Forestry Methods 2

The AccuCarbon project develops and demonstrates a pioneering service on estimating the carbon storage of tropical forest. Combining data on moving cultivation from remotely sensed data with models on biomass accumulation enables the precise and accurate estimation of biomass carbon density even in challenging landscapes with shifting cