LAI Data for Carbon Balance of Fast Growing Forests in Italy

Summary:

This study is a part of the long term Kyoto-Experiment, a JRC-IES research project included in the framework of the CARBOEUROPE cluster of projects aimed to understand and quantify the carbon balance at European level. In this context, the University of Milano-Bicocca is collecting LAI and fAPAR field measurements in order to develop local relationships between canopy properties and carbon exchanges and to validate moderate resolution remote sensing products.The investigated sites are fast growing forests located along the alluvial plain of the Ticino and Po rivers, in northern Italy. LAI was measured with different techniques: LiCor LAI-2000 PCA, hemispherical digital camera and destructive sampling.

More information about the LAI and other data collected by this project can be found in the presentation by Colombo et al. "CarboEurope - LAI Mapping in Italy" at the CEOS LAI-Intercomparison Meeting, Missoula - Montana. 2004 (http://landval.gsfc.nasa.gov/LPVS/LPV_LAI_meeting04.html).

Data Citation:

Cite this data set as follows:

Colombo, R. Meroni, M. Busetto, L. Seufert, G. 2004. CarboEurope - LAI data for Carbon Balance of Fast Growing Forests in Italy. Available on line from (http://www.disat.unimib.it/telerilevamento/downlds.htm). Remote Sensing of Environmental Dynamics Lab. DISAT-UNIMIB, Italy.

1. Data Set Overview:

This data set include field data collected in two experiments, namely the:

1) JRC-KYOTO EXPERIMENT;

2) PROGETTO KYOTO: RICERCA SUI CAMBIAMENTI CLIMATICI E IL CONTROLLO DEI GAS SERRA IN LOMBARDIA.

1) The JRC Kyoto Experiment is a long-term flux-monitoring research held in the Ticino Natural Park (Ticino Site), in northern Italy. The JRC Kyoto experiment is a pilot study aimed to develop tools and process understanding for monitoring the full balance of greenhouse gases in a local, but representative series of land-use types, from pristine forest to poplar- and rice-plantations.

2) The Kyoto Project is an integrated three years project (2004 - 2006) that aims at providing the basic information (data, scenarios, policies) for monitoring and reduction of greenhouse gases emissions on a regional scale. The Kyoto Project, natural follow up of the Piano Regionale per la Qualita’ dell'Aria (Regional Plan for Air Quality) aims to achieve the following objectives:
        i) Climate study: the study of climate evolution and current trends in Lombardia also considering the frequency and intensity of extreme meteorological events;
        ii) Emissions: the implementation and update of the emission inventory for the six greenhouse gases prescribed by the Kyoto Protocol and targets for EU policies;
        iii) Terrestrial cycles and other greenhouse gasses: the improvement of monitoring networks for greenhouse gases emissions and net carbon fluxes among the atmosphere, soil and agro-forestry systems in Lombardia;
        iv) Externalities: evaluation of economic, environmental and health externalities linked with the actual climate change;
        v) Scenarios, policies and measures: the analysis of the expected emissions changes related to different economic and/or control policies (e.g. clean development mechanisms joint implementation, emission trading) and the implementation of protocols for actions to be undertaken in the operational fields responsible for greenhouse gases emissions;
        vi) Public communication: build-up a web site open to policy makers, stakeholders, scientific community and general public;

Related Data Sets:

1) JRC-KYOTO EXPERIMENT: http://ies.jrc.cec.eu.int/Climate_Change.40.0.html
2) PROGETTO KYOTO: http://www.flanet.org/ricerca/kyoto.asp
3) LAB. Telerilevamento Home page: http://www.disat.unimib.it/telerilevamento

Title of Investigation:

LAI Data for Carbon Balance of Fast Growing Forests in Italy

Investigators:

Name                      Email
Colombo, R             roberto.colombo@unimib.it
Meroni, M.              michele.meroni@unimib.it
Busetto, L.               lorenzo.busetto@unimib.it

