About Martin Brandt

My name is Martin Brandt and I’m a physical geographer with focus on remote sensing and environmental change. My regional focus is in the Sahel of Western Africa, especially Mali and Senegal, where I’ve spent a lot of time doing field work. I’m working with every kind of data, from high resolution RapidEye to coarse scale time series. All processing is done using free and open software, escecially GRASS GIS, QGIS, and R in a Linux environment. Currently I’m working in the micle project which aims to find linkages between climate, evironment and migration in the West African Sahel. With this blog, I want to share results, but also simple methods on data handling. Beside producing interesting maps in the office, I’m especially interested in explaining remote results locally on the ground, which includes vegetation surveys, but also “socializing pixels” by intensive work with humans living and acting within the pixel.

How does conflict affect land use? New publication!

Population and Environment in the Middle East

In November 2015, me and a colleague (Michael Degerald, visit his blog here) asked the question: how is agriculture affected in the areas seized by the Islamic State (aka ISIS, ISIL, Da’esh)? We couldn’t find much information to answer our question, so we decided to investigate it ourselves.

At first we wanted to look at changes in productivity indicated by satellite measured greenness, but later we decided to go a step deeper and look at land use activity as an indicator of land abandonment (as I had done in a previous publication). As the project moved on, more people became interested, and eventually three more co-authors were added: Petter Pilesjö (Lund University), Martin Brandt and Alexander Prishcepov (both from Copenhagen University).

Together, we conducted a land use classification based on NDVI data from MODIS based on the seasonality of the land surface. We distinguished between single cropped cropland…

View original post 725 more words

Advertisements

Expanding the study area: from West Africa to South China

People who know me and my research know that I am in love with the Sahel and its people, and this will never change. However, I recently had the chance to expand my research area from the semi-arid Sahel to humid China, more specifically the South China Karst. I spent around 4 months in China in 2017, and there are some major publications on the way. This area is particularly interesting, because millions of trees have been planted, and while we have a hard time to find a human footprint in satellite data over Sahel, it is more than obvious in China. It is also a very beautiful area:

Our research in the media

Our article in Nature Ecology & Evolution got some media attention, both in the Danish and the international press.

Here is our article:

Brandt, M.; Rasmussen, K.; Peñuelas, J.; Tian, F.; Schurgers, G.; Verger, A.; Mertz, O.; Palmer, J. R. B.; Fensholt, R. Human population growth offsets climate-driven increase in woody vegetation in sub-Saharan Africa. Nature Ecology & Evolution 2017, 1, 0081.

New publications!

woody-plants-workshop-in-copenhagen19_01_2017_group-photo-2

Our fantastic team had some new publications within the past half year, they are all worth to have a look:

This one deals with the question if agricultural intensification in Sahel causes an increase or decrease in NDVI trends. Surprisingly, we find a negative NDVI trend coupled with an increase in cropped areas which means that fallowed fields have a substantially higher NDVI than cropped fields.

Open access! Here we use great data sources to document dynamics in woody vegetation in central Senegal. Field data from 2000 to 2015, fantastic aerial photos from 1994, repeat photography from 1994 and 2015, satellite imagery at 50 cm resolution from 2005-2015, and finally MODIS time series. We find a high spatial and temporal dynamic, encroachment, die off, etc. It’s a very a colourfully illustrated study which will make you feel like travelling to Senegal..

A great story as well: We document how conservation projects in Southern China are able to impact on vegetation trends and propose an index which allows to put the invested money (for conservation projects) in relation with vegetation trends to be able to determine the project effectiveness.

A very clever way to combine optical and passive microwave satellite data: We assume that optical satellite data senses the green part of the vegetation and the passive microwaves the green plus non-green parts. So we combine both to estimate the non green vegetation (i.e. the wood) and look at global trends from 2000 to 2012 which allows us to map gradual gains and losses in woody cover.

Publishing in the open access journal Remote Sensing (MDPI)

remotesensing-logo

With our last article published, we are closing the special issue on land degradation for the open access journal Remote Sensing and I want to share some experiences here.

In total, 24 articles were submitted, 13 of them were published, 6 rejected without going to review, and 5 were rejected after review.

