Woody plant cover estimation in drylands from Earth Observation based seasonal metrics

Brandt, M., Hiernaux, P., Tagesson, T., Verger, A., Rasmussen, K., Diouf, A.A., Mbow, C., Mougin, E., Fensholt, R., 2016. Woody plant cover estimation in drylands from Earth Observation based seasonal metrics. Remote Sensing of Environment 172, 28–38.

Download a free copy here (until 28 December 2015)

Trees, shrubs and bushes are an important element of savanna ecosystems and for livelihoods in dryland areas dependent on fuel–wood supply. During the past decades, several studies have seriously questioned prevailing narratives of a widespread and Sahel-wide decrease in woody cover, commending the relevance of large scale woody cover monitoring systems.


Most studies estimating tree canopy cover with remote sensing rely on high resolution imagery which allow direct mapping at a scale recognizing trees of a certain size as objects. However, imageries with a spatial resolution of 1–5 m are cumbersome to process, expensive, susceptible to clouds, and do only provide a static situation for a limited spatial area. Moreover, considering trees as objects, smaller isolated woody plant are missed. Moreover, the reliability of global tree cover products in semi-arid regions with open tree cover is contested.

We suggest an approach driven by vegetation phenology including in situ measured woody cover data across the Sahel and seasonal metrics from time series of MODIS and SPOT-VGT. The method is an indirect estimation of the canopy cover of all woody phanerophytes including trees, shrubs and bushes, and is based on the significant difference in phenophases of woody plants as compared to that of the herbaceous plants. In the Sahel, annual herbaceous plants are only green during the rainy season from June to October and senescence occurs after flowering in September towards the last rain events of the season. The leafing of most trees and shrubs is longer, with several evergreen species, and many woody species green-up ahead of the rains during the last month of the dry season, while annual herbaceous are dependent on the first rains to germinate.


Figure from Brandt et al., 2016: Seasonal distribution of woody leaf mass depending on the phenological type, modeled within the STEP primary production simulation model (Mougin, Lo Seen, Rambal, Gaston, & Hiernaux, 1995). The months of the wet season during which herbaceous grow are highlighted in a shaded box. Illustrations of typical herbaceous growing curves can be found in Mougin et al. (2014).

We tested 10 metrics representing the annual FAPAR dynamics for their ability to reproduce in situ woody cover at 43 sites (163 observations between 1993 and 2013). Both multi-year field data and satellite metrics are averaged to produce a steady map. Multiple regression models using the integral of FAPAR from the onset of the dry season to the onset of the rainy season, the start date of the growing season and the rate of decrease of the FAPAR curve achieve a cross validated r2 /RMSE (in % woody cover) of 0.73/3.0 (MODIS) and 0.70/3.2 (VGT). The extrapolation to Sahel scale shows an almost nine times higher woody cover than in the global tree cover product MOD44B which only captures trees of a certain minimum size. The derived woody cover map of the Sahel is made publicly available and represents an improvement of existing products and a contribution for future studies of drylands quantifying carbon stocks, climate change assessment, as well as parametrization of vegetation dynamic models.


Download the woody cover map for Sahel here

Find the full article here

Special Issue: Remote Sensing of Land Degradation and Drivers of Change


Together with my colleagues Rasmus Fensholt, Stephanie Horion and Torbern Tagesson we edit a special issue for the open access journal Remote Sensing and we want to encourage everyone working with remote sensing and land degradation to submit a well prepared manuscript. The submission deadline is 31 May 2016. Find below the introduction text from the journal website:

Human and climate induced degradation of arable lands has been of major concern for livelihoods and food security particularly in drylands during recent decades, supporting and affecting the wellbeing of more than one-third of the global population. Monitoring vegetation productivity is of great importance because crop and livestock production is the most essential economic activity, especially in arid and semi-arid regions of the world.

