REMOTELY SENSED INFORMATION FOR CROP MONITORING AND FOOD SECURITY

Transcription

1 REMOTELY SENSED INFORMATION FOR CROP LEARNING OBJECTIVES Lesson 2 Commonly Used Remote Sensing Data At the end of the lesson, you will be able to: recognize the types of data most commonly used in remote sensing analysis of crop monitoring; and list a number of institutions making available remote sensing data for crop monitoring. 1 of 65

2 INTRODUCTION This lesson presents the most important data commonly used for remote sensing analysis of crop monitoring. In particular, this lesson will illustrate through concrete examples the need and the potential uses for those kind of data. Furthermore, this lessons will provide some links to several data repositories. 2 of 65

3 USING SATELLITE IMAGES FOR VEGETATION MONITORING Satellite images can play a role in providing: 1. direct information about crop stage, conditions and health by observing the spectral properties of green vegetation and analysing their relationship with biomass/yield; and 2. estimates of the meteorological variables that drive crop development such as temperatures and rainfall, in those areas which are not sufficiently covered by meteorological stations. 3 of 65

4 Introduction: Crop monitoring using remote sensing Spectral response of green & dry, vegetation, bare soil, water -> spectral signature -> Interest to observe vegetation from distance (remote sensing / télédétection) -> using satellites, airplanes, now unmanned airborne vehicles (depending on the revisit frequency and resolution needed) Let us look at the satellites used for crop monitoring 4 of 65

5 USING SATELLITE IMAGES FOR VEGETATION MONITORING Let s see a short history of the use of satellite imagery for monitoring crop development. 1. spectral properties 70s - Landsat 80s - AVHHR NOAA 90s - VEGETATION SPOT 2000s MODIS/MERIS 5 of 65

6 Geostationary satellite Reminder on some satellite characteristics Orbit: geostationary versus near-polar, sun synchronous (the satellite covers a given area of the world at a constant local solar time) Orbit cycle: time for the satellite to pass over the same point on the Earth's surface directly below the satellite (called the nadir point) Swath: portion of the Earth s surface seen by the satellite -> determines the revisit period Polar satellite Swath of a Polar sensor

7 Adjacent swaths (ascending orbits of Landsat8 in 1 day, swath width 185 km)

8 Adjacent swaths (VEGETATION) S1- product (daily synthesis), swath width 2200 km

9 Adjacent swaths (VEGETATION) S10-product (10-daily synthesis)

10 Characteristics of images Pixel representing brightness with a digital number e.g Electromagnetic energy may be detected: photographically (chemical reactions on the surface of light-sensitive film to detect and record energy variations) electronically / digital (subdividing the image into small equal-sized and shaped areas, called picture elements or pixels, and representing the brightness of each area with a numeric value or digital number)

11 Fine and coarse resolution Images where only large features are visible are said to have coarse or low resolution whereas images with fine or high resolution allow detecting smaller objects. Generally speaking, the finer the resolution, the smaller the image frame / area seen. There is a trade-off between coverage and detail / data volume (Earth s surface = km 2 -> 500 MB/band with 1 km data and 1 byte/pixel; volume x with 10m data, with 1m data )

12 USING SATELLITE IMAGES FOR VEGETATION MONITORING 70s - Landsat 80s - AVHHR NOAA 90s - VEGETATION SPOT 2000s MODIS/MERIS A Landsat full image, from October 1972, shown as a false color composite. The use of spectral data was studied extensively by using satellite imagery after the launch of the first civil Earth observation satellite (Landsat1) in of 65

13 USING SATELLITE IMAGES FOR VEGETATION MONITORING Popup Window Landsat Landsat imagery is available since 1972 from seven satellites with 4-5 sensors evolving over thirty years: MSS (Multi-Spectral Scanner, 4 bands) on Landsat 1-3 ( ), TM (Thematic Mapper, 7 bands) on Landsat 4 & 5 ( ), ETM+ (Enhanced Thematic Mapper Plus, 8 bands) on Landsat 7 (1999-), OLI (Operational Land Imager, 9 bands) & TIRS (Thermal InfraRed Sensor, 2 bands) on LDMC/Landsat 8 (2013-). Landsat supplies high resolution visible and infrared imagery (MSS: 79m, TM 30m) with thermal imagery (120m on L4 & 5, 100m on L7, 60m on L8) and a panchromatic image (15m) also available from the ETM+ and OLI sensors. (Source: Landsat-7 on orbit. For more information visit: 13 of 65