So we have an acceptance rate of 54%. Interestingly, this is very close to the 2013 statistics for all submissions, and it also reflects the overall quality of the submissions, which is average. The quality of the articles that were published in the end is ok, some are good, but rather not exceptional.

The interaction between us and the MDPI editorial staff was professional, smooth and efficient. Everything was prepared nicely for us and we could concentrate on the scientific part without any  managing aspects.

Having now 7 articles published in this journal (2 as first author) within the past 3 years, I can fully recommend publishing in Remote Sensing. Yes, the quality of the articles can not be compared with the leading journal “Remote Sensing of Environment” (here we have 4 articles now published within the past 2 years), and it is for sure easier to publish in Remote Sensing (RS) with less critical editors and reviewers. However, if you have an overall good quality article (not exceptional), there are several reasons for going for RS instead for the armada of Elsevier and Springer journals:

  • Open access: research should be available to everyone and not limited to rich countries and rich universities. Remote Sensing of Environment for example is not available at my former university, and in many German universities Elsevier journals are generally unavailable. Buying open access in these journals is possible but too expensive. Thousands of academics boycott Elsevier.
  • The authors of the article keep the rights on their research and are able to distribute their work freely.
  • Rapid processing, most of our 7 articles were published after around 2 months. The main reason is the professional editorial staff who do this work as full time job.
  • The articles are downloaded thousands of times and reach a wide audience.

One may argue the large number of average quality articles being published (being an online only journal, there are no issues and thus no article limit) swamps the scientific market and reduces importance of individual scientific work, but this is a general problem of science these days. One may also argue that the publisher MDPI is a company making money with each article they publish (and there is no limitation), so their aim is probably to publish as many articles as possible, and this is not beneficial for being critical. This might be true, however, in the end it’s up to the academic editors and the reviewers to decide if an article is published, not the company, and even the Nature and Science groups have their own mass publishing journals (Scientific Reports, Science Advances) now. Scientific publishing is about making profit.

Many people think that open access journals like RS are commercial companies making money (“you pay to get your paper published”), whereas articles published in Elsevier & co are non-commercial and real science. Here one should not forget that companies like Elsevier make billions of $$ profit each year, of which the reviewers see nothing and the editors do it as free time job being poorly paid. The universities pay absurd sums to make the articles available for their students, but many universities can not, and do not want to support this any more, but rather support the open access publishing by paying the publishing waves. In the end, this is much cheaper for the university and the article is freely available for everyone.

My personal recommendation: If you think you have an exceptional article dealing with remote sensing, there is no way around Remote Sensing of Environment, the reputation of this journal is untouchable. However, not every study we do has outstanding results, so if you do not want to wait more than a half year for a likely rejection, I personally can fully recommend Remote Sensing, the processing is rapid but still professional.

Brandt, M.; Tappan, G.; Diouf, A.A.; Beye, G.; Mbow, C.; Fensholt, R. Woody Vegetation Die off and Regeneration in Response to Rainfall Variability in the West African Sahel. Remote Sens. 2017, 9, 39.

Integrating meteorological data in biomass prediction models

Diouf, A.A.; Hiernaux, P.; Brandt, M.; Faye, G.; Djaby, B.; Diop, M.B.; Ndione, J.A.; Tychon, B. Do Agrometeorological Data Improve Optical Satellite-Based Estimations of the Herbaceous Yield in Sahelian Semi-Arid Ecosystems? Remote Sens. 2016, 8, 668.

Quantitative estimates of forage availability at the end of the growing season in rangelands are helpful for pastoral livestock managers and for local, national and regional stakeholders in natural resource management. For this reason, remote sensing data such as the Fraction of Absorbed Photosynthetically Active Radiation (FAPAR) have been widely used to assess Sahelian plant productivity for about 40 years.

This study combines traditional FAPAR-based assessments with agrometeorological variables computed by the geospatial water balance program, GeoWRSI, using rainfall and potential evapotranspiration satellite gridded data to estimate the annual herbaceous yield in the semi-arid areas of Senegal.