The United Nations Convention to Combat Desertification’s (UNCCD) definition of desertification, or dryland degradation states that: “Land degradation in arid, semi-arid and dry sub-humid areas resulting from various factors, including climatic variations and human activities” followed by “land degradation means reduction or loss, in arid, semi-arid and dry sub-humid areas, of the biological or economic productivity and complexity of rainfed cropland, irrigated cropland, or range, pasture, forest and woodlands”…  (UNCCD homepage, www.unccd.int).

This definition implies that change in vegetation productivity is a key indicator (but not the only one) of land degradation. Along with land mismanagement often caused by human pressure, climatic variability is a major determinant of land degradation. Given the harsh nature of the climate in drylands, it is of great policy relevance to understand potential damaging interactions between land degradation and climate change. Indeed, climate-induced changes in air temperature and soil moisture might inflict soil erosion, salinization, crusting, and loss of soil fertility or depletion of seed banks in dryland ecosystems.

Continuous long-term Earth Observation (EO) satellite data provides the only suitable means of temporally and spatially consistent data analysis across multiple scales, and EO based metrics of vegetation productivity and land degradation are of great interest for the assessment and monitoring of environmental changes in dryland regions.

This forthcoming special issue welcomes research papers focusing on: (i) monitoring ecosystem productivity (both vegetation and economic productivity) and ecosystem complexity (i.e., biodiversity); (ii) studying the impact of climate change and human pressure on land degradation processes; (iii) uncovering the driving mechanisms of observed changes in vegetation productivity. The primary region of interest will be drylands, but studies covering other parts of the globe are also welcomed.


  • EO-based methods for monitoring land degradation;
  • Vegetation/climate/anthropogenic productivity indicators;
  • Human versus climate-induced land degradation;
  • Land-use land-cover change in monitoring land degradation;
  • Multi-temporal/time-series analysis/multiple datasets;
  • Local to global scales;
  • Drivers attribution;
  • Case studies on land degradation and climate change;
  • Field evidence of degradation linked with EO data.

Fodder Biomass Monitoring in Sahelian Rangelands Using Phenological Metrics from FAPAR Time Series

From: Diouf, A.A., Brandt, M., Verger, A., Jarroudi, M.E., Djaby, B., Fensholt, R., Ndione, J.A., Tychon, B., 2015. Fodder Biomass Monitoring in Sahelian Rangelands Using Phenological Metrics from FAPAR Time Series. Remote Sensing 7, 9122–9148. doi:10.3390/rs70709122

Livestock farming constitutes the most widespread human activity and the dominant land use in rangeland ecosystems. At a global scale, it contributes 40% of the agricultural gross domestic product, and provides income for more than 1.3 billion people and nourishment for at least 800 million food-insecure people. In particular for the West African Sahel, livestock constitutes the first renewable resource and is mainly characterized by an extensive use of pastures in rangelands.


Since 1987 the Centre de Suivi Ecologique (CSE) operationally estimates the total annual biomass in Senegal in order to monitor the fodder availability of the national pastoral rangelands. Field data is collected along 1 km transects at 24 sites at the end of the wet season. Here, herbaceous and woody leaf biomass is measured and summed to the total available fodder biomass. This is done each year since 1987.

cse2 cse1


Using a linear regression with satellite images, the field data is projected to whole Senegal and gives stakeholders an estimation on the quantity and distribution of fodder biomass. Between 1987 and 1999, this method was implemented using the seasonal integrated NDVI (i.e., seasonal weighted average) from the Advanced Very High Resolution Radiometer (AVHRR) of the National Oceanic and Atmospheric Administration (NOAA) satellites acquired in Local Area Coverage (LAC) format at the CSE receiving station in Dakar. Since 2000, the 1-km SPOT-VEGETATION NDVI have been used.

In this context, we developed a new operational system for monitoring total fodder biomass, including both herbaceous and woody leaf biomass. The proposed method is based on multiple linear regression models using phenological variables derived from the seasonal dynamics of the FAPAR SPOT-VEGETATION time series and ground measurements of total biomass production collected in different Sahelian ecosystems in Senegal over 15 years.


