from the Canadian Anesthesiologists’ Society, Vizient, University of Colorado, and the Institute for Healthcare Improvement. Further details of The BMJ policy on financial interests are here: https://www.bmj.com/sites/default/files/attachments/resources/2016/03/16-current-bmj-education-coi-form.pdf .
Provenance and peer review: Commissioned; not externally peer reviewed.
2023 Theses Doctoral
Rhodes-Bratton, Brennan
The disproportionate concentration of unhealthy food in communities of color in the United States may contribute to health inequities and food insecurity. Gentrification has been associated with residents’ increased adverse health outcomes in its early and rapid phases. This study adds to the growing body of research by examining the relationship between gentrification, the food environment, food habits (the interplay between food chances and food choices), and health in New York City. I used a mixed methods approach to assess the food landscape in NYC between 1990 and 2014, using group-based trajectory modeling, the National Establishments Time-Series database, census data, and in-depth interviews with mothers from the Columbia Center for Children’s Environmental Health study. I found that the growth in the food environment was unevenly distributed. While healthy food chances declined across all examined neighborhoods, unhealthy food chances quickly grew, commanding dominance. It was gentrifying neighborhoods; however, that surprisingly experienced the most remarkable growth in unhealthy food chances compared to other neighborhoods. A cross-tabulation of the food chance trajectories of New York City census tracts indicated the presence of food ecologies that exhibit both healthy and unhealthy food chances. There was a strong association between the type of food ecology and gentrification status (p < 0.001). The in-depth interviews corroborated these findings and revealed that food insecurity is a by-product of gentrification in two ways. First, neighborhoods in the early stages of gentrification are inundated with unhealthy food chances, such as fast-food chains, without adequate access to quality, fresh, healthy foods. Secondly, when healthy food chances finally arrive in resource-deprived areas through gentrification, families are forced to relocate to areas without access to fresh, affordable, healthy foods due to the increased cost of living. This cycle of food insecurity is inequitable due to historical racial segregation, exploitative capitalistic markets, and racist stereotypes. Speculators invest in unhealthy food chances aligned with pre-existing stereotypes, assumptions, and beliefs that such communities do not or will not consume healthier foods. Therefore, a cycle of structural racism reinvents itself through this investment in unhealthy food chances, constructing food deserts and swamps bestowed upon communities experiencing poverty and disproportionate adverse cardiovascular health conditions. Strengthening policy focused on the relationship between gentrification mitigation and health outcomes is needed.
This item is currently under embargo. It will be available starting 2026-06-14.
Advertisement
Supported by
A chemical reaction involving emissions from cars and buildings can negate their environmental benefits. New research shows what big cities can do about it.
By Austyn Gaffney
Oak trees with leaves like outstretched hands and thick, stately trunks reduce the effects of extreme heat around New York City’s five boroughs by shading sidewalks and sucking up planet-warming carbon.
These workhorses also naturally emit a chemical called isoprene. Alone, it is harmless. But when isoprene meets nitrogen oxides, a group of gasses emitted when fossil fuels are burned to power buildings and vehicles, the chemical reaction creates ground-level ozone, a harmful pollutant that can cause respiratory issues.
“The trees themselves don’t create any problems, in fact they solve a lot of problems for us in cities,” said Andrew Reinmann, associate professor of environmental science at the City University of New York and an author of a new study examining the isoprene emissions of city trees. “The problem is cars and fossil fuel combustion in cities that can essentially start to negate or take away some of the benefits that trees are providing us.”
The study, published in July in Environmental Science and Technology , looked at citywide data for every six-by-six-inch square of tree canopy across the five boroughs. The researchers analyzed the types of trees and modeled what could happen to ground-level ozone pollution if the city grew enough trees to cover 40 percent of its land.
In one scenario, if the city only planted oaks, ground-level ozone in Manhattan could increase more than 30 percent. In a second scenario, continuing to plant a mixed canopy like the one that exists today could increase peak ground-level ozone by about 10 percent, depending on the borough.
New York City repeatedly violates federal health-based air quality standards for ozone on hot days, according to the study. Every year more than 400 New Yorkers are estimated to die from ground-level ozone pollution , and more than 4,300 adults and children in the city visit emergency rooms for ozone-related asthma.
We are having trouble retrieving the article content.
Please enable JavaScript in your browser settings.
Thank you for your patience while we verify access. If you are in Reader mode please exit and log into your Times account, or subscribe for all of The Times.
Thank you for your patience while we verify access.
Already a subscriber? Log in .
Want all of The Times? Subscribe .
COMMENTS
Environmental Research is a multi-disciplinary journal publishing high quality and novel information about anthropogenic issues of global relevance and applicability in a wide range of environmental disciplines, and demonstrating …. View full aims & scope. $3590. Article publishing charge. for open access.
Abstract. Environment study deals with the processes in water, air, land, soil, and organisms which cause pollution or degradation of the environment. It is an interdisciplinary approach and its ...
This study analyses lease negotiations between these two parties in Ohio, noting disadvantages on the side of the landowner in the process. Ben Farrer. Robert Holahan. Soleil Smith. Research 19 ...
Environmental science is the multidisciplinary study of all aspects of the Earth's physical and biological environments. It encompasses environmental chemistry, soil science, ecology ...
Topics & Ideas: Environmental Chemistry. The impact of cobalt mining on water quality and the fate of contaminants in the environment. The role of atmospheric chemistry in shaping air quality and climate change. The impact of soil chemistry on nutrient availability and plant growth in wheat monoculture.
Environmental studies (EVS or EVST) is a multidisciplinary academic field which systematically studies human interaction with the environment.Environmental studies connects principles from the physical sciences, commerce/economics, the humanities, [1] and social sciences [2] to address complex contemporary environmental issues. It is a broad field of study that includes the natural environment ...