It showed that a machine-learning model combining FAPAR seasonal metrics with various agrometeorological data provided better estimations of the in situ annual herbaceous yield (R2 = 0.69; RMSE = 483 kg·DM/ha) than models based exclusively on FAPAR metrics (R2 = 0.63; RMSE = 550 kg·DM/ha) or agrometeorological variables (R2 = 0.55; RMSE = 585 kg·DM/ha). All the models provided reasonable outputs and showed a decrease in the mean annual yield with increasing latitude, together with an increase in relative inter-annual variation. In particular, the additional use of agrometeorological information mitigated the saturation effects that characterize the plant indices of areas with high plant productivity.

The date of the onset of the growing season derived from smoothed FAPAR seasonal dynamics showed no significant relationship (0.05 p-level) with the annual herbaceous yield across the whole studied area. The date of the onset of rainfall was significantly related to the herbaceous yield and its inclusion in fodder biomass models could constitute a significant improvement in forecasting risks of a mass herbaceous deficit at an early stage of the year.

remotesensing-08-00668-ag

New publications

Check out our new publications:

This paper raises awareness of data scale issues in environment migration research, for example the declining station network used in the CRU (3.2) dataset:

test

With a case study of a Senegalese village we examine how (and if) migration patterns are linked with climate and environmental changes.Screenshot from 2016-07-14 14-19-10.png

This paper uses LASSO and Random Forest to select the best variables from a huge pool of indices to predict wood volume in Tadjikistan. Red Edge bands show to be of high value.

Screenshot from 2016-07-14 14-27-41.png

Downloading and processing MODIS with fast and free tools

Here’s a simple way of processing MODIS: downloading, extracting the bands from HDF as GEOTIFF, merging tiles and reprojecting. Area extraction, NDVI calculation, cloud filtering and more can be done in GRASS for example. It’s Linux based, but can easily used in Windows. My way might not be the most elegant, but it’s fast and works fine.

The first step is to select the tiles, the time frame and product at http://reverb.echo.nasa.gov/reverb/. I take MOD09Q1 (250 m, 8 day, collection 6 reflectance) as example. Once orderd, we get a TXT file containing all file names. We download all files using wget. After download i mass rename all files to remove the creation date (using e.g. total commander) so the files are named like: MOD09Q1.A2000065.h21v05.006.

Then I extract Band 1, Band 2 and the Quality flag as GeoTiff. “aus” is the folder were the TIFFS will end up. Find out the name of the bands using gdalinfo.

wget -i modis.txt
aus="/media/data/MODIS/tiffs/"
for i in *.hdf; do gdal_translate -of GTiff 'HDF4_EOS:EOS_GRID:"'$i'":MOD_Grid_250m_Surface_Reflectance:sur_refl_b01' $aus`basename $i .hdf`_REF_band1.tif; done
for i in *.hdf; do gdal_translate -of GTiff 'HDF4_EOS:EOS_GRID:"'$i'":MOD_Grid_250m_Surface_Reflectance:sur_refl_b02' $aus`basename $i .hdf`_REF_band2.tif; done
for i in *.hdf; do gdal_translate -of GTiff 'HDF4_EOS:EOS_GRID:"'$i'":MOD_Grid_250m_Surface_Reflectance:sur_refl_qc_250m' $aus`basename $i .hdf`_QA.tif; done

I did not find a clean way for merging, so i suggest my “dirty” way, it needs to be repeated for each year and band, this is 5 minutes of copying but you can use it again and again and again for the next years then….here’s an example for band 1 year 2000. If you use a 16 day product every second line will give an error, but that you can ignore.