14 USING SATELLITE IMAGES FOR VEGETATION MONITORING 70s - Landsat 80s - AVHHR NOAA 90s - VEGETATION SPOT 2000s MODIS/MERIS It is only with the growing availability of low resolution multi-spectral satellite images from the meteorological satellite series NOAA AVHRR in the early 80 s, that similar analyses are extended to large areas, including many countries in arid and semiarid climates. Image of a tornado outbreak taken by the NOAA AVHRR sensor. 14 of 65

15 USING SATELLITE IMAGES FOR VEGETATION MONITORING Popup Window NOAA AVHRR An artist's rendering of the NOAA POES. AVHRR (Advanced Very High Resolution Radiometer) is a broad-band scanner, sensing in the visible (red), near-infrared, mid infrared and thermal infrared portions of the electromagnetic spectrum, carried on NOAA s (National Oceanic Atmospheric Administration) Polar Orbiting Environmental Satellites (POES) and more recently also on the European MetOp satellite. 15 of 65

16 USING SATELLITE IMAGES FOR VEGETATION MONITORING 70s - Landsat 80s - AVHHR NOAA 90s - VEGETATION SPOT 2000s MODIS/MERIS Most of the early studies (e.g. from the80 s and the 90 s) relate to the use of different sensors of the NOAA AVHRR series. These images were typically available at the national and multinational level with a 1km resolution (LAC) and at the continental and global level with a 4,6 km resolution (GAC) or below. It is only at the end of the 90 s that the French-Belgian-Swedish satellite SPOT is equipped with a 1km resolution sensor (with B, R, NIR and SWIR bands) for vegetation monitoring at global scale called VEGETATION. Athens (Greece) as seen by the Spot-5 satellite in of 65

17 USING SATELLITE IMAGES FOR VEGETATION MONITORING Popup Window SPOT SATELLITE The SPOT (Satellites Pour l Observation de la Terre or Earth-observing Satellites) is a high-resolution (10-20m resolution, 60 km swath), optical imaging Earth observation satellite system operating from space, which was set up in 1978 by France in partnership with Belgium and Sweden. The SPOT system includes a series of satellites and ground control resources for satellite control and programming, image production, and distribution. The company SPOT Image is marketing the highresolution images, which SPOT can take from every corner of the Earth. Spot-5 Satellite. The first VEGETATION sensor (1km resolution, 2200 km swath) was launched on SPOT 4 in April of 65

18 USING SATELLITE IMAGES FOR VEGETATION MONITORING 80s - AVHHR NOAA 70s - Landsat 90s - VEGETATION SPOT 2000s MODIS/MERIS Several so called medium resolution sensors (maximum m) are operational since the year Amongst the best known are the MODIS and MERIS sensors belonging to the TERRA/AQUA and ENVISAT platforms respectively. Envisat view of Europe as a whole, using the wide angle capability of MERIS. 18 of 65

19 USING SATELLITE IMAGES FOR VEGETATION MONITORING Popup Window MODIS and MERIS MODIS and MERIS sensors can potentially observe the Earth s surface in 1-3 days (similar to the 1 day of NOAA/AVHHR and SPOT VEGETATION), but their spatial resolution increases from 1km to 250 m (MODIS) and 300 m (MERIS) respectively. The higher spatial resolution means that they include more topographic detail on ground features and for example rivers, small lakes, or irrigated areas are much better visible than with NOAA AVHRR or SPOT VEGETATION. On the other hand the data volume as compared to the other sensors increases by a factor 16 (ca. 4*4 pixels corresponding to each 1km pixel). In practice they provide spatially more detailed information and are typically used at national or regional scale, while NOAA AVHRR and SPOT VEGETATION images remain the most common at continental and global level. 19 of 65

20 Popup Window Terra and Aqua satellites Terra and Aqua are multi-national NASA scientific research satellite. Terra was launched on December It is the flagship of the Earth Observing System (EOS). The name "Terra" comes from the Latin word for Earth. Terra crosses the equator at 10:30 am. Aqua was launched on 2002 and studies the precipitation, evaporation, and cycling of water. The name "Aqua" comes from the Latin word for water. Acqua crosses the equator at 1:30 pm. Artist's rendering of the Terra spacecraft. Aqua satellite and its instruments. 20 of 65