A model with three variables – large seasonal integral (LINTG), length of growing season and end of season decreasing rate – performed best (MAE = 605 kg DM/ha; R² = 0.68) across Sahelian ecosystems in Senegal (data for the period 1999-2013). A model with annual maximum (PEAK) and start date of season showed similar performances (MAE = 625 kg DM/ha; R² = 0.64), allowing a timely estimation of forage availability. The subdivision of the study area using metrics related to ecosystem properties increased overall accuracy (MAE = 489.21 kg DM/ha; R² = 0.77). LINTG was the main explanatory variable for woody rangelands, whereas for areas dominated by herbaceous vegetation it was the PEAK metric. The proposed approach outperformed the established single-predictor model (MAE = 818 kg DM/ha and R² = 0.51) and should improve the operational monitoring of forage resources in Sahelian rangelands.


In the future, such early warning models should enable stakeholders to take decisions as early as September (current year as biomass shortage) with regard to livestock by triggering protocols designed for livestock management (e.g., Opération de Sauvegarde du Bétail ) in Senegal.

see the full document here: MDPI

Text and Figures: A.A. Diouf; Fotos: M. Brandt

What Four Decades of Earth Observation Tell us about Land Degradation in the Sahel

From: Mbow, C.; Brandt, M.; Ouedraogo, I.; de Leeuw, J.; Marshall, M. What Four Decades of Earth Observation Tell Us about Land Degradation in the Sahel? Remote Sens. 2015, 7, 4048-4067.

Land degradation mechanisms are related to two main categories, one related to climate change and one associated with local human impact, mostly land use change such as expansion of cultivation, agricultural intensification, overgrazing and overuse of woody vegetation. Land degradation characteristics, triggers and human influence are manifold and interrelated. Some of the indicators can be monitored using Earth Observation techniques (underlined in red):


During the last four decades, the Sahel was affected by below-normal precipitation with two severe drought periods in 1972–73 and in 1983–84. Because of this negative climate trend, many studies prioritized the Sahel “crisis” in terms of productivity loss and land degradation. These negative perceptions have been opposed with recent findings of improved greenness mostly in relation to recent improvement in rainfall.


The assessment of land degradation and quantifying its effects on land productivity have been both a scientific and political challenge. After four decades of Earth Observation applications, little agreement has been gained on the magnitude and direction of land degradation in the Sahel. The number of Earth Observation datasets and methods, biophysical and social drivers and the complexity of interactions make it difficult to apply aggregated Earth Observation indices for these non-linear processes. Hence, while many studies stress that the Sahel is greening, others indicate no trend or browning. The different generations of satellite sensors, the granularity of studies, the study period, the applied indices and the assumptions and/or computational methods impact these trends.



While there is a clearly positive trend in biomass production at Sahel scale, a loss in biodiversity and locally encroaching barren land are observed at the same time. Multi-scale Earth Observation analyses show that neither the desertification nor the greening paradigms can be generalized, as both attempt to simplify a very complex reality. Heterogeneity is an issue of scale, and very coarse-scaled vegetation trend analyses reveal a greening Sahel. However, locally-scaled studies are not uniform, observing greening and degradation at the same time.

We suggest several improvements: (1) harmonize time-series data, (2) promote knowledge networks, (3) improve data-access, (4) fill data gaps, (5) agree on scales and assumptions, (6) set up a denser network of long-term fields-surveys and (7) consider local perceptions and social dynamics, as local people’s perception of land degradation/improvements often disagree with Earth Observation analyses.

Thus, to allow multiple perspectives and avoid erroneous interpretations caused by data quality/scale issues/generalizations, we recommend combining multiple data sources at multiple scales. Furthermore, we underline the relevance of field data and experience, and results achieved by remote sensing techniques should not be interpreted without contextual knowledge.

