The CHOVE-CHUVA Earth observation platform to monitor socio-environmental dynamics in Mato Grosso, Brazil

Abstract

Remote sensing science is expected to produce spatio-temporal indicators to help societies to address major global challenges. In this regard, we have implemented the CHOVE-CHUVA web platform to monitor socio-environmental dynamics in the Brazilian Amazon state of Mato Grosso. Result of a long-term collaboration between research labs, local NGOs, and administrations, this Space for Climate Observatory initiative relies on two major pillars: (1) visualizing and computing spatio-temporal indices derived from Earth Observation data and (2) collecting citizen information as part of collaborative science. A major asset of the platform is to gather, visualize, and process data covering a wide range of themes such as land status, land use, climate, natural vegetation, agriculture, and hydrology. The collaborative information refers to land use types that are still unusual in Mato Grosso, i.e., forest restoration and low-carbon agricultural practices. The implementation of the platform was based on a French open source geospatial data infrastructure named PRODIGE. Prospects for enhancing the platform include integrating new thematic information, making better use of raw Earth Observation data, improving interactions with end-users to better capture their interpretation of socio-environmental dynamics, and improving the platform’s efficiency to update data and process large study areas.

Introduction

Worldwide societies are facing critical global challenges as evidenced in the Sustainable Development Goals (SDGs), e.g., mitigate and adapt to climate change, promote sustainable management of forests, preserve biodiversity, and ensure food and energy production¹. In this context, remote sensing science is expected to play an essential role by producing manageable spatio-temporal indicators to monitor the achievement of these SDGs². In addition, Earth Observation (EO) images serve as critical tools not only to inform and communicate socio-environmental dynamics but also to actively engage communities and decision-makers, fostering collective actions to address global change challenges. Indeed, the continuous development of new sensors, coupled with easier access to data and algorithms, is transforming both the theory and practice of remote sensing science³, which has evolved rapidly during the last decades to accompany the advent of the digital era. These advances open up new perspectives for reducing the gap between data sciences and applications.

The most important trend relates to the ever-improving potential to process big Earth Observation data. Public and private initiatives to store and share access to a large amount and variety of spatial information, including remote sensing data, combined with new approaches to process these data (e.g., Google Earth Engine^4,5 and data cubes^6–8), resulted in the multiplication of large-scale products on a wide range of applications. For example, considerable efforts have been made to map land use^9–11 on a global^12,13 or national scale^14,15. Similarly, many complementary initiatives have focused on more specific themes such as forest monitoring^16,17), water bodies^18,19, fires²⁰, or climate variables²¹.

Another important trend in remote sensing science refers to collaborative and open science. This includes collaborative implementation of land use maps by gathering various research units with complementary expertise (e.g., the MapBiomas project in Brazil¹⁵). It also regards sharing image processing algorithms as R packages or Python libraries, for example. More challenging, seminal studies suggest moving towards citizen science for Earth Observation²². Citizen science has long existed in the field of geographical studies²³ and is gaining importance in the field of remote sensing, particularly in the form of “volunteer geographical information”²⁴. Examples of collaborative citizen science applied to Earth Observation include the Open Street Map²⁵ or the geo-wiki projects, designed for online sharing of land cover training and validation sample points²⁶. Nonetheless, few of these crowd-sourced projects have been integrated into operational monitoring systems to date²².

While the scientific community has easier access to remote sensing images, innovative image processing technologies, and reference samples from citizens or collaborative initiatives, major challenges remain to transform data into information and knowledge at all scales from local to global, and in communicating the latter to civil society, particularly to decision makers^24,27,28. As of 2025, the major issues to be addressed in this regard can be summarized as follows:

• Product inadequacy: Global, national, or regional scale products may not be relevant for specific local applications (e.g., the difficult assessment of forest policies based on remote sensing data²⁹) so end-users may be reluctant to use them.

• Product diversity: Various products on the same theme (e.g., land use or deforestation) may exist so that non-remote sensing experts may be confused when choosing the product that best suits their requirements.

• Product dispersion: Products from different themes are often available on different data portals so that end-users are not able to cross information easily.

• Product handling: Products often need to be downloaded in geographic file formats, which may not be easy to handle for non-experts in Geographic Information Sciences.

• Information extraction: Products are often easy to visualize, but also difficult to process when it comes to extracting relevant information on specific study areas.

• Contributing user: end-users are generally regarded as passive consumers of information rather than being encouraged to contribute actively.

