Elsevier

Remote Sensing of Environment

Volume 152, September 2014, Pages 136-149
Remote Sensing of Environment

Mapping short-rotation plantations at regional scale using MODIS time series: Case of eucalypt plantations in Brazil

https://doi.org/10.1016/j.rse.2014.05.015Get rights and content

Highlights

  • Mapping of Eucalyptus plantations at 250 m resolution from 2003 to 2010

  • Binary classification method based on MODIS 250 m NDVI time series

  • Test of seven matching functions, Bounding Envelope is the most efficient

  • Validation of the final map with various types of data at different scales

Abstract

Short-rotation plantations are extending worldwide due to the increased demand for pulp and wood. Reliable estimations of recent expansion of short-rotation plantation areas and associated land use changes are a prerequisite to assess their environmental impact on regional carbon and water cycles, and on climate. A binary classification methodology using MODerate resolution Imaging Spectroradiometer (MODIS) 16-day 250 m NDVI time series was developed and applied to classify Eucalyptus plantations across Brazil. The identification of Eucalyptus plantations specific patterns in the time series was based on the calculation of matching functions between the NDVI time series and a ~ 2 years long reference time series. Among the seven tested matching functions, the bounding envelope was the most successful. This method was robust to residual noise on the NDVI time series, and a threshold coefficient for the binary classification was adjusted using an omission-commission criteria. With this method, it was possible to detect any presence of Eucalyptus between 2003 and 2009 at monthly time-steps, including the periods of bare soils between two rotations that are typically 6–7 years long. The dates of first afforestation, of clear-cut at the end of a rotation, and of re-planting at the beginning of a new rotation were retrieved from the NDVI time series with a precision of ~ 66 days. The final almost continuous tri-dimensional map (space and time) was validated with three different datasets, from local to regional data. All three datasets gave similarly high global accuracy statistics, but a global underestimation of Eucalyptus areas compared to large scales census was observed. Discrepancies and way to improve the Eucalyptus area estimates were discussed in this study. The developed methodology could be applied to other short-rotation tree plantations.

Introduction

Tracking the land uses and land cover changes at a regional scale is of critical importance to analyze the modifications of global biogeochemical cycles and the impacts of environmental policies. Several global land cover maps have been produced from classification of remote sensing data (MODIS land cover product, USGS-IGBP, UMD, GLC2000, GlobCover, etc.). The classification algorithms were an ensemble supervised decision trees, e.g. for MODIS MCD12Q1 product (Friedl, Sulla-Menashe, Tan, Schneider, Ramankutty, Sibley and Huang, 2010), unsupervised classification followed by post-classification refinement, e.g. USGS-IGBP product (Loveland, Reed, Brown, Ohlen, Zhu, Yang and Merchant, 2000), clustered supervised and unsupervised classification, e.g. GlobCover (Bontemps, Defourny, Van Bogaert, Arino, Kalogirou and Ramos Perez, 2011). Such global maps obviously have a small number of classes and have a coarse spatial resolution, and are therefore of limited interest to monitor the area covered by specific crops or plantations. In parallel to the development of these global maps, researchers have used the same satellite image resources to produce maps of crop classes at farm or landscape levels in order to assess regionally and annually the land use changes of the main crops (e.g. Arvor et al., 2011, Brown et al., 2013, Epiphanio et al., 2010, Galford et al., 2010, Wardlow et al., 2007). All these studies have shown the potential of satellite image series to classify different crops and cropping systems, and therefore to assess the consequences of agricultural practices on land use changes. Indeed, the knowledge of the crop, forest or grassland phenology, together with their spectral signature, makes it possible to greatly improve the precision of the determination of subclasses. As a consequence, it is difficult to get a unified methodology and many different methods have been used to classify coarse resolution satellite image time series for the production of crop maps, each method depending on the objective of the study and of the crop type under consideration (García-Mora, Mas, & Hinkley, 2011).

While considerable efforts have been made to follow the deforestation of tropical forests, little has been done to follow the large expansion of forest plantations, while they can represent a significant area compared to deforestation in tropical and subtropical areas (Hansen et al., 2013). For instance, the area of Eucalyptus plantations in Brazil was 3.5 million ha in 2006 and now reaches approximately 5 million hectares (ABRAF, 2012). The Ministry of Agriculture of Brazil expects an increase of forest plantations up to a total of 9 million hectares by 2020 (Ministério da Agricultura website), with probably more than three quarters planted with Eucalyptus species. However, simulated scenarios show that the forest sector will need to reach 13.5 million hectares in 2020 to meet the expected demand for wood (EMBRAPA-Florestas, 2010). For comparison, perennial and annual croplands covered approximately 60 million hectares and native and cultivated pastures 160 million hectares in Brazil ((IBGE, 2006), compiled in Barretto, Berndes, Sparovek, and Wirsenius (2013)). Recent expansion of Eucalyptus plantations is probably the most important agrarian change that have occurred in recent years in some parts of Brazil (Kröger, 2012). Most of commercial Eucalyptus plantations in Brazil are managed to produce pulpwood, charcoal for steel industry, biomass for energy, and timber for solid wood items and different types of reconstituted panels (ABRAF, 2012). Selection of productive hybrids and clones, improved silviculture, and ideal soil and climate conditions of Brazil for eucalypt-based forestry have led to mean productivities of 40–45 m3 ha 1 yr 1 in commercial plantations across Brazil (Gonçalves et al., 2013).