21 USING SATELLITE IMAGES FOR VEGETATION MONITORING Popup Window Envisat Envisat ("Environmental Satellite") was a large multi-sensor Earth-observing satellite which was launched on 2002 and was operational until April This European Space Agency (ESA) satellite is the most advanced environmental spacecraft ever built and the largest civilian Earth observation satellite put into space. Its objective was to service the continuity of European Remote- Sensing Satellite missions, providing additional observational parameters to improve environmental studies. Model of Envisat in original size. 21 of 65

22 USING SATELLITE IMAGES FOR VEGETATION MONITORING This table resumes the properties of the most common sensors used for vegetation monitoring. 1. spectral properties Properties of the most common optical low and medium resolution sensors Sensor Platform Spectral range Number of bands Resolution Swath width Repeat coverage Launch AVHHR AVHRR NOAA POES 6-19 METOP VIS, NIR, MIR, TIR VIS, NIR, MIR, TIR SEAWIFS Orbview-2 VIS, NIR 8 VEGETATION SPOT 4, 5 MODIS EOS AM1/PM1 VIS, NIR, SWIR VIS, NIR, SWIR, TIR MERIS ENVISAT VIS, NIR 15 PROBA-V PROBA-V SENTINEL 3 SENTINEL VIS, NIR, SWIR VIS, NIR, SWIR m 2400km 12 hours m 2400km 12 hours m 4500m 1500km 2800km 1day m 2200km 1day m 2330km <2days m (1200m) 300m (1000m) 1150km <3days km 1 day m 1270km <2 days Foreseen of 65

23 USING SATELLITE IMAGES FOR VEGETATION MONITORING Let s consider the strengths and drawbacks of the low resolution systems 1. spectral properties Strengths Thanks to their large swath width, low resolution systems have a much better synoptic view and temporal revisit frequency compared to high resolution sensors. The individual scenes span a width of up to 3000 km, such that the entire Earth surface is scanned every day and the specific costs per ground area unit are very low. All the low and medium resolution sensors that have proven their validity for land surface observation and vegetation analysis do normally find their applications also in agriculture. Weaknesses The low spatial resolution is an intrinsic drawback for the sensors, with pixel sizes of about 1 km², i.e. far above typical field sizes. As a consequence, recorded spectral radiances are mostly mixed information from several surface types. This seriously complicates the interpretation (and validation) of the signal as well as the reliability of the derived information products. However, for crop monitoring, early warning and yield forecasting at the national scale a 1 km² resolution is quite suitable. 23 of 65

24 USING SATELLITE IMAGES FOR VEGETATION MONITORING Please indicate whether the following statements about low resolution systems characteristics are true or false. True False 1. One of the intrinsic drawbacks of low spatial resolution is the small swath width. 2. For crop monitoring at the national scale a 1km 2 resolution is an adequate pixel size. 3. In low resolution systems specific costs per ground area unit are quite high. Please select the answers of your choice. When you have finished, click on "Check Answer". 24 of 65

25 USING SATELLITE IMAGES FOR VEGETATION MONITORING Meteorological variables such as rainfall and temperatures are measured at station level They can also be derived from satellites, most commonly from geostationary satellites such as METEOSAT. 2. meteorological variables Data such as land, sea and cloud surface temperatures are measured directly by the satellite (as infrared radiation) while rainfall data have to be modelled based for example on cloud temperatures or other factors. 25 of 65

26 RAINGAUGES AND INTERPOLATED RAINFALL DATA At the global scale, raster data of rainfall are available as interpolations of rainfall stations. However resolution is generally low, in the range of several degrees, and rainfall data are often not available in real time. Popup Window Historical and climatological datasets: CMAP and GPCC Examples of these datasets are: the CPC Merged Analysis of Precipitation (CMAP) pentad and monthly data from 1979 at 2.5 degrees resolution; and the Global Precipitation Climatology Centre (GPCC ) of the German Weather service (DWD ) which provides different monthly datasets at 1 and 2,5 degrees resolution starting as early as of 65

27 RAINGAUGES AND INTERPOLATED RAINFALL DATA GTS station data: example from October 2010 Africa has a limited network of rain gauge stations and for diverse reasons - such as economic and civil insecurity coupled with a low perceived relevance of weather services - a large number of existing rainfall records is incomplete. 27 of 65