In this context, important initiatives relying on Earth Observation data to derive spatio-temporal information are worth highlighting. For example, the Global Forest Watch³⁰ or Tropical Forest Monitoring³¹ are unmissable projects that provide easy-to-understand dashboards on land use changes in forest ecosystems worldwide or in tropical moist forests, respectively. ClimateCharts.net³² and the global climate monitor³³ enable users to visualize time series of climate variables for a specific location. The ASAP (Anomaly Hotspots of Agricultural Production) platform provides comparisons of NDVI time series from Sentinel-2, Landsat, and MODIS data for a given location³⁴. The Hydroweb platform processes and releases remote sensing-based measurements on water surface elevation for lakes and rivers, water quality, and water surface dynamics³⁵. The TerraClass^36,37 and MapBiomas^15,38 projects allow to view land use distribution and transitions in Brazil at various spatial scales (biome, state, or municipality). For its part, the WHISP (What is in that plot) project has been launched to accompany the implementation of the European Deforestation Regulation (EUDR) by allowing the retrieval and comparison of information on deforestation for specific study areas, using various global products and advocating for a “convergence of evidence” approach³⁹. WHISP relies on another interesting initiative named Earth Map⁴⁰ that takes advantage of the power of Google Earth Engine to derive zonal statistics for a wide variety of spatial variables without programming skills.

In line with these examples, the Space for Climate Observatory (SCO) international initiative was launched in 2019 to combat and adapt to the impacts of climate change using satellite data. SCO goals are threefold: (1) facilitate interoperability of satellite and local data and thus “make them talk”; (2) establish vulnerability indicators in support of climate change mitigation or coping strategies; and (3) pool these tools and transpose them to other territories through an open data rationale. In practice, there are today about 122 projects in 51 countries that take advantage of remote sensing data to address specific issues such as mapping tropical deforestation^41,42 or mangroves^43,44.

In this paper, we introduce the SCO CHOVE-CHUVA web platform⁴⁵ that aims at monitoring territorial dynamics in the Brazilian Amazon state of Mato Grosso. This platform stands out for 1) its multi-thematic approach based on a wide variety of data (i.e., on land use, climate, vegetation, agriculture, and hydrology), 2) its ability to display indices calculated “on-the-fly” for a user-defined study area without any prerequisite skills, and 3) its collaborative modules that enable end-users to contribute to the mapping of specific land use classes. In section 2, we justify the choice of Mato Grosso state as the study area and introduce the local partners. The computational architecture of the platform is described in section 3. Section 4 presents the main functionalities of the platform. Main achievements and future perspectives are discussed in section 5.

Study area and local partners

Study area

The CHOVE-CHUVA project focuses on socio-environmental dynamics in the Brazilian state of Mato Grosso (Fig. 1). This state concentrates most of the socio-environmental challenges faced in the Amazon and has thus long been studied by the remote sensing community. State administrations, in particular, rely on EO data to support the implementation of public policies and the definition of restoration actions in degraded areas to meet biodiversity targets and climate commitments. In this regard, EO-based platforms contribute to increasing the transparency of government actions.

Figure 1.

Presentation of the CHOVE-CHUVA web platform and location of the Mato Grosso study area.

The state of Mato Grosso is located in the central-western region of Brazil and covers approximately 906,000 km². It is distinguished by its ecological diversity due to a wide variety of vegetation types (namely, three biomes: the Amazon rainforest, the Brazilian Cerrado savanna, and the Pantanal wetlands), climatic regimes, and ecological processes that shape its natural landscapes. It is known for hosting a major agricultural frontier, supplying commodities to national and international markets. As of 2024, pastures and croplands in Mato Grosso represent 213,659 km² and 118,963 km², respectively³⁸, mainly located in former savannas and forest areas. In addition to agricultural expansion, agricultural intensification gained in importance during the 2000s, especially through the adoption of double cropping systems (i.e., two harvests per rainy season, usually soy and maize or soy and cotton)⁴⁶, favored by a long rainy season (in some regions), sometimes completed by irrigation systems⁴⁷. In this regard, irrigation represents an additional pressure on water resources already impacted by the proliferation of small dams and water tanks to support a diversification strategy to develop fish production⁴⁸. Finally, urbanization and industrialization increased the demand for energy and led to the construction of numerous hydropower dams⁴⁹.

The progress of the agricultural frontier has generated dramatic environmental damage⁵⁰. Mato Grosso has long been acknowledged for its high deforestation rates related to agricultural expansion⁵¹. Moreover, concerns related to forest degradation, i.e., damages in forest structure, composition, and function with no change in land use⁵², have emerged. The main drivers of forest degradation are forest fires, landscape fragmentation, selective logging, and climate change. The latter represents another major environmental issue per se since perspectives towards a reduced rainy season⁵³ and increased extreme events (especially droughts)⁵⁴ are expected to impact both cropping systems⁵⁵ and natural ecosystems⁵⁶.

At the same time, Brazil has committed to better protecting the environment by voluntarily reducing its greenhouse gas emissions⁵⁷. Thus, environmental considerations have emerged to improve the preservation of natural ecosystems (e.g., through the creation of protected areas) and regulate the agricultural sector through the adoption of low-carbon agricultural practices⁵⁸. This evolution in Brazil is supported by a Sectoral Plan for Mitigation and Adaptation to Climate Change and for the Consolidation of a Low-Carbon Economy in Agriculture (ABC and ABC+ plans)⁵⁹. A major objective of this plan is to encourage the adoption of sustainable management practices such as the restoration of degraded pastures, the reforestation of riparian vegetation, and the promotion of agricultural practices such as crop-livestock integration systems⁶⁰ or no-tillage practices.