There is a critical need to monitor the rapid expansion of short-rotation plantations like Eucalyptus plantations at regional scales. An accurate estimation of recent changes of Eucalyptus areas is a prerequisite to assess, for example, the environmental impact of afforestation or the impact of incentive policies on wood production at a regional scale (Cossalter & Pye-Smith, 2003). Eucalyptus plantations are mostly found in the Minas Gerais and São Paulo states (about half of total Eucalyptus areas in Brazil), but recent expansions mainly occur in the Mato Grosso do Sul and Para states (ABRAF, 2012). New Eucalyptus plantations are mainly established on pastures, or follow the conversion of other forest plantations (mainly pine or low-productive Eucalyptus stands), but few statistical data are available on these land use changes. It is likely that future afforestation rates will also be influenced by availability of suitable land (Piketty, Wichert, Fallot, & Aimola, 2009). Use of high resolution satellite images (e.g. Landsat) to monitor Eucalyptus expansion would be very difficult at large scales for both technical reasons and lack of data in some areas. Coarse resolution satellite images would be more suitable to follow large areas, and a resolution close to 250 m like the one of MODIS sensor onboard Terra satellite (for Red and Near infrared bands) can be sufficient to monitor commercial Eucalyptus plantations (le Maire et al., 2011b, Marsden et al., 2010). Another advantage of MODIS sensor is to have more than 12 years of continuous global data, with a high observation frequency.

Eucalyptus plantations have never been mapped across Brazil using remote sensing data, despite their rapid expansion, their economic importance and their large scale environmental impacts. The specific objectives of this study were 1) to develop a MODIS-based methodology to map Eucalyptus plantations and 2) to provide information on rotation practices through the detection of planting dates after afforestation and successive clearcuts, 3) to evaluate the potential of the method for regional area estimations.

Section snippets

Theoretical background

Classification of vegetation at coarse spatial resolution and decadal temporal scale is associated with two major issues: the classification method itself and the validation of the classification. Classification methods generally rely on the multispectral signals and their changes along the year, processed through classification algorithms. Classifying MODIS pixels at a 250 m spatial resolution is a challenge because spectral information is low at this sensor spatial resolution, and landscape

Choice of the best distance function

The selection of the best distance function was based on the omission–commission graph obtained on the high resolution Landsat classification (Fig. 4a). The classification results are very different depending on the chosen distance function. Following the approach of Boschetti et al. (2004), it was possible to state that some classification methods (i.e. some distance functions) were “dominant” to other ones, and therefore should be preferred. The Boundary Envelope (BE) and the Standardized

Map accuracy assessment

The BE method was shown to be well suited for Eucalyptus classification, both in terms of accuracy and robustness. Validation of such vegetation classification at these scales is a matter of numerous discussions, as emphasized in Strahler et al. (2006). Errors of classification could come from: 1) radiometric errors (sensor, atmospheric corrections, etc.) and percentage of clouds in the time series; 2) the mixed pixel issue and landscape heterogeneity: 3) the misclassification errors due to the

Conclusion

This study mapped the distribution of fast-growing Eucalyptus plantations across a large part of Brazil, using MODIS 250 m NDVI time series. The classification method is based on the NDVI variations during the two first years of a rotation, by computing a matching function based on the Bounding Envelope. Bounding Envelope distance was the best distance function among the seven functions tested. The classification method with BE function was efficient to classify the NDVI time series, robust to

Acknowledgements

This study was conducted in the frame of the SIGMA European Collaborative Project (FP7-ENV-2013 SIGMA — Stimulating Innovation for Global Monitoring of Agriculture and its Impact on the Environment in support of GEOGLAM — project no. 603719). The Landsat area includes a JECAM site (Joint Experiment of Crop Assessment and Monitoring). We thank the forest companies Duratex, International Paper of Brasil, and Fibria Celulose for providing the data. We thank NASA for sharing MODIS and Landsat data.

References (66)

  • J.L.d.M. Gonçalves et al.