28 RAINGAUGES AND INTERPOLATED RAINFALL DATA GTS Gauge locations (April 2013) The coverage of GTS (Global Telecommunication System for Africa) of the WMO (World Meteorological Organization) includes less than 500 regularly reporting stations (most of them in South Africa) out of less than 1100 GTS stations On the other side, rainfall information is crucial for crop growth and is therefore an important factor for the monitoring of agricultural and pastoral production and consequently for food security. In addition to the poor availability of rainfall records, there is a steady decline of the standard observation network, which is a strong limitation for climate related research as well as for operational agricultural monitoring. 28 of 65

29 RAINFALL ESTIMATES Because of the presented limitations in the availability and quality of measured data, rainfall estimates have been developed. Rainfall estimates provide a spatial and temporal overview of the amount of rainfall based on a variety of input data, which generally depend on: satellite observations; and climate circulation models. Let s have a look at them of 65

30 RAINFALL ESTIMATES Two common examples of estimated rainfall data currently available for Africa - which we will discuss more in depth in a few screens - are the following: the rainfall forecasts of the European Centre for Medium-Range Weather Forecast (ECMWF); and the rainfall estimates (RFE) produced by the Climate Prediction Centre (CPC) of the NOAA, and used operationally by FEWS-NET of 65

31 ECMWF DATA The European Centre of Medium-Range Weather Forecast (ECMWF), based in Reading in the UK, produces mainly daily operational medium-range weather forecasts at the global scale. This data: is used for weather prediction, research and commercial purposes; concentrates mainly on medium-range forecasts, defined from 3 to 15 days; and is produced by the ECMWF s own model and assimilation system, the Integrated Forecast System (IFS). ECMWF is also known for its seasonal weather forecasts. 31 of 65

32 ECMWF GLOBAL CIRCULATION MODELS Global Circulation Model (GCM) 0,125 0 of resolution daily forecasts 10 days forecasts with a temporal resolution of 3/6 hours ECMWF forecasts are based on a high resolution global circulation model called HRES which simulates a series of meteorological variables at a 3 to 6 hours time steps including rainfall, temperature. Daily reanalysis and 10-daily forecasts are produced at a resolution of about 16 km (since 2008). This model is regularly updated / improved. ECMWF also builds a consistent archive with a resolution of about 80 km called ERA Interim (for ECMWF Re-Analysis). This archive starts in 1979 till present (with a 3 months delay for collecting measurements). 32 of 65

33 RFE COLD CLOUD DURATION MODEL The RFE is a rainfall estimate of NOAA's Climate Prediction Centre currently used by FEWS-NET and several United Nations agencies - such as the Food and Agriculture Organization (FAO) and World Food Programme (WFP) - for agricultural monitoring in a large number of African countries. It basically uses satellite imagery from the geostationary Meteosat Second Generation (MSG) and estimates convective rainfall as a function of top of cloud temperatures and using Global Telecommunication System (GTS) stations for calibration. 33 of 65

34 RFE - COLD CLOUD DURATION MODEL There are two versions of rainfall estimates RFE The two RFE versions are produced with slightly different methodologies and different input data: RFE 1.0 RFE 2.0 RFE 2.0 uses additional techniques to better estimate rainfall while continuing the use of CCD and GTS. Two new satellite rainfall estimation instruments are incorporated into RFE 2.0, namely, the Special Sensor Microwave/Imager Click on (SSM/I) the buttons board to read Defence more. Meteorological Satellite Program satellites, and the Advanced Microwave Sounding Unit (AMSU) on board NOAA satellites. RFE 2.0 rainfall estimates are available only from RFE 1.0 and RFE 2.0 are not directly comparable! 34 of 65

35 RAINFALL ESTIMATES Several other rainfall estimates exists, such as Tropical Rainfall Measuring Mission (TRMM) Tropical Applications of Meteorology using SATellite (TAMSAT) Other methods combine global circulation model outputs with rainfall estimates, like for example: FAO Rainfall Estimate (FAO RFE) (currently stopped in 2010) The next screen presents an overview of the existing rainfall estimates methods 35 of 65