Download the full article here: Paper at MDPI

see also:

Knauer, K., Gessner, U., Dech, S., Kuenzer, C., 2014. Remote sensing of vegetation dynamics in West Africa. International Journal of Remote Sensing 35, 6357–6396. doi:10.1080/01431161.2014.954062

Be cautious using GIMMS3g v0 for trend analysis!

GIMMS3g NDVI is widely used to assess vegetation trends from local to global scale. And even though it is the best long term dataset available right now (July 2015), people should be aware that a serious error affects the dataset in semi arid areas, like the Sahel, which has severe impacts on trend analysis. The dry season values around 2004 suddenly drop, which is not happening in reality. It is definitely related to the sensor change from NOAA16 to NOAA17, and also the last change to NOAA18 has an impact on the time series. The good thing is that the wet season values should be usable. The developers are aware about this problem and a new version is on it’s way. To illustrate the issue, find attached the GIMMS3g NDVI (v0) averaged over the Sahel belt, note the drop at the sensor change (first blue line) and the recovery at the next sensor change:


To further illustrate how serious this problem affects trend analysis, the same area is shown in the VOD dataset, which is found to be reliable:vod

a rough overlay highlights the different directions of annual trends:


The problem can be solved by excluding the dry season and using the small integral, calculated in TIMESAT:


see further literature:

  • Tian, F., Fensholt, R., Verbesselt, J., Grogan, K., Horion, S., Wang, Y., 2015. Evaluating temporal consistency of long-term global NDVI datasets for trend analysis. Remote Sensing of Environment 163, 326–340. doi:10.1016/j.rse.2015.03.031
  • Horion, S., Fensholt, R., Tagesson, T., Ehammer, A., 2014. Using earth observation-based dry season NDVI trends for assessment of changes in tree cover in the Sahel. International Journal of Remote Sensing 35, 2493–2515. doi:10.1080/01431161.2014.883104
  • Pinzon, J.E., Tucker, C.J., 2014. A Non-Stationary 1981–2012 AVHRR NDVI3g Time Series. Remote Sensing 6, 6929–6960. doi:10.3390/rs6086929
  • Liu, Y.Y., van Dijk, A.I.J.M., McCabe, M.F., Evans, J.P., de Jeu, R.A.M., 2013. Global vegetation biomass change (1988–2008) and attribution to environmental and human drivers. Global Ecology and Biogeography 22, 692–705. doi:10.1111/geb.12024
  • Jonsson, P., Eklundh, L., 2004. TIMESAT–a program for analyzing time-series of satellite sensor data* 1. Computers & Geosciences 30, 833–845.

New project started: BICSA – Biophysical Changes in the Sahel

To continue my research about the Sahel area, I received a Marie Curie Individual Fellowships (Call: H2020-MSCA-IF-2014) for the coming 2 years. The new project which started on May the 1st is named BICSA – Biophysical Changes in the Sahel: Ground and Satellite Based Evidence Across Scales and Disciplines. This is the abstract of the successful application:

Human and climate induced desertification has been a major issue for livelihoods and food security in drylands. In this context, the Sahel has been subject to various controversial studies. Earth Observation (EO) studies show a positive trend in vegetation greenness over the last decades, which has been interpreted as an increase in biomass and contradicts prevailing narratives of widespread degradation. However, new scientific outcome suggests a massive loss in biodiversity, which again contradicts the beneficial effects of the greening theory. These apparent oppositions result from little investment that has been made in studying long-term ground data. Thus, the overall purpose of this project is to assess the opposing trends of biomass increase and species decline in the Sahel. By combining a range of long-term in-situ field data records (1980s-today) with EO time series and Very High Resolution (VHR) satellite imagery, an improved understanding on the role of trees, herbs and species on the greening Sahel will be achieved. Trends will be translated in ecosystem services and beneficial effects on livelihoods. Knowing the underlying biophysical mechanisms of the Sahel greening will resolve contradictions regarding the greening/desertification paradigms and thus be basis for future studies. Furthermore, the scientific understanding of linkages between ground and satellite data and their applicability across scales will be improved. New monitoring methods of biophysical variables address challenges in land management and food security. To achieve this, I will be trained in cutting edge skills (EO time series; object based mapping; field monitoring of vegetation productivity/biodiversity; socializing pixels; ecological services). Finally, this project will encourage a North-South collaboration in common scientific interest that is relevant for development and environmental research.