Overview. The Journal of Environmental Studies and Sciences offers a venue where relevant interdisciplinary research, practice and public policies can be recognized and evaluated. Increasingly, environmental studies integrate many different scientific and professional disciplines.
Short-term air pollution exposure is associated with lower severity and mixed features of manic episodes in hospitalized bipolar patients: A cross-sectional study in Milan, Italy. Michele Carugno, Dario Palpella, Alessandro Ceresa, Angela Cecilia Pesatori, Massimiliano Buoli. Article 110943. View PDF. Article preview.
Ecology is the study of how organisms interact with each other and their environment. It considers processes that occur at the population, community and ecosystem levels and has a particular focus ...
2005 — Volumes 97-99. Previous. Page 1 of 3. Read the latest articles of Environmental Research at ScienceDirect.com, Elsevier's leading platform of peer-reviewed scholarly literature.
Environmental studies is a constantly evolving field. Some of the most compelling research areas today include climate change, limnology, and conservation. By Lindsay VanSomeren Lindsay VanSomeren. Contributing Writer Lindsay VanSomeren writes education content for BestColleges, focusing on topics related to environmental and biological science
Addresses environmental concerns and research with coverage from more than 300 environmental studies journals and book reference content. Gale In Context: Opposing Viewpoints Supports science, current events, and language arts and presents all sides of important issues, empowering learners to develop information literacy and critical-thinking ...
OPEN ACCESS. Environmental Research: Ecology is a multidisciplinary, open access journal devoted to addressing important macroscale challenges at the interface of ecology, biodiversity and conservation. The journal bridges scientific progress and methodological advances with assessments of environmental change impacts on ecosystems, and the responses of those ecosystems to change, including ...
By Roxana Shafiee and Dan Schrag. Electrification through heat pumps is a primary strategy for decarbonizing residential heating in Massachusetts. The state has announced ambitions for 65% of residential-scale heating to be heat pump equipment by 2030 and 90% by 2040, earmarking nearly 4 billion dollars in incentives between 2022 and 2024 to ...
Environment More than 4 billion people may not have access to clean water ... a study suggests. ... membership organization dedicated to public engagement in scientific research and education (EIN ...
Qualitative research methods examine a wide range of topics in the study of environment and resource management. This first review on the topic highlights innovative and impactful research over the past few decades, drawing from social science disciplines that include sociology, geography, anthropology, political science, public policy, and psychology. We describe qualitative research methods ...
The International Journal of Environmental Research (IJER) is an international and multidisciplinary platform for researchers across the globe to swiftly publish, share and discuss new findings and developments in environmental science, engineering, and management.IJER is an interdisciplinary journal concerned with all aspects of the environment. These include but are not limited to air, water ...
The methodological needs of environmental studies are unique in the breadth of research questions that can be posed, calling for a textbook that covers a broad swath of approaches to conducting research with potentially many different kinds of evidence. Written specifically for social science-based research into the environment, this book ...
OPEN ACCESS. Environmental Research: Climate is a multidisciplinary, open access journal devoted to addressing important challenges concerning the physical science and assessment of climate systems and global change in a way that bridges efforts relating to impact/future risks, resilience, mitigation, adaptation, security and solutions in the broadest sense.
2) Renewable Energy. Renewable energy is another fairly mainstream topic in which there is much to learn and research. Although scientists have identified many forms of sustainable energy, such as wind, solar, and hydroelectric power, questions remain about how to best implement these energy sources.
New research on environmental sustainability from Harvard Business School faculty on issues including the role of companies to mitigate climate change, corporate social responsibility, reporting to stakeholders, government relations and development of Chief Sustainability Officers. ... A study of 30 years of corporate advocacy by Jonas Meckling ...
Introduction. Research environments matter. Environmental considerations such as robust cultures of research quality and support for researchers are thought to be the most influential predictors of research productivity.1, 2 Over 25 years ago, Bland and Ruffin1 identified 12 characteristics of research‐favourable environments in the international academic medicine literature spanning the ...
Environmental Science and Pollution Research (ESPR) serves the international community in all broad areas of environmental science and related subjects with emphasis on chemical compounds. Covers all areas of Environmental Science and related subjects. Publishes on the natural sciences, but also includes the impacts of legislation, regulation ...
Welcome to Environmental Studies! Immerse yourself in an exciting academic adventure where you'll have the opportunity to design your own degree, delve into coursework from various departments across several schools, and make your college degree your own.From protecting wildlife to promoting sustainability, from understanding climate change to addressing environmental racism, our ...
According to our observations, the best estimation model in the study area was the "OSR + SVM" model (R2 = 0.9590, RMSE = 13.9887, MAE = 10.8075), which considered four environmental factors.
Experimental ecology is the scientific study of ecological relationships and processes using controlled experiments, mostly which focus on understanding how living organisms interact with their natural environment. Experimental ecologists have multiple methods to conduct experiments such as manipulating environmental variables in controlled settings, which help investigate how these factors ...
The newly released Concordat for the Environmental Sustainability of Research and Innovation Practice, co-developed by the UK research and innovation community—including universities, research organisations, funders, and their partners—represents a broad ambition for the sector to transition to a sustainable future. 1 Concordat signatories are asked to prioritise action on leadership and ...
The disproportionate concentration of unhealthy food in communities of color in the United States may contribute to health inequities and food insecurity. Gentrification has been associated with residents' increased adverse health outcomes in its early and rapid phases. This study adds to the growing body of research by examining the relationship between gentrification, the food environment ...
Undoing the Benefits of City Trees: A study found that a chemical reaction involving emissions from cars and buildings can negate the environmental benefits of trees. New research shows what big ...