36 RAINFALL ESTIMATES Popup Window Tropical Rainfall Measuring Mission (TRMM) The Tropical Rainfall Measuring Mission (TRMM) is a joint space mission between NASA and the Japan Aerospace Exploration Agency (JAXA) designed to monitor and study tropical rainfall. The TRMM Microwave Imager (TMI) is a passive microwave sensor designed to provide quantitative rainfall information over a wide swath under the TRMM satellite. By carefully measuring the minute amounts of microwave energy emitted by the Earth and its atmosphere, TMI will be able to quantify the water vapor, the cloud water, and the rainfall intensity in the atmosphere over the tropical belt (35 degrees north and south latitudes). More info at: 36 of 65

37 RAINFALL ESTIMATES Popup Window Tropical Applications of Meteorology using SATellite (TAMSAT) TAMSAT stands for Tropical Applications of Meteorology using SATellite data and ground-based observations and is a research group at the University of Reading (UK). Routine products of the group are a ten-daily (dekadal), monthly and seasonal rainfall estimates for Africa derived from Meteosat thermal infra-red (TIR) channels based on the recognition of convective storm clouds and calibration against ground-based rain gauge data. Since 01/2014, daily estimates are also produced (daily rainfall is needed for e.g. flood monitoring, weather index based insurance ) More info at: 37 of 65

38 COLD CLOUD DURATION MODEL TAMSAT rainfall estimates for October 2012 The TAMSAT group at the University of Reading produces rainfall estimates also based on the cold cloud duration model, but differently from the RFE algorithm, historical gauge data are used for calibration which is done at the level of geographical calibration windows. The advantage is that the product is not influenced by the available GTS station data and is therefore consistent over time The disadvantage is that the fixed calibration windows sometimes lead to sharp transitions between windows. It may also be poorly calibrated in areas with few gauge data. TAMSAT data are available as 10-daily and monthly data for Africa since 1983 and the operational production of the data at continental level has been supported by the JRC. 38 of 65

39 COLD CLOUD DURATION MODEL In tropical Africa, rain is mostly created by convective storms and the clouds below a given temperature produce rainfall. It is possible to derive estimates of rainfall by measuring the cloud top temperatures and the length of time a cloud is below the critical threshold temperature. This methodology is based on the Cold Cloud Duration (CCD) derived from TIR data. The threshold below which a convective cloud is precipitating is determined for a given region (or calibration area) by comparing the cases of rain / no rain versus CCD > 0 / CCD = 0 (contingency table). For rainy pixels (containing gauges), the dekadal rainfall is then regressed against their dekadal CCD. This calibration against historical gauge data allows to estimate rainfall in near-real time without any station data. 39 of 65

40 COLD CLOUD DURATION MODEL High cold cloud, raining Warm low cloud, not raining High cold cloud, not raining Warm low cloud, raining Popup Window CCD method s validation To establish the utility of the method, it must subsequently be validated by comparing estimates and gauge data from some area or period distinct from that used for the calibration. Alternatively, validation may consist of testing the use of the estimates in a quantitative application, such as a catchment runoff model, again using a separate period or area from that used for the calibration. Within the estimation procedure there are many decisions to be taken and numerical values to be assigned. 40 of 65

41 RAINFALL ESTIMATES Commonly available satellite derived rainfall estimates DEFINITION Distributed by Coverage Resolution Frequency Archive NOAA RFE V2 NOAA-CPC FewsNet Africa 0.1 deg daily, dekadal 2001-present ARC V2 CPC Africa 4 km? Daily, dekadal 1983-present TRMM 3B42-RT V7 NASA Global 0.25 deg 3 hourly 1998-present CMORPH CPC Global 0.25 deg Daily (3hr present PERSIANN University of averages) Global 10 km 6 hourly 2000-present Arizona, USA Tropical Applications of Meteorology using Satellite data and ground-based observations (TAMSAT) University of Reading, UK Africa 0.04 o (4 km) Dekadal, monthly Now daily 1983-present Multi Sensor Precipitation Estimate (MPE ) EUMETSAT Africa, Europe 4 km 5, 15 and 30min 2006-present 41 of 65

42 Need for validation of RAINFALL ESTIMATES All the data described in the previous screens are estimates and do therefore contain errors and show deviations in different regions of Africa. At the same time, validation or better calibration of these estimates is again hindered by the scarce availability of ground measurements. This makes it extremely difficult to assess the quality or reliability of each dataset and can have important implications for food security applications. Overestimation or underestimation of RFE (according to RFE/NDVI relationship). 42 of 65