Ground- and satellite-based evidence of the biophysical mechanisms behind the greening Sahel

Making use of 27 years of ground measurements, we were able to find evidence of the role of trees and grass on the greening of the Senegalese Sahel. This was made possible by a close collaboration with our colleagues from the CSE, the Centre de Suivi Ecologique in Dakar. Moreover, woody species abundance data provided by Gray Tappan from 1983 shows changes in biodiversity over 30 years. We thus provide ground based evidences against the conventional view of irreversible degradation in the Sahel.


  • biodiversity;
  • biomass monitoring;
  • degradation;
  • greening;
  • Sahel;
  • vegetation change


After a dry period with prolonged droughts in the 1970s and 1980s, recent scientific outcome suggests that the decades of abnormally dry conditions in the Sahel have been reversed by positive anomalies in rainfall. Various remote sensing studies observed a positive trend in vegetation greenness over the last decades which is known as the re-greening of the Sahel. However, little investment has been made in including long-term ground-based data collections to evaluate and better understand the biophysical mechanisms behind these findings. Thus, deductions on a possible increment in biomass remain speculative. Our aim is to bridge these gaps and give specifics on the biophysical background factors of the re-greening Sahel. Therefore, a trend analysis was applied on long time series (1987–2013) of satellite-based vegetation and rainfall data, as well as on ground-observations of leaf biomass of woody species, herb biomass, and woody species abundance in different ecosystems located in the Sahel zone of Senegal. We found that the positive trend observed in satellite vegetation time series (+36%) is caused by an increment of in situ measured biomass (+34%), which is highly controlled by precipitation (+40%). Whereas herb biomass shows large inter-annual fluctuations rather than a clear trend, leaf biomass of woody species has doubled within 27 years (+103%). This increase in woody biomass did not reflect on biodiversity with 11 of 16 woody species declining in abundance over the period. We conclude that the observed greening in the Senegalese Sahel is primarily related to an increasing tree cover that caused satellite-driven vegetation indices to increase with rainfall reversal.

Brandt, M., Mbow, C., Diouf, A.A., Verger, A., Samimi, C. & R. Fensholt (2015) Ground and satellite based evidence of the biophysical mechanisms behind the greening Sahel. Global Change Biology.

Smoothing/Filtering a NDVI time series using a Savitzky Golay filter and R

Coarse scaled NDVI (or FAPAR, EVI…) images have a high temporal frequency and are delivered as Maximum Value Composites (MVC) of several days, which means the highest value is taken, assuming that clouds and other contaminated pixels have low values. However, especially in areas with a rainy season, the composites over 10-16 days still contain clouds and other disturbances. This figure illustrates a raw MODIS NDVI MVC of 16 days, and it’s obvious that several pixels are contaminated, and it’s also obvious that analyses will be affected by the noisy data.

unfiltered MODIS image

unfiltered MODIS image

Datasets like GIMMS or GEOV1 provide already filtered data, but e.g. MODIS, SPOT VGT, PROBA-V and AVHRR data are raw. The solution is to smooth the time series, using a filter, which calculates a smooth time series and interpolates the bad values using the previous and following images. Here is the same image, but smoothed with a Savitzky Golay filter.

filtered and smoothed MODIS image

filtered and smoothed MODIS image

The data are mostly delivered with quality rasters, rating the quality of each pixel. This can be used to either filter the raster and set bad quality pixels to  NA, or to weight the pixels. When the new time line is calculated, low weighted (i.e. contaminated) pixels are less considered in the calculation process. One possibility is the software TIMESAT (Eklundh & Jönsson, 2011), which offers different filter techniques. Here is an example how timesat smoothes the time line with a Savitzky Golay filter, omitting “bad” pixles and creating new rasters.