# 2000
gdal_merge.py -o 2000001_band1.tif *2000001*
gdal_merge.py -o 2000009_band1.tif *2000009*
gdal_merge.py -o 2000017_band1.tif *2000017*
gdal_merge.py -o 2000025_band1.tif *2000025*
gdal_merge.py -o 2000033_band1.tif *2000033*
gdal_merge.py -o 2000041_band1.tif *2000041*
gdal_merge.py -o 2000049_band1.tif *2000049*
gdal_merge.py -o 2000057_band1.tif *2000057*
gdal_merge.py -o 2000065_band1.tif *2000065*
gdal_merge.py -o 2000073_band1.tif *2000073*
gdal_merge.py -o 2000081_band1.tif *2000081*
gdal_merge.py -o 2000089_band1.tif *2000089*
gdal_merge.py -o 2000097_band1.tif *2000097*
gdal_merge.py -o 2000105_band1.tif *2000105*
gdal_merge.py -o 2000113_band1.tif *2000113*
gdal_merge.py -o 2000121_band1.tif *2000121*
gdal_merge.py -o 2000129_band1.tif *2000129*
gdal_merge.py -o 2000137_band1.tif *2000137*
gdal_merge.py -o 2000145_band1.tif *2000145*
gdal_merge.py -o 2000153_band1.tif *2000153*
gdal_merge.py -o 2000161_band1.tif *2000161*
gdal_merge.py -o 2000169_band1.tif *2000169*
gdal_merge.py -o 2000177_band1.tif *2000177*
gdal_merge.py -o 2000185_band1.tif *2000185*
gdal_merge.py -o 2000193_band1.tif *2000193*
gdal_merge.py -o 2000201_band1.tif *2000201*
gdal_merge.py -o 2000209_band1.tif *2000209*
gdal_merge.py -o 2000217_band1.tif *2000217*
gdal_merge.py -o 2000225_band1.tif *2000225*
gdal_merge.py -o 2000233_band1.tif *2000233*
gdal_merge.py -o 2000241_band1.tif *2000241*
gdal_merge.py -o 2000249_band1.tif *2000249*
gdal_merge.py -o 2000257_band1.tif *2000257*
gdal_merge.py -o 2000265_band1.tif *2000265*
gdal_merge.py -o 2000273_band1.tif *2000273*
gdal_merge.py -o 2000281_band1.tif *2000281*
gdal_merge.py -o 2000289_band1.tif *2000289*
gdal_merge.py -o 2000297_band1.tif *2000297*
gdal_merge.py -o 2000305_band1.tif *2000305*
gdal_merge.py -o 2000313_band1.tif *2000313*
gdal_merge.py -o 2000321_band1.tif *2000321*
gdal_merge.py -o 2000329_band1.tif *2000329*
gdal_merge.py -o 2000337_band1.tif *2000337*
gdal_merge.py -o 2000345_band1.tif *2000345*
gdal_merge.py -o 2000353_band1.tif *2000353*
gdal_merge.py -o 2000361_band1.tif *2000361*

# We can remove the other files now 
rm MOD*
# Then reproject everything to a geographic system, maybe into a new folder:
mkdir lat
aus="/media/data/MODIS/tiffs/lat/"

for i in *_band1.tif; do gdalwarp -t_srs EPSG:4326 $i $aus`basename $i .tif`.tif; done
for i in *_band2.tif; do gdalwarp -t_srs EPSG:4326 $i $aus`basename $i .tif`.tif; done
for i in ???????_QA.tif; do gdalwarp -t_srs EPSG:4326 $i $aus`basename $i .tif`.tif; done

What we have now can already be used, but you can continue in GRASS to calculate NDVI and filter the data.

 

Recent woody vegetation trends in Sahel

Our new paper looks at recent dynamics in woody vegetation in Sahel and finds some interesting patterns which are mainly controlled by human population density.

Martin Brandt, Pierre Hiernaux, Kjeld Rasmussen, Cheikh Mbow, Laurent Kergoat, Torbern Tagesson, Yahaya Ibrahim, Abdoulaye Wele, Compton J. Tucker, Rasmus Fensholt. Assessing woody vegetation trends in Sahelian drylands using MODIS based seasonal metrics. Remote Sensing of Environment, 2016, 183, 215-225.

  • Woody cover trends are estimated for Sahel based on MODIS dry season metrics.
  • Interannual fluctuations in foliage density are attenuated to monitor woody plant trends.
  • Increases (decreases) are seen in areas of low (high) human population.
  • Recent decreases only partially offset a general post-drought increase in Sahelian woody cover.

Woody plants play a major role for the resilience of drylands and in peoples’ livelihoods. However, due to their scattered distribution, quantifying and monitoring woody cover over space and time is challenging. We develop a phenology driven model and train/validate MODIS (MCD43A4, 500 m) derived metrics with 178 ground observations from Niger, Senegal and Mali to estimate woody cover trends from 2000 to 2014 over the entire Sahel at 500 m scale.