43 Need for validation of RAINFALL ESTIMATES Example of Angola Feb 2014 ECMWF Rainfall anomaly Tamsat Rainfall anomaly Need for gauge data to validate the different rainfall products 43 of 65

44 RAINFALL ESTIMATES Why is the use of rainfall estimates especially important in Africa? Meteo stations are not available on the market in Africa. Low density meteorological network. Incomplete rainfall records. Rainfall estimates are more accurate than rain gauge data. Limitations in sharing ground station measurements. Please select the option of your choice (2 or more) and press "Check Answer". 44 of 65

45 COLD CLOUD DURATION MODEL Combine each rainfall estimates method with its related characteristic. 1. It uses the Special Sensor Microwave/Imager (SSM/I), and the Advanced Microwave Sounding Unit (AMSU). 2. It uses historical rainfall data from stations for calibration. RFE 1.0 RFE It uses rainfall data from stations for the dekadal estimates. TAMSAT Click each option, drag it and drop it in the corresponding box. When you have finished, click on "Check Answer". 45 of 65

46 RAINFALL ESTIMATES- DOWNLOADING DATA The following are some of the most common sites for downloading rainfall estimates: Cold cloud duration data TAMSAT Research Group, Department of Meteorology, University of Reading FEWS NET Data Portals Africa Rainfall Estimates - Climate Prediction Center - NOAA IRI/LDEO Climate Data Library Earth System Research Laboratory- NOAA PSD : CPC Merged Analysis of Precipitation Interpolated rain-gauge data* German Global Precipitation Climatology Centre (GPCC) #detail Atmospherical models* European Centre for Medium-Range Weather Forecasts NOAA Climate.gov: Science & Services for Society JRC - FOODSEC Meteodata Distribution Page - Europa of 65

47 TEMPERATURE DATA FOR TEMPERATE AREAS AND MOUNTAINOUS CLIMATES In temperate climates and in mountainous areas of tropical climates, temperature data can also play an important role in crop monitoring, since temperatures becomes an important additional factor to water availability for crop development. Sowing only takes place when required temperatures for germination are reached and when the risk of cold shocks is low. In many areas of the world measured temperatures are even more difficult to obtain than rainfall data. This is why again, satellite derived temperatures can be of use. The NOAA AVHRR sensor for example has a thermal band, such as also the geostationary METEOSAT. Otherwise temperature data too are available from global circulation models such as the ECMWF. There is a slight difference though: satellites measure surface temperature, while we usually need air temperatures. 47 of 65

48 VEGETATION INDICES Vegetation indices are mathematical transformations of the original multi-spectral data which are aimed at enhancing the information about vegetation properties while reducing the effect of external factors (mainly atmospheric and soil effects). Variation of reflectance for different ground targets (or spectral signature) The basic principle of vegetation indices is the strong absorption of light by leaf pigments in the visible (mainly red) range of the spectrum and on the high reflectance of leaf mesophyll in the near infrared range. This strong reflectance difference between the red and the near infrared channels is typical for vegetation and allows discrimination from other targets such as soil or water. Vegetation indices are based on the large difference of green vegetation and soil reflectance between the red and near infrared wavelengths of electromagnetic radiation. 48 of 65

49 VEGETATION INDICES Interaction of leaf structure with (visible and NIR) light RED and BLUE largely absorbed for use in photosynthesis As visible in this picture, the red and blue components of incoming radiation are absorbed by the surface of green leafs, while near infrared (NIR) radiation deeply penetrates the leaf but a significant portion of it is reflected backwards. While blue and red light is largely absorbed, NIR radiation is partially reflected. Since this behaviour is only true for green and healthy leafs, vegetation indices are generally proportional to the photosynthetic efficiency of green vegetation. 49 of 65

50 VEGETATION INDICES Although a large number of vegetation indices has been developed, most of them are conceptually similar and partially redundant. The table below shows an overview of the commonly used simple vegetation indices (VIs). NDVI SR EVI Click on the buttons to read more. SAVI 50 of 65