Interactions with local partners to define platform’s content

Result of long-term collaborations (more than 20 years) between research laboratories and local partners, the project gathers a multidisciplinary consortium with complementary interests in supporting sustainable development in Mato Grosso. Beyond academic partners, the platform was co-constructed with local stakeholders to identify their needs in maps and indices to be implemented primarily on the platform. These stakeholders represent non-government organizations (NGO) and local administrations that share a common interest in sustainable development, albeit from different perspectives:

• ICV (Instituto Centro de Vida) is an NGO that builds shared solutions for the sustainability of land use and natural resources. ICV works on issues of transparency, environmental governance, and public policies, especially at the municipal level in northern Mato Grosso. ICV has a long experience in the production and publication of spatial information through GIS web platforms, but mainly with a restricted focus on deforestation and selective logging. In this project, they were interested in the possibility of easily accessing additional information (particularly on precipitation and hydrology) and collaborative modules.

• FEC (Fundação Ecologica Cristalino) is an environmental NGO that contributes to the conservation and sustainable development in northern Mato Grosso through the management of protected areas, social communication, education for conservation, and scientific research. FEC has little experience in the field of GIS, but has extensive experience in scientific collaboration (especially in ecology and climatology). In charge of preserving the Cristalino conservation unit, FEC is interested in monitoring threats to protected areas, especially deforestation, forest degradation, and fires.

• CAT (Clube dos Amigos da Terra) is a non-profit association that promotes the technological development of agribusiness while respecting the environment. CAT organizes technological events to disseminate low-carbon agricultural practices, strengthens women’s participation, and contributes to environmental education within the school network. CAT also acts as a certification organism for the Roundtable for Responsible Soy (RTRS). CAT has little experience in the field of GIS and is interested in the possibility of monitoring and communicating on positive environmental actions carried out locally: reforestation of riparian vegetation, adoption of integrated practices, and certifications of farms.

• SEMA-MT is a public administration that promotes the control, preservation, conservation, and recovery of the environment, as well as formulates, proposes, and executes the environmental policies in Mato Grosso. SEMA-MT produces and publishes spatial data, notably through the SEMA geoportal⁶¹, which enables users to visualize and access a wide variety of data, but does not allow for the computation of spatio-temporal indices.

Methods: implementation of the CHOVE-CHUVA web platform

The CHOVE-CHUVA platform is based on an open source geospatial data infrastructure named PRODIGE⁶². PRODIGE is a Digital Commons for the acquisition, storage, cataloging, enhancement, and distribution of geographic data. It is identified and referenced in the French inter-ministerial open source software catalog, named SILL⁶³. The CHOVE-CHUVA platform is using dockerized environments supported by the symfony and phalcon frameworks on a Linux Debian-based operating system. The database management system is PostgreSQL, with its PostGIS spatial cartridge and MapServer to display data on maps. Data processing is carried out and automated with the R software environment. The results are processed by API (Rplumber) before being interpreted by the Chart.js JavaScript library, enabling the generation of interactive graphics (Fig. 2).

Figure 2.

Architecture of the CHOVE-CHUVA web platform.

PRODIGE is a software package made up of core and peripheral modules (Fig. 3), which can be activated or deactivated as required, and whose data management is based on the Postgres open source relational database and its dedicated PostGIS module. The core modules used in the implementation of the CHOVE-CHUVA platform are:

• Admin: PRODIGE’s central module, integrating user and rights management and enabling global configuration of the platform.

• Download: Resource download module to perform transformations, manage vector and raster data types, re-projection, or geographic slicing.

• WxS diffusion: A module that natively offers REST API-type web dissemination services in GeoJSON format compliant with OGC standards (WMS, WFS, WMTS).

• Visualizer: Geographic data visualization module in the form of a dynamic map that can be edited by the administrator but also manipulated and exported by the end user. The interactive map viewer lets you navigate through a platform’s data assets, query data, and cross-reference geographic information on a map.

• API: Resource access API (data series) with authentication and access rights management.

• Quality: Module for comparing data produced to a standard with regard to the naming of columns and their type, enabling the creation of data that conforms to a standard.

• Map composer: Module for composing interactive maps, based on the Mapserver cartographic rendering engine and enabling map context generation (OWS GeoJSON format).

• Catalog: The metadata catalog is the search engine for the platform’s data, maps, and services. It provides discovery and access to referenced databases and cartographic products, facilitating the exchange and sharing of information.

The core modules are complemented by the following peripheral modules:

• Collaborative: Collaborative work tools based on a customized content management system, enabling electronic document management, calendar sharing, mailing, and community building tools.

• Edition: Communication module based on a customized CMS, enabling management of editorial content, news, and forms.

• Dataviz: Interactive graphical data visualisation tools configurable by the user.

• Territorial basis: Back-office-controlled module for breakdown and presentation of resources by geographical area, enabling information to be retrieved by API.

• Dedicated application: Custom application generator enabling administrators to create applications integrating dashboards, interactive cartography, and geographic data entry forms.