Technical Contacts:

Name Email Phone
Colombo, R. roberto.colombo@unimib.it +00 39 0264482819

Future Modifications and Plans:

A LAI measurement campaign will be carried out in 2005 in order to validate MODIS LAI global products in the patchy agro-ecosystem landscape typical of the Ticino alluvial plain. The planned strategy is to collect LAI data in pixel pairs of 1km x 1km centered on the plantation sites and to develop appropriate transfer functions using high-resolution satellite imagery. In this context, an intensive ground campaign (LAI, fAPAR, specific leaf area, canopy water content and pigments) is foreseen simultaneously with a hyperspectral airborne CASI 2 acquisition.

2. Data Characteristics:

This dataset contain LAI data collected during two experiments aimed to develop methods for extracting parameters from remotely sensed data to be incorporated in ecosystem models.

i) The first experiment was conducted at the JRC Kyoto experimental site, a long-term flux-monitoring site located in a traditional poplar plantation in the Ticino Natural Park. This data set contains data reported in Colombo et al., (2003) and in Meroni et al., (2004). The download file is an excel file which includes two data sheets (2001-LAI-data.xls). The first sheet contains a description of the test site and a summary of data used. The second one includes all LAI measures acquired around 20 June 2001 in an area of 1x1km. LAI data were collected with the sampling scheme shown in Meroni et al., (2004), sampling a patchy area of about 1x1km characterized by different plantations and woodland.
A summary of the measured parameters is reported in the following table:

X, Y COORD	UTM WGS84, Zone 32
LAIEO	Effective LAI of the overstory (only crown) computed using all 5 LAI-2000 rings
4R_LAIEO	Effective LAI of the overstory (only crown) computed using first 4 LAI-2000 rings
LAIE(U+O)	Effective total LAI (measured at ground level: overstory + understory); computed using all 5 LAI-2000 rings
4R_LAIE(U+O)	Effective total LAI (measured at ground level: overstory + understory); computed using first 4 LAI-2000 rings
SAIWINTER	Stem Area Index measured in winter (no foliage)
4R_SAIWINTER	Stem Area Index measured in winter using first 4 rings
LAI_GREEN	Difference between LAIEO and SAIWINTER
4R_LAI_GREEN	Difference between LAIEO and SAIWINTER using 4 rings
LAIU_S	Understory LAI computed as the difference between total LAI and overstory LAI
4R_LAIU_S	Understory LAI computed as the difference between total LAI and overstory LAI, using first4 rings
FCO	Overstory fractional cover (from forestry meas.)
FCO	Understory fractional cover (from Digital photos meas.)

ii) The second experiment was conducted in the framework of the Kyoto Project, a regional project founded from national institutions. The experiment site is a short rotation poplar forest (SRF) located in northern Italy (Vigevano site). An intensive measurement campaign was conducted through the 2004 season and several field data were collected with different temporal frequencies. LAI measurements were collected with the sampling scheme shown in Colombo et al. (2004), sampling two poplar fields of about 500x500m. Besides data provided by the eddy covariance tower, a summary of the measured parameters is reported in the next table:

midday fAPAR/fIPAR	fAPAR/fIPAR collected around midday with a Sun Scan Canopy Analysis System (SS 1, Delta-T)
daily fAPAR/fIPAR	fAPAR collected through the day (10 measurements).
eLAIo (LAI-2000 PCA)	Effective LAI of the overstory computed using first 4 LAI-2000 rings
eLAIo (HC)	Effective LAI of the overstory computed using a Nikon digital hemispherical camera
Leaf Area	Mean destructive Leaf area of six plants
LAI (destructive)	Plantation LAI computed scaling Leaf Area with tree density
SLA	Specific Leaf Area
SLA vertical profile	Vertical profile of the specific leaf area (only one day)
Leaf Fresh and dry weight	Leaf fresh weight measured in field and Leaf dry weight
Diameter and Height	Mean diameter and height of the trees at the SRF site
Nadir Photos	Digital photo for site characterization
Chl (Minolta Spad)	Relative chlorophyll concentration
EWT/FMC	Equivalent water thickness and fuel moisture content
EWT/FMC vertical profile	Vertical profile of equivalent water thickness and fuel moisture content (only one day)
Daily EWT/FMC	Daily variations of leaf water content (only one day)
Daily MTA	Daily mean tilt angle measured by a protractor
Fc	Fractional cover computed by LAI-2000 PCA
Above/Below Biomass	Total net primary production
Management	Irrigations
Start/end of season	For meteorological and phenology