51 VEGETATION INDICES NDVI SR EVI SAVI NDVI= (NIR-RED)/(NIR+RED) The Normalized Difference Vegetation Index (NDVI) is one of the oldest, best known, and most frequently used VIs. The normalized difference formulation makes it suitable for a wide range of conditions. It can, however, saturate in dense vegetation conditions when the Leaf Area Index (LAI) becomes high. 51 of 65

52 VEGETATION INDICES NDVI SR EVI SR = NIR / RED The Simple Ratio (SR) index is another old and well known VI. The SR is simply the ratio between the near infrared and red reflectances. As the NDVI it is both easy to understand and effective over a wide range of conditions and it can saturate in dense vegetation when LAI becomes very high. SAVI 52 of 65

53 VEGETATION INDICES NDVI SR EVI SAVI EVI = 2,5 [(NIR-RED)/(NIR+6RED-7,5BLUE+1)] The Enhanced Vegetation Index (EVI) was developed to improve the NDVI by optimizing the vegetation signal in high LAI regions by using the blue reflectance to correct for soil background signals and reduce atmospheric influences, including aerosol scattering. This VI is therefore most useful in high LAI regions, where the NDVI may saturate. 53 of 65

54 VEGETATION INDICES NDVI SR EVI SAVI SAVI = [(NIR-RED)/(NIR+RED+L)] * (1+L) The Soil-Adjusted Vegetation Index (SAVI) includes information on the soil properties. The factor L depends on the distance of RED and NIR from the so called soil line of bare soils. This information must be available. A development of the SAVI which minimizes soil noise is the so called TSAVI (Transformed SAVI). 54 of 65

55 VEGETATION INDICES Among the different vegetation indices based on the NIR and RED related behaviour of vegetation, it is the NDVI the most popular indicator for studying vegetation health and crop production. Research in vegetation monitoring has shown that NDVI is closely related to the leaf area index (LAI) and to the photosynthetic activity of green vegetation. NDVI is an indirect measure of primary productivity through its quasi-linear relation with the fapar (Fraction of Absorbed Photosynthetically Active Radiation). Global NDVI image for October 2012 Cloudy areas (white) are visible in China and along the west coast of Central Africa. Parts of Siberia and Canada are covered by snow (blue). Click on the image to enlarge it. 55 of 65

56 VEGETATION INDICES Which of the following statements about Vegetation Indices is correct? VI are the optimal input data for land cover classification. VI are a direct measurement of crop yield. VI are generally proportional to the photosynthetic efficiency of green vegetation. Please select the answer of your choice. 56 of 65

57 VEGETATION INDICES - ISSUES All spectral devices operating in space are exposed to orbital drift and sensor degradation. Even with sophisticated radiometric calibration methods it is difficult to have satellite image time series of several decades that are perfectly consistent in time. For data from the widely used NOAA AVHHR sensor the problem is further complicated by the fact that the images of the last 25 years belong to different sensors mounted on different satellites subject to different degradation scenarios. The 2 sensors of the SPOT-VGT series (1 and 2) mark a clear progress in terms of improved consistency over time. However, the time series is only 14 years long so far. 57 of 65

58 VEGETATION INDICES - ISSUES Besides aerosol and water vapor related problems, cloud contamination remains the biggest problem for low resolution images. For most applications 10-daily images are used, where the daily images are composited in so called Maximum Value Composites (MVC) to eliminate at least the most perturbing atmospheric artifacts. The assumption here is that both clouds and atmospheric effects such as vapour and haze, are generally lowering NDVI, so by taking the maximum NDVI values during a time step of some days these effects are automatically reduced. Although very helpful, MVC cannot fully eliminate all the atmospheric noise present in the images. 10 More on Compositing and MVC A ten-days time step (or dekad) is also a reasonable period for monitoring changes in crop phenology and is widely used in agro-meteorology and crop monitoring. 58 of 65

59 VEGETATION INDICES - ISSUES Popup Window Compositing and Maximum Value Composites (MVC) Compositing is one of the procedures to get better land products. It consists of a combination of daily images to generate a single image that results in a cloud free land product over temporal intervals. MVC compares all the images taken by a sensor, such as MODIS, during a pre-defined period of time and selects the pixels with the highest vegetation index value since it is assume that contamination reduces the VI values [Viovy et. al. 1992]. Another approaches uses the mean value over the time period instead of the maximum. Although the method clearly improves the quality of land images it also has side effects: 1)adjacent pixels of an MVC image can belong to different days of the time period, e.g. in a 10-daily composite they can theoretically be 10 days apart; and 2)in vegetation monitoring, MVC s of the growing phase will select more pixels of the latest days of the time period while in senescence phases the opposite happens. Finally MVC cannot eliminate all cloud problems especially in areas with nearly continuous cloud cover during the whole compositing period. 59 of 65