• Co-production: Online geographic data editing module integrated with the map viewer module and based on the main API.

• Table editing: Module for online editing of tabular data, enabling data to be searched in tables and edited according to user rights.

• Urbadata: Data harvesting module in DCAT format.

PRODIGE installation documentation is available on the main web page of the PRODIGE project⁶² and the essential base application containers deployed on the servers for the successful implementation of the CHOVE-CHUVA platform are detailed in supplementary information (Table S1).

Figure 3.

Prodige’s core and peripheral modules.

Results: presentation of the CHOVE-CHUVA web platform

The CHOVE-CHUVA web platform is organized in four seminal blocks: 1) listing of thematic layers, 2) identification of user-defined study areas, 3) computation and visualization of spatio-temporal indices, and 4) collection of collaborative data (Fig. 1). On this basis, the platform aims to meet two major objectives identified with local partners: 1) to carry remote sensing-based information to non-remote sensing experts and 2) to include citizen science principles by enabling users to contribute to the mapping of specific land use classes.

Spatio-temporal visualization of socio-environmental dynamics

The first objective of the CHOVE-CHUVA web platform is to enable users to visualize and process spatial information to derive spatio-temporal indices to qualify and quantify territorial dynamics for user-defined areas (Fig. 4). These indices can be computed for administrative units (i.e., municipalities), land status (e.g., indigenous lands, conservation units, or public settlements), or by manual delineation of an area of interest. As defined with local partners, a major asset of the platform is to gather spatio-temporal information covering a wide range of themes such as land status and land use, climate, natural vegetation, agriculture, and hydrology (Fig. 5).

Figure 4.

Visualization of a spatio-temporal index (e.g., irrigation) and legend in the CHOVE-CHUVA web platform for a defined study area (e.g., Sorriso municipality).

Figure 5.

Illustration of spatio-temporal indices for a manually-defined study area (red polygon).

Land status and land use

The first information about land status and land use includes information on conservation units, indigenous lands, and public settlements. In addition, it also integrates information on the Rural Environmental Registry (CAR – Cadastro Ambiental Rural) and consolidated rural areas (Fig. 5B).

• Conservation units, indigenous lands, and public settlements: It refers to information on land status. A conservation unit is a territorial area subject to a special administration regime with conservation objectives for its environmental resources. An indigenous land is a territorial area inhabited by one or more indigenous communities, which becomes the property of the Union (i.e., Brazil) after a regular administrative process. Finally, public settlements are areas divided and granted in plots for farmers who cannot afford to buy a rural property. Shapefiles with the limits of these land units are available at⁶¹.

• Rural environmental registry: The Rural Environmental Registry (CAR) is a national public electronic registry implemented throughout the country in 2012. CAR is mandatory for all rural properties and aims to integrate environmental information from rural properties to compose a database for control, monitoring, and environmental and economic planning. The registration of rural properties in the CAR is done through an electronic system in which owners or legal representatives of the areas enter geospatial information that characterizes the rural property, including its boundaries and the location of Permanently Protected Areas (PPAs) and Legal Reserve Areas (LRAs), as defined by the Brazilian Forest Code. After submission, the data declared in the CAR are analyzed and approved by the competent environmental agency, which verifies the conformity of the information and the adherence to legal requirements. This data can be found at⁶¹.

• Consolidated rural areas: The Brazilian Forest Code, published in 2012, defines a “consolidated rural area” (CRA) as an area within a rural property that was occupied (i.e., cleared and with established agricultural or infrastructure use) prior to July 22, 2008. Environmental infractions that occurred before that date are granted amnesty and landowners are exempted from the obligation to restore the degraded areas. The map of consolidated rural areas in 2008 was produced based on the visual interpretation and digitization of satellite imagery at ICV and reviewed by analysts from SEMA-MT⁶⁴. This product is available at⁶¹.

• Land use: We included land use maps from the MapBiomas initiative^15,38. MapBiomas is a collaborative network formed by NGOs, universities, and companies to monitor land use dynamics in the Brazilian territory. Land use maps have been produced annually at 30 m spatial resolution since 1985 (Fig. 5D). The collection 10 has 30 legend classes (Level 4) and covers 40 years of land cover and land use data in Brazil (1985 – 2024).

Climate

The climate of Mato Grosso is characterized by a well-defined rainy season between October and April, primarily controlled by the South American Monsoon System (SAMS)⁶⁵. Although quite regular, rainfall regimes are also affected by strong spatio-temporal variability. This variability is exacerbated by strong interannual variations, correlated to El Niño Southern Oscillation (ENSO)⁶⁶. However, while climate conditions have proven to be ideal for supporting the development of an intensive agricultural model in the region, the deforestation associated with this model also impacts the global, regional, and local climate. Ironically, climate change is expected to reduce the duration of the rainy season in the long term and thus to hinder the adoption of intensive double cropping systems in the future^55,67,68. In other words, climate change calls into question the sustainability of the agricultural model in Mato Grosso⁵⁸.