This data set contains data not yet published. The downloadable file is an excel file which includes three sheets (2004-SRF.xls). The first one includes a description of the test site and a graphical representation of the sampling scheme. The second one includes the time table of all the measurements collected at the Short Rotation Experimental Site, while the last one includes all LAI and fPAR data acquired through the 2004 growing season. The parameters reported in the last data sheet are the following: (n.a. indicates not available measure, while n.y.c. indicates not yet calculated values)

Date	Date of measurements
DOY	Day of the Year of measurements
Fapar	Average Fraction of Absorbed Photosyntetically Active Radiation and its standard deviation (sd)
Fipar	Average Fraction of Intercepted Photosyntetically Active Radiation and its standard deviation (sd)
ePAIo	Average Effective Plant Area Index, measured with LAI200 PCA and its standard deviation (sd)
ePAIo	Average Effective Plant Area Index, measured with hemispherical camera and its standard deviation (sd)
LAI dest.	Average Destructive LAI and its standard deviation (sd)

Site boundaries: (All latitude and longitude given in degrees and fractions)

Site (Region)	Westernmost Longitude	Easternmost Longitude	Northernmost Latitude	Southernmost Latitude	Geodetic Datum
Ticino (Italy): Traditional Poplar Plantation	9.05083	9.07	45.20639	45.19306	WGS 84
Vigevano (Italy): Short Rotation Poplar Forestry	8.85472	8.89	45.30167	45.28833	WGS 84

Time period:

The data set covers the period 2001/06/20 to 2005/10/01.

3. Data Application and Derivation:

Typical Application of Data:

Regional level application for biogeochemical models. The data set of the Ticino Site (Traditional Poplar Plantation) and the LAI map is available for MODIS validation purposes but it should be used with attention since the measured area is only 1x1km. This dataset include all LAI measurements collected in two days in correspondence of the DAIS hyperspectral airborne survey (20 June 2001). The data set of the Vigevano Site (Short Rotation Poplar Forest) is about 500x500m but include data collected through the 2004 season in two different fields. Such data can be used only to test algorithms starting from Landsat data. Challenging issues are posed by LAI retrieving in our patchy agricultural area with MODIS data (unmixing approach is needed.

Theory of Measurements:

LAI and leaf mean tilt angle measurements were collected with the LAI-2000 Plant Canopy Analyzer (Licor Inc., Nebraska). LAI-2000 measures the gap fraction P(I) in five zenith angles (I) ranges with midpoints of 7°, 23°, 38°, 53° and 67°. LAI was determined by inverting a simple radiative transfer model according to Welles and Norman (1991) from the LAI-2000 gap fraction measurements. Leaf mean tilt angle was determined assuming a uniform leaf azimuth distribution and a constant leaf normal angle, using the Lang graphical method.

LAI map derivation: A LAI map of the plantation at the Ticino Site was obtained by the inversion of an optimal spectral subset of the DAIS multi-view angle data set using the best performing inversion configuration, whit a cost function that combines a weighted least squares approach and the exploitation of prior knowledge about the model variables.

4. Quality Assessment:

Errors, limitations and known problems with the Data:

In this section we provide an overview of the errors found in LAI field data and in the fine scale LAI map obtained from DAIS airborne images.