60 VEGETATION INDICES - ISSUES Temporal smoothing techniques are commonly used in time series analysis and the number of different algorithms for temporal filtering continues to grow. The aim of the smoothing techniques is to remove artefacts related for example to undetected clouds and poor atmospheric conditions. Also, possibly occurring data gaps should be filled. Temporal smoothing also follows the principle of MVC that high NDVI values are generally good measurements, while sudden dips in NDVI are mainly due to cloud contamination and atmospheric perturbation. Differently from MVC it looks at good quality data before and after the current composite, while MVC operates inside the compositing period. NDVI profiles from different land cover types before (top) and after (bottom) smoothing with the Whittaker filter. A large number of algorithms has been proposed for smoothing out the bad dekads and smoothed images are usually appearing much cleaner than the original ones. 60 of 65

61 VEGETATION INDICES - ISSUES Which among the following factors affect the quality of vegetation indices for low-res images? Sensor degradation Aerosol and water vapor Cloud contamination Sensor temperature Senescent vegetation Please select the option of your choice (2 or more) and press "Check Answer". 61 of 65

62 VEGETATION INDICES - DOWNLOADING DATA The following are some of the most common sites for downloading low and high resolution data: USGS FEWS NET Data Portal - DevCoCast project website LP DAAC: ASTER and MODIS Data Products Reverb / ECHO Nasa Global - BOKU IVFL University of Natural Resources and Life Science MetopS10-AVHRR Vito SPOT-VEGETATION programme webpages STAR - Global Vegetation Health Products COPERNICUS - Global Land Service Products /vci/vh/index.php 62 of 65

63 COMMONLY USED ANCILLARY DATA Example of secondary data layers Administration Rivers Buffer main roads Roads Land Use Soil TM image DEM While time series of remote sensing data provide information on the status of agricultural vegetation, additional data layers are needed to contextualize this information. Ancillary data Data derived from existing sources (other than remote sensing) used to assist in classification and analysis of remotely sensed data. These data have been collected for other purposes, often independent from crop monitoring. Examples: topographic, demographic or socioeconomic data. 63 of 65

64 COMMONLY USED ANCILLARY DATA For mapping purposes but also for possible spatial aggregation administrative unit layers (1) and topographic information (2) are needed. This is important also because other data that can be used for comparison such as agricultural statistics (3) are generally aggregated by administrative units. Furthermore it is important to distinguish agricultural vegetation from natural vegetation which has a different behaviour for example in terms of spectral signature, and crop masks (7) can be helpful in knowing the spatial distribution of crops. Any other thematic layers on the studied areas are also helpful in crop monitoring, such as for example soil and land use maps (6), geomorphology and hydrology (4). Since the ultimate goal of crop monitoring in semi-arid countries is often to contribute relevant information to food security analysis, socio-economic data (5) also play an important role. This can be for example information on livelihood groups, vulnerability information, food consumption and nutrition data. In the temporal domain, crop calendars (8) are extremely important to understand the seasonal patterns and distribution of crop cycles. 1. Administrative boundaries 2. Topographic information 3. Agricultural statistics 4. Geomorphology and hydrology 5. Socio-economic data 6. Land use, land cover and soils 7. Crop masks 8. Crop calendars 64 of 65

65 IF YOU WANT TO KNOW MORE Additional reading About TAMSAT: Grimes, D. I. F., E. Pardo-Igúzquiza, and R. Bonifacio, 1999: Optimal areal rainfall estimation using raingauges and satellite data. Journal of Hydrology, 222(1-4), RFE rainfall RFE2 Xie, P. and Arkin, P. A. 1997, A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bulletin of the American Meteorological Society 78(11): " Rainfall estimates validation and comparison: Dinku, T., P. Ceccato, E. Grover Kopec, M. Lemma, S. J. Connor, and C. F. Ropelewski, 2007: Validation of satellite rainfall products over East Africa s complex topography. International Journal of Remote Sensing, 28(7), About NDVI: of 65