The climate theme resumes information developed by researchers from the project consortium based on rainfall estimates from the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS)²¹. CHIRPS is a near-global precipitation dataset created from cooperation between USGS (United States Geological Survey) and CHC (Climate Hazards Center) scientists in order to develop techniques for producing rainfall maps, especially in areas where in-situ data is scarce. Covering - (and all longitudes) and ranging from 1981 to the near-present, CHIRPS incorporates satellite imagery and in-situ station data to create gridded precipitation time series for analysis of trends and climate extremes.

The maps and indices illustrate the evolution of:

• Precipitation estimates: Monthly and annual rainfall since 1981 (Fig. 5F).

• Temporality of the rainy season: Onset dates, end dates, and duration of the rainy season, based on the Anomalous Accumulation method^68,69 (Fig. 5G).

• Evolution of precipitation extremes: Number of rainy days, Single Day Intensity Index, maximum consecutive 5-days rainfall, and maximum number of consecutive wet days, as defined by the Expert Team on Climate Change Detection and Indices (ETCCDI)⁵⁴.

• Distribution of annual precipitation by daily intensity: Types of precipitation (light, moderate, intense, or heavy) and their contribution to annual precipitation.

Natural vegetation

The state of Mato Grosso has long been considered as a major hotspot for deforestation in the Amazon. According to official statistics from the Prodes project for the “Annual Monitoring of Native Vegetation Suppression”, led by the National Institute for Space Research (INPE), 202,329 km² of forest and 14,622 km² of non-forest vegetation had been cleared in the Amazonian part of the state by 2023⁷⁰. Meanwhile, vegetation loss reached 158,932 km² and 9,975 km² in the Cerrado and Pantanal biomes, respectively⁷⁰. Beyond deforestation, concerns about forest degradation have emerged over the last decades. In the Amazon, forest degradation (at least 364,748 km² between 2001 and 2018⁷¹) exceeds deforestation (215,784 km² during the same period⁷⁰)^71,72. For these reasons, we focused on deforestation, forest degradation, and three of its major drivers (i.e, forest fires, selective logging, and landscape fragmentation).

• Deforestation: The Prodes project has been monitoring natural vegetation loss in Brazilian biomes, most notably the Amazon, using remote sensing since 1988^73,74 (Fig. 5H). Prodes relies on images from the LANDSAT and Sentinel-2 satellites. In the Amazon, it tracks the loss of both forest and non-forest vegetation⁷³. For this region, forest clearings larger than 6.25 ha are considered deforestation, while the minimum mapped area for non-forest vegetation is 1 ha⁷⁴. This long-term monitoring system has proven to be of great importance for actions and planning of public policies, particularly in the Amazon.

• Forest degradation: The European Commission’s Joint Research Centre mapped changes in tropical moist forests (TMF) using 40 years of Landsat time series¹⁷ (Fig. 5I). The maps of “degradation year” at 30 m resolution depict the year of forest disturbances related to forest degradation, i.e., a temporary disturbance in the forest due to selective logging, fire, or weather events.

• Forest fires: Forest loss due to fire²⁰ was mapped within the extent of the global 30 m resolution 2001–2022 forest loss dataset¹⁶. Fires include wildfires, escaped fires from slash-and-burn agriculture, hunting and other human activities, and intentionally set fires (e.g., for land grabbing).

• Selective logging: It includes maps of logged areas between 2013 and 2021 in the state of Mato Grosso⁷⁵. The methodology is based on the Logging Monitoring System (SIMEX) developed by the Amazon Institute for Man and the Environment (Imazon), with adaptations by ICV and then validated by SEMA-MT. Logged areas corresponding to operations carried out under a forestry authorization were considered legal, whereas the others were classified as illegal. The indices shown on the platform refer to the annual and cumulative areas of selective logging.

• Landscape metrics: Landscape metrics related to fragmentation (e.g., number and density of forest patches) and heterogeneity (e.g., Shannon Diversity Index) were analyzed using the R package landscapemetrics. These metrics were computed based on land use maps from the MapBiomas initiative¹⁵ to assess landscape structure dynamics, considering both compositional and configurational changes.

Agriculture

Agriculture is the main cause of land use changes in Mato Grosso⁵¹. It implies both 1) land use conversions of forests and savannas due to agricultural expansion and 2) land use modifications due to the intensification of agricultural practices. Whereas land use conversions can be observed in land use maps (see section 4.1.1), we here focus on agricultural practices monitored with remote sensing data⁷⁶. We provide information on the adoption of intensive double cropping systems in Mato Grosso, considering the sequential cultivation of soybean and maize or soybean and cotton during the same rainy season⁴⁶. We also added information on irrigation, especially used to ensure water supply at the end of the rainy season⁴⁷.

• Cropping systems: The crop maps are derived from annual land cover classification data for the state of Mato Grosso from 2001 to 2017⁷⁷ (Fig. 5 - E). These data were produced using a machine learning algorithm (Support Vector Machine) to classify MODIS image time series at 250 m spatial resolution⁷⁸. The classifications consider 14 classes, including 8 agricultural classes that allow the discrimination of major cropping systems in Mato Grosso. The corresponding index shows the evolution in the proportion of single and double cropping systems for an area of interest.