Quality Assessment Activities:

i) To asses the quality of indirect LAI measurements we compared such data with direct estimation of leaf area obtained by destructive sampling. The LAI derived from destructive measurements was always greater than the LAI computed by LAI-2000 and this difference can be due to the observed clumping at crown level. The LAI-2000 and the destructive LAI measurements are strongly linear related with an RMSE of 0.32 m²/m².

ii) The LAI map derived from DAIS reflectance data was evaluated in terms of the RMSE of LAI retrieval calculated comparing the estimated LAI map with LAI-2000 ground measurements at selected sites. The issue of matching the spatial footprint of LAI-2000 measurements and the one of model inversion (2.5x2.5m pixel resolution) was accomplished considering the image spatial resolution with respect to the ground area sampled with LAI-2000, represented by a circle segment with radius approximately ranging from 14 to 42m (see Meroni et al., 2004 for detail). We claim that the best performances is achieved exploiting multi view DAIS data and prior knowledge information about the model variables (RMSE of 0.39 m²/m²).

A LAI map was also derived with a semi-empirical model approach, performing a regression analysis between ground collected LAI measurements and different spectral vegetation indexes calculated from DAIS data. SWIR corrected spectral vegetation indexes were found to improve LAI maps accuracy, and the best relationship was found between LAI measurements and the NDVIcor vegetation index, with a r² of 0.82. More sophisticated approaches (e.g. geometric - turbid medium hybrids, 3D descriptions) might permit a better exploitation of the canopy reflectance anisotropy. Finally, we remark that poplar plantations exhibit low LAI values and should be considered as sparse canopy forest with a signal that is also affected form significant understory influences

5. Data Acquisition Materials and Methods:

Sites description:

The investigated sites are fast growing forests located along the alluvial plain of the Ticino and Po rivers, in northern Italy. The sites belong to a patchy agro-ecosystem landscape, composed by different land uses that produce high variability in canopy properties at regional level. The selected fast growing forest sites represent the main plantation typologies in this region, short rotation poplar forestry and traditional poplar plantation.
The Ticino site is a traditional poplar plantation with tree density of about 300 trees ha-1 and tree height ranging from 3 to 24 m. This kind of plantation is typically clear-cut every 8-12 years for wood production.
The Vigevano Site is a Short Rotation Forest (SRF) with tree density around 12000 trees ha-1 and tree height starting from 0 m at the beginning of the growing season and reaching about 3.5 m at the end of the season. This kind of plantation is typically clear-cut every 1 or 2 years for bio-fuel production.
Both sites are equipped with a scaffold tower for the measure of gas exchanges with eddy covariance technique and are managed by the Institute for Environment and Sustainability (IES) of the Joint Research Centre (JRC, Ispra, Italy) of the European Commission.

Sampling scheme for LAI Field measurements:

LAI was measured with different techniques: LiCor LAI-2000 PCA, hemispherical digital camera and destructive sampling. Field data were collected employing a nested sampling design consisting in a main plot (Elementary Sampling Unit, ESU) sampled by individual measurements collected along a transect within each ESU. The forest plantations exhibit geometrical properties typical of row-crops (i.e. spatial pattern). Therefore, for each plantation type , the shape and the dimensions of the ESUs have been carefully defined in order to ensure unbiased estimates of the plantation average properties (e.g. the ESU tree density must be equal to the plantation tree density). In particular, the spatial sampling scheme has been defined by the number of samples (SRF: 10 gap fraction measurements and 3 hemispherical photos; PP: 5 gap fraction measurements and 3 hemispherical photos) collected along the transect, the ESU size (SRF: 3.5x1.5 m; PP: 6x12 m), and finally by the location and number of the ESUs (SRF: 10; PP: 2 to 20 (Table 1). The ESU number and location have been selected on the basis of the knowledge about the site heterogeneity in order to sample the variability in vegetation structure in a sampling area ranging from 250x250m to 1x1km.