    Integrating genetic and silvicultural strategies to minimize abiotic and biotic constraints in Brazilian eucalypt plantations

    Forest Ecology and Management

    (2013)
  • E. Keogh

    Exact indexing of dynamic time warping

  • G. le Maire et al.

    MODIS NDVI time-series allow the monitoring of Eucalyptus plantation biomass

    Remote Sensing of Environment

    (2011)
  • G. le Maire et al.

    Leaf area index estimation with MODIS reflectance time series and model inversion during full rotations of Eucalyptus plantations

    Remote Sensing of Environment

    (2011)
  • G. le Maire et al.

    Tree and stand light use efficiencies over a full rotation of single- and mixed-species Eucalyptus grandis and Acacia mangium plantations

    Forest Ecology and Management

    (2013)
  • M.P. Lenney et al.

    The status of agricultural lands in Egypt: The use of multitemporal NDVI features derived from landsat TM

    Remote Sensing of Environment

    (1996)
  • S. Lhermitte et al.

    A comparison of time series similarity measures for classification and change detection of ecosystem dynamics

    Remote Sensing of Environment

    (2011)
  • Z. Li et al.

    Mapping rubber tree growth in mainland Southeast Asia using time-series MODIS 250 m NDVI and statistical data

    Applied Geography

    (2012)
  • C. Marsden et al.

    Relating MODIS vegetation index time-series with structure, light absorption and stem production of fast-growing Eucalyptus plantations

    Forest Ecology and Management

    (2010)
  • C. Marsden et al.

    Modifying the G′DAY process-based model to simulate the spatial variability of Eucalyptus plantation growth on deep tropical soils

    Forest Ecology and Management

    (2013)
  • M.-G. Piketty et al.

    Assessing land availability to produce biomass for energy: The case of Brazilian charcoal for steel making

    Biomass and Bioenergy

    (2009)
  • K. Soudani et al.

    Evaluation of the onset of green-up in temperate deciduous broadleaf forests derived from Moderate Resolution Imaging Spectroradiometer (MODIS) data

    Remote Sensing of Environment

    (2008)
  • E. Vintrou et al.

    Crop area mapping in West Africa using landscape stratification of MODIS time series and comparison with existing global land products

    International Journal of Applied Earth Observation and Geoinformation

    (2012)
  • B.D. Wardlow et al.

    Analysis of time-series MODIS 250 m vegetation index data for crop classification in the U.S. Central Great Plains

    Remote Sensing of Environment

    (2007)
  • R.E. Wolfe et al.

    Achieving sub-pixel geolocation accuracy in support of MODIS land science

    Remote Sensing of Environment

    (2002)
  • C. Woodcock et al.

    The factor of scale in remote sensing

    Remote Sensing of Environment

    (1987)
  • R. Zurita-Milla et al.

    Downscaling time series of MERIS full resolution data to monitor vegetation seasonal dynamics

    Remote Sensing of Environment

    (2009)
  • ABRAF

    Chapter 01, Planted forests in Brazil. Statistical yearbook 2012

    (2012)
  • D. Arvor et al.

    Classification of MODIS EVI time series for crop mapping in the state of Mato Grosso, Brazil

    International Journal of Remote Sensing

    (2011)
  • H. Bagan et al.

    Land cover classification from MODIS EVI times-series data using SOM neural network

    International Journal of Remote Sensing

    (2005)
  • A.G.O.P. Barretto et al.

    Agricultural intensification in Brazil and its effects on land-use patterns: an analysis of the 1975–2006 period

    Global Change Biology

    (2013)
  • S. Bontemps et al.

    GLOBCOVER 2009 Products description and validation report, Vers. 2.2

    (2011)
  • M. Christina et al.

    Almost symmetrical vertical growth rates above and below ground in one of the world's most productive forests

    (2011)
  • Cited by (52)

    • Exploring a uniform procedure to map Eucalyptus plantations based on fused medium–high spatial resolution satellite images

      2021, International Journal of Applied Earth Observation and Geoinformation
      Citation Excerpt :

      However, Eucalyptus is an evergreen broadleaf tree with a high degree of spectral similarity to other evergreen broadleaf tree species. Previous Eucalyptus mapping studies mainly focused on medium and coarse spatial images such as MODIS, Landsat and Sentinel-2 (Forstmaier et al., 2020; le Maire et al., 2014; Somers et al., 2010). Although these free archived images greatly facilitate the mapping of Eucalyptus plantations by identifying its short-rotation phenological features, the mixed pixel problem and frequent cloud cover in tropical and subtropical regions introduce large uncertainties into the analysis of such datasets.

    • Multi-sensor prediction of Eucalyptus stand volume: A support vector approach

      2019, ISPRS Journal of Photogrammetry and Remote Sensing
    View all citing articles on Scopus
    View full text