• Irrigation: We included maps of center pivot irrigation systems and derived indices on the area and number of pivots between 1985 and 2019. The data was published in 2021 in the second edition of the Survey of Centre Pivot Irrigated Agriculture in Brazil, released by ANA (Brazilian National Water Agency), Embrapa (Brazilian Agricultural Research Corporation), and INPE⁷⁹.

Hydrology

Whereas deforestation in Mato Grosso has significantly decreased since the mid-2000s, the development model did not change radically, remaining focused on agricultural expansion, agricultural intensification, diversification, and industrialization. Such evolution questions the potential impacts on other natural resources beyond forest and savanna ecosystems. This issue is especially valid for water resources since 1) crop and pasture expansion have long affected riparian vegetation⁸⁰, 2) agricultural intensification implies the use of agrochemicals with potential impacts on water quality⁸¹, 3) agricultural diversification through fish farming implies the construction of numerous small water reservoirs⁴⁸, and 4) industrialization (associated with rapid urbanization) requires energy provided by hydropower dams⁴⁹.

In this section, we thus provide important information on rivers, sources, and water reservoirs in Mato Grosso. The Brazilian Forest Service prepared a hydrographic map at a scale of 1:25,000, further validated by the SEMA-MT technical department. This product represents a breakthrough in the treatment of the hydrographic theme, for which the State only had mapping at a scale of 1:100,000, carried out by the Brazilian Institute of Geography and Statistics (IBGE). Based on it, we computed basic statistics for user-defined areas (Fig. 5 - C):

• Rivers: Number of sources and cumulative length of watercourses.

• Reservoirs: Number and cumulative area of water reservoirs.

• Riparian vegetation: Proportion of non-forest (based on MapBiomas land use maps) in 300 m buffer areas around water bodies.

Collaborative modules to collect citizen data

The second objective of the project is to collect collaborative data from citizens and local partners. The collected information especially refers to the monitoring of land use types that are still unusual in Mato Grosso and for which it is thus difficult to obtain sufficiently complete information to perform statistically significant analysis (e.g., validation of satellite image classifications). To collect this collaborative information, a tool has been implemented to allow users to locate and delineate land use types on Sentinel-2 satellite images (Fig. 6). The land use types of interest are:

• Forest restoration: Many farmers are encouraged to restore natural vegetation in degraded riparian areas to comply with environmental legislation. These refer to many small areas spread across the state territory.

• Crop-Livestock-Forest integrated systems: Farmers are encouraged to adopt low-carbon agricultural practices as part of the ABC low-carbon agriculture plan⁸². Yet, there is still little information on the level of adoption of such practices, e.g., crop-livestock-forest integrated systems.

• RTRS Certified farms: Farmers are encouraged to comply with high-standard social and environmental rules in order to get access to specific markets. In this regard, the RTRS certification aims to ensure that soy is produced in environmentally correct, socially appropriate, and economically viable conditions⁸³.

Figure 6.

Illustration of the tool for collecting collaborative data. Users can manually delineate areas corresponding to land use classes of interest (forest restoration, agricultural practices, and RTRS certification) and add complementary information (e.g., area, date).

Discussion and perspectives

The current version of the CHOVE-CHUVA web platform⁴⁵ represents an original achievement that opens new perspectives for future improvements. In this regard, the platform is expected to evolve in the coming years to enhance 1) the information content, 2) the collaborative science part of the project, and 3) the efficiency of the platform implementation.

Information content

To date, the thematic layers and derived spatio-temporal indices encompass a large diversity of themes. In order to improve the information content of the platform, we will focus on a few specific points.

Additional thematic information

A first improvement concerns the integration of additional information to complete the existing themes already included in the platform. For example, recent data on vegetation cover⁸⁴, vegetation regrowth³⁷, or vegetation height in tropical forests⁸⁵ provide valuable new information on natural vegetation and deserve to be disseminated to a wide audience.

Regarding croplands, the maps of crop types and practices should be updated, possibly using the R sits package developed by the same authors who released the maps currently used in the platform⁸⁶. In addition, other maps of croplands, such as SojaMaps⁸⁷, could be included for comparison. Maps of center pivot irrigation systems should also be updated and completed. While the MapBiomas initiative recently released maps of such irrigation systems, innovative techniques based on deep learning algorithms to detect these systems are regularly published⁴⁷. Importantly, since a major challenge for the agricultural sector in Brazil refers to the adoption of low-carbon agricultural practices promoted through the ABC plan, we should consider including specific maps of integrated crop-livestock systems^82,88. In addition, although cattle ranching is a major driver of land use changes in Mato Grosso, pasture (201,982 km² in 2024³⁸) is here only considered as one land use class among others in the MapBiomas product. In the future, we intend to include additional information related to pasture management techniques since pasture intensification and restoration is a key strategy to curb deforestation⁸⁹.