Test Site	ESU Number	ESU Size (m)	ESU Samples			Data Availability
Test Site	ESU Number	ESU Size (m)	L	H	D	Data Availability
SRF1 and SRF2	5	3.5x1.5	10	3	6	2004
PP1	10	6x12	5	3	0	2001-2003
PP2 to PP11	2	6x12	5	3	0	2003

Table 1: ESU number, size and number of samples collected within the ESU by LAI2000 (L), hemispherical digital camera (H) and destructive samples (D).

Gap fraction measurements were acquired with LAI-2000 PCA under diffuse radiation conditions at sunset, using two inter-calibrated sensor units equipped with a 45° view cap. The effective LAI was computed excluding the fifth ring (67° midpoints). In this condition, the radius of the circle segment spanned from the sensor FOV is approximately 1.7 times the canopy height. To estimate the effective LAI of the overstory, the below canopy measurements were acquired positioning the sensor unit above the understory, while the LAI of understory was quantified with two approaches. The first one computes the LAI of the understory as the difference between the total LAI (measured by an additional transect positioning the sensor at ground level) and the overstory LAI. With the second approach, the LAI of the understory is computed by considering the below measurements for the overstory LAI as above measurements for the ground level transect in the gap fraction calculation (Colombo et al., 2002). Finally, the magnitude of stem and branch contributions to LAI was appraised by collecting wintertime gap fraction measurements (trees without foliage).

The effective LAI was also measured at the SRF site through analysis of hemispherical photographs acquired with a NIKON 4500 Coolpix digital camera, equipped with a FE-E8 8mm fisheye lens converter. In order to enhance the separation between sky and non-sky pixels, an improved classification method was developed. With this method, the image is divided into 121 sectors of 150x150 pixels and an unsupervised isodata classification is applied to each sector to assign pixels to sky or vegetation classes. This method reduces the uncertainties in LAI estimation with respect to the traditional approach in which a threshold is subjectively selected for the classification of the entire image. It also reduces the errors due to the acquisition of images in non-homogeneous sky illumination conditions.

Destructive sampling was performed at SRF1 and SRF2 only, providing the leaf area for a single tree. Six plants were sampled at each measurement campaign and the leaf area index at the plantation level was computed by scaling the average leaf area of the trees with the tree density.

Remote sensing data:

On 20/06/01, from 10.30 to 11.00 a.m. local solar time (solar zenith angle and solar azimuth angle, were 23° and 150°, respectively), four aerial stripes were acquired by the Digital Airborne Imaging Spectrometer (DAIS 7915) of the German Aerospace Centre (DLR). The DAIS sensor acquires data in 79 different spectral bands, spawning from 0.4 um to 12.6 um, with a spectral resolution of 15-30 nm in the range 0.4 -1.0 um, of 45 nm in the range 1.5 -1.8 um, of 20 nm in the range 2.0 -2.5 um, and finally of 0.9 nm in the range between 8.0 and 12.6 um. The DAIS instantaneous field of view (IFOV) is equal to 0.189°, while the FOV is of ±29°. With such a configuration, the aircraft was planned to fly with average altitude of 1800 m, resulting in a swath of 1280 m with a spatial pixel size of 2.5x2.5m.
The fine scale LAI maps were generated from hyperspectral airborne imagery at the Ticino experimental site only. LAI maps were generated with two different approaches: semi-empirical models (regression) and inversion of canopy reflectance models (PROSPECT + SAIL ). Field campaigns aimed to the validation of these products were devoted to the measurement of vegetation canopy properties (i.e. leaf area index and mean leaf inclination) and non-canopy properties affecting the signal, such as atmospheric properties and background reflectance (i.e. optical thickness, soil reflectance).
The inversion of the canopy reflectance model exploiting prior knowledge about the model variables on a selected subset of the spectral bands of a multi-view hyperspectral DAIS 7915 (DLR) data set, provided a LAI map with an RMSE of 0.40 m²m^-2.