A Savitzky Golay filter applied on MODIS using TIMESAT

A Savitzky Golay filter applied on MODIS using TIMESAT

Filtering is also possible in R, and it’s very simple. First one has to decide if one wants to work with quality files, or simply use the raw data, both is possible. GIS software like GRASS has modules which allow an easy use of the quality files:



However, filtering without quality flags also provides reasonable results. Now we assume we have a time series of MODIS data from 2005-2010, with 23 images each year. This data is loaded in R in a raster stack or brick, called MODIS, bad values are masked as NA. We load the libraries, and create a function which uses the Savitzky Golay filter from the signal package. The parameters of the function need to be adapted (p, n, ts) (http://www.inside-r.org/packages/cran/signal/docs/sgolayfilt), also the time frame.



fun <- function(x) {
z=substituteNA(v, type=”mean”)
MODIS.ts2 = ts(z, start=c(2005,1), end=c(2010,23), frequency=23)
x=sgolayfilt(MODIS.ts2, p=1, n=3, ts=30)

MODIS.filtered <- calc(MODIS, fun)

MODIS.filtered is a new brick containing the smoothed time series. Compare the raw with the filtered tims series:

l=cellStats(MODIS, stat=mean)
MODIS.raw.ts = ts(l, start=c(2005,1), end=c(2010,23), frequency=23)
l=cellStats(MODIS.filtered, stat=mean)
MODIS.filt.ts = ts(l, start=c(2005,1), end=c(2010,23), frequency=23)

One may find out the perfect fitting parameters by looking at the whole area, playing with the parameters:

l=cellStats(MODIS, stat=’mean’)
MODIS.ts = ts(l, start=2005, end=c(2010,23), frequency=23)
sg=sgolayfilt(MODIS.ts, p=3, n=9, ts=20)
sg.ts = ts(sg, start=2005, end=c(2010,23), frequency=23)


Eklundh, L.; Jönsson, P. Timesat 3.1 Software Manual. 2011.

Woody vegetation and land cover changes in the Sahel of Mali (1967–2011)

Another very interesting publication using object based methods to detect single trees on very high resolution imagery is online.

Raphael Spiekermann, Martin Brandt, Cyrus Samimi, Woody vegetation and land cover changes inthe Sahel of Mali (1967–2011), International Journal of Applied Earth Observation and Geoinformation, Volume 34, February 2015, Pages 113-121.

It can be downloaded for free until late October using this link: http://authors.elsevier.com/a/1PeJA14ynR~DWs 


  • Woody cover, species and land cover change over 44 years are analyzed.
  • Object-based classifications are applied with high resolution images of 1967 and 2011.
  • Climate and especially human impact have caused extensive changes.
  • Changes are not always negative and a variety of spatial variations are shown.


In the past 50 years, the Sahel has experienced significant tree- and land cover changes accelerated by human expansion and prolonged droughts during the 1970s and 1980s. This study uses remote sensing techniques, supplemented by ground-truth data to compare pre-drought woody vegetation and land cover with the situation in 2011. High resolution panchromatic Corona imagery of 1967 and multi-spectral RapidEye imagery of 2011 form the basis of this regional scaled study, which is focused on the Dogon Plateau and the Seno Plain in the Sahel zone of Mali. Object-based feature extraction and classifications are used to analyze the datasets and map land cover and woody vegetation changes over 44 years. Interviews add information about changes in species compositions. Results show a significant increase of cultivated land, a reduction of dense natural vegetation as well as an increase of trees on farmer’s fields. Mean woody cover decreased in the plains (−4%) but is stable on the plateau (+1%) although stark spatial discrepancies exist. Species decline and encroachment of degraded land are observed. However, the direction of change is not always negative and a variety of spatial variations are shown. Although the impact of climate is obvious, we demonstrate that anthropogenic activities have been the main drivers of change.