Hydrological dynamics will be further characterized considering variations in water surface elevation (WSE) derived from radar altimetry. This data source has demonstrated significant potential for revealing complex hydrologic processes in the Pantanal biome⁹⁰. Specifically, we will include radar altimetry-based WSE observations from the Theia Hydroweb Operational River Water Level dataset³⁵. Hydroweb time series (2002 to present) are calculated at virtual stations using satellite acquisitions from the Sentinel-3, Sentinel-6, SWOT, Jason-3, and Jason-2 missions.

Urban land use should also be taken into account in the future. Indeed, about 80% of Amazonian populations live in cities, and urbanization of Amazonian territories is an emerging concern with implications for energy production, waste management, and exacerbation of global warming in urban heat islands⁹¹.

With regard to climate indices, an initial improvement will consist of including additional variables. For example, the Brazilian Daily Weather Gridded Data (BR-DWGD) based on the interpolation of in situ data provides valuable long-term information on precipitation, temperature, solar radiation, wind speed, and relative humidity in Brazil since 1961⁹². In this regard, it enables a fine analysis of climate change to be completed with outputs of climate models by 2050 and 2100, depending on various scenarios of greenhouse gas emissions. Such data are essential to raise public awareness of global change issues, which could be supported by the implementation of the well-known warming stripes on the past and future evolution of air temperatures⁹³. We will also consider including air quality maps from Copernicus Atmosphere Monitoring Service (CAMS)⁹⁴ that provide daily analyses of the worldwide long-range transport of atmospheric pollutants. This issue is especially relevant in the context of increasing fire frequency and intensity in the Amazon and their consequent impacts on human health⁹⁵. Finally, innovative initiatives such as mapping most influential precipitationsheds (MIPs), i.e., the upwind surface areas providing evapotranspiration to a specific area’s precipitation⁹⁶, appear relevant.

Data comparison

The CHOVE-CHUVA web platform was initially designed to inform citizens about a wide range of socio-environmental themes. To achieve this, we opted to include only one source of data per theme. However, for many applications, several products exist, and users should be able to compare them for the sake of transparency. More specifically, the main purpose of comparing data is to make users aware of the divergences and convergences between products. For example, there are inconsistencies between the rivers and sources mapped on the state scale in the “hydrology” section and the rivers and sources mapped in the Rural Environmental Registry (CAR River Sources and CAR Permanent Preservation Area) in the “land tenure” section of the platform. In the same line, the deforestation charts show divergences between estimates by Prodes and TMF due to different methodologies and definitions. Whereas these differences do not generally affect trend analysis at a large scale (e.g., state or biome), significant discrepancies may arise at a local scale (e.g., municipality or user-defined area) and should be reported to the end-users in order to avoid misleading interpretation due to biased data or inappropriate cartographic representations⁹⁷. So, whereas the deforestation maps currently shown on the platform are official data from the Prodes project, other maps could be displayed, e.g., Global Forest Change¹⁶, Tropical Moist Forest¹⁷, TerraClass³⁷, or MapBiomas¹⁵. Since some of these products are already included in the platform for other themes than deforestation (i.e., MapBiomas data for land use evolution and TMF for forest degradation), we intend to implement a specific tool to compare deforestation maps for user-defined areas. More generally, product comparison is of the utmost importance at a time when important environmental policies, such as the European Deforestation Regulation (EUDR), are being established, raising issues about the implementation of effective monitoring methods based on a “convergence of evidence” approach between independent and heterogeneous products^39,98.

Collaborative science

An important asset of the CHOVE-CHUVA platform is to allow users to contribute to the collection of spatial information on predefined themes (e.g., low-carbon agricultural practices or reforestation areas). In this regard, the collaborative science dimension of the platform is limited to a citizen data perspective, i.e., users’ contribution refers to objective information on land use practices in support of future scientific research. Yet, we also wish to allow users to provide a more subjective interpretation of the spatio-temporal dynamics shown on the platform, helping them to create narrative mappings of their own interests⁹⁷, e.g., users may comment on causes of fire or on reasons that convinced them to adopt new agricultural practices. Facilitating bi-directional communication between EO and collaborative information should empower citizens to participate in environmental management and allow the platform to evolve into a citizen observatory²⁴. This could be done through surveys on specific themes like climate change perception, for example. Whereas previous studies based on field campaigns in the Amazon have evidenced discrepancies between climate change perception by local communities and scientific measurements^99,100, a web platform providing easy access to both surveys and climate data would represent an excellent opportunity to gather more respondents for large-scale analysis.

That said, and beyond the laudable intentions, we must recognize that the implementation of collaborative modules remains exploratory and very difficult. In practice, for the moment, it has raised more ethical and technological questions than it has helped to solve environmental problems. For example, additional efforts are needed to better manage differentiated access rights or conflicting changes in certain areas. Furthermore, despite our intention to keep the modules as simple as possible, we are encountering difficulties in encouraging users to contribute to the platform. As a result, and with the aim of attracting more users from diverse backgrounds, we are now considering offering tutorials explaining how to use the platform in environmental education projects. In this regard, we will soon be collaborating with universities and local governments to train teachers across the state, with the ultimate goal of using the platform as a tool for students to 1) analyze local landscape dynamics and climate variability and 2) enrich local environmental knowledge on the platform.