Over the 15 year period we observed an average increase of 1.7 (± 5.0) woody cover (%) with large spatial differences: No clear change can be observed in densely populated areas (0.2 ± 4.2), whereas a positive change is seen in sparsely populated areas (2.1 ± 5.2). Woody cover is generally stable in cropland areas (0.9 ± 4.6), reflecting the protective management of parkland trees by the farmers. Positive changes are observed in savannas (2.5 ± 5.4) and woodland areas (3.9 ± 7.3).

The major pattern of woody cover change reveals strong increases in the sparsely populated Sahel zones of eastern Senegal, western Mali and central Chad, but a decreasing trend is observed in the densely populated western parts of Senegal, northern Nigeria, Sudan and southwestern Niger. This decrease is often local and limited to woodlands, being an indication of ongoing expansion of cultivated areas and selective logging.

We show that an overall positive trend is found in areas of low anthropogenic pressure demonstrating the potential of these ecosystems to provide services such as carbon storage, if not over-utilized. Taken together, our results provide an unprecedented synthesis of woody cover dynamics in the Sahel, and point to land use and human population density as important drivers, however only partially and locally offsetting a general post-drought increase.

graph_abstract

Assessing Future Vegetation Trends and Restoration Prospects in the Karst Regions of Southwest China

Our latest article is not located in the Sahel, however, the method to assess the future persistence of vegetation trends is highly interesting in the context of ecosystem stability and resistance. The article is open access and freely available.

Tong, Xiaowei; Wang, Kelin; Brandt, Martin; Yue, Yuemin; Liao, Chujie; Fensholt, Rasmus. 2016. “Assessing Future Vegetation Trends and Restoration Prospects in the Karst Regions of Southwest China.” Remote Sens. 8, no. 5: 357.

 

To alleviate the severe rocky desertification and improve the ecological conditions in Southwest China, the national and local Chinese governments have implemented a series of Ecological Restoration Projects since the late 1990s. In this context, remote sensing can be a valuable tool for conservation management by monitoring vegetation dynamics, projecting the persistence of vegetation trends and identifying areas of interest for upcoming restoration measures.

In this study, we use MODIS satellite time series (2001–2013) and the Hurst exponent to classify the study area (Guizhou and Guangxi Provinces) according to the persistence of future vegetation trends (positive, anti-persistent positive, negative, anti-persistent negative, stable or uncertain). The persistence of trends is interrelated with terrain conditions (elevation and slope angle) and results in an index providing information on the restoration prospects and associated uncertainty of different terrain classes found in the study area.

The results show that 69% of the observed trends are persistent beyond 2013, with 57% being stable, 10% positive, 5% anti-persistent positive, 3% negative, 1% anti-persistent negative and 24% uncertain. Most negative development is found in areas of high anthropogenic influence (low elevation and slope), as compared to areas of rough terrain. We further show that the uncertainty increases with the elevation and slope angle, and areas characterized by both high elevation and slope angle need special attention to prevent degradation. Whereas areas with a low elevation and slope angle appear to be less susceptible and relevant for restoration efforts (also having a high uncertainty), we identify large areas of medium elevation and slope where positive future trends are likely to happen if adequate measures are utilized.

The proposed framework of this analysis has been proven to work well for assessing restoration prospects in the study area, and due to the generic design, the method is expected to be applicable for other areas of complex landscapes in the world to explore future trends of vegetation.

graphical_abstract

The Hurst exponent, a measure of the persistence of a time series, can be easily calculated on a raster time series in R:

library(raster)
library(rgdal)
library(pracma)

# set working directory with the raster files, here in TIF format, 
# and load the files in a rasterbrick

setwd("/media/2016_Xiaowei/")
gsn = brick(list.files(pattern='*.tif'))
# test on the average of the study area

g=cellStats(gsn, stat='mean')
gsn.ts = ts(g, start=c(2001,1), end=c(2013,1), frequency=1)
plot(gsn.ts)

h=hurstexp(gsn.ts)$Hs
h

# run the function on the rasterbrick gsn

fun=function(x) { 
  v=as.vector(x)
  if (is.na(v[1])){ NA } else
  gsn.ts = ts(v, start=c(2001,1), end=c(2013,1), frequency=1)
  x=hurstexp(gsn.ts, display=F)$Hs 
}
h <- calc(gsn, fun)
plot(h)