Platform architectural efficiency

As expressed in the present paper, this first release of the CHOVE-CHUVA web platform provides many options to visualize thematic layers and compute spatio-temporal indices. Yet, we expect to improve the overall efficiency of the platform in the coming months.

Improve data access and processing

At the moment, data are downloaded and stored in local servers and updated manually. Thus, a first objective would consist of accessing thematic layers as Web Mapping Services (WMS) to ensure automatic updates. However, the use of WMS may slow down the visualization process and the computation of spatio-temporal indices. Indeed, whereas the indices are processed on-the-fly for small manually delineated areas (< 500 km²), the same indices at the administrative level (municipalities and protected areas) are computed in anticipation to speed up the display. Ideally, we would have liked to do all processes on-the-fly, as for the small areas, but the processing time tended to increase exponentially with increasing study area.

In addition, we intend to include more raw Earth observation information. Especially, the display of time series of vegetation indices appears very interesting to monitor and comment on agricultural practices, deforestation and forest degradation, or forest restoration and natural regeneration. Such time series are usually produced based on MODIS, Landsat, or Sentinel-2 time series³⁴, depending on the required spatial and temporal resolution and the historicity. Recent initiatives to facilitate access to such data in large repositories via data cubes⁸ offer new opportunities to process data online before displaying it on web platforms such as CHOVE-CHUVA.

Improve the visualization and interaction with end-users

In order to enhance the user’s experience on the platform, we intend to include several key improvements. First, users should be able to define their area of interest by selecting polygons from one of the various datasets shown on the platform. For example, a polygon delineating a property in the Rural Environmental Registry could be selected to get the exact limits of the study area. Complementarily, users should also be able to upload polygons corresponding to their area of interest. Ideally, they should also be allowed to process multiple polygons for specific applications. For example, farms certified by the RTRS are spread throughout the Mato Grosso state. To assess whether land use dynamics differ between certified and non-certified farms, this information should, therefore, be analyzed together. In the same line, we will improve the visualization of charts, including the possibility of producing multiple charts to enhance inter-site comparison. In addition, we intend to add the possibility for users to download the values of spatio-temporal indices to be further processed locally and shared. Finally, we will optimize the web platform for better use on cell phone and tablet screens, ensuring seamless access and usability on any device.

Conclusion

The CHOVE-CHUVA project is a Space for Climate Observatory (SCO) initiative to develop a web platform to monitor socio-environmental dynamics in the Southern Amazon Brazilian state of Mato Grosso. The platform gathers spatio-temporal information on different themes such as land status, land use, natural vegetation, agriculture, hydrology, and climate. It allows computing spatio-temporal indices for user-defined areas, e.g., a municipality, a protected area, or a private property. In that regard, the platform aims to solve some of the issues identified in section 1 to bridge the gap between EO-derived information and end users. It also includes tools to collect collaborative citizen information on important initiatives such as forest restoration, low-carbon agricultural practices, or RTRS certification. Future perspectives for the improvement of the web platform include: 1) integrating new thematic information, 2) enabling data comparison, 3) improving the use of raw spatial data, 4) improving the interactions with end-users to better capture their interpretation of socio-environmental dynamics, 5) improving the efficiency of the platform to better update the data and process large study areas, 6) promoting CHOVE-CHUVA in academic and non-academic networks, and 7) extending the platform to other ecosystems, e.g. the Pantanal wetland.

Author contributions

The concept for this work was developed by D. A., A. B. and A. B.. D. A., V. D., B. F., U. R., J. B., A. B., V. S., A. D., C. d. S., M. S., R. F. and P. K. listed the spatio-temporal indices to be displayed. D. A., A. B., U. R., J. D. and L. R. processed the data and implemented the web platform. D. A. wrote the original version of the paper and all authors contributed on specific sections. D. A., A. B., and A. B. managed the administration and acquisition of project funds. All authors have read and approved the published version of the manuscript.

Funding

This work was supported by 1) the French National Center for Space Studies (CNES) through the CHOVE-CHUVA Space For Climate Observatory (SCO CHOVE-CHUVA), 2) the CNES through the SEMTI-SENT project, 3) the French National Research Agency (ANR) through the TELKANTE LAB project, 4) the recruitment of two engineers under the France Relance plan, and 5) the European Joint Programme (EJP) for the institutional support and for fostering scientific collaboration, and the ANR for the financial support of the Project: Soil ecosystem services under sustainable intensification of agriculture - looking for innovative mapping and monitoring at multiple scales (ID number 31/SOIL-ES).

Data availability

The data (spatio-temporal indices) that support the findings of this study are openly shown at www.sco.chove-chuva.org and are available from the corresponding author on request.

Declarations

Competing interests

The authors declare no competing interests.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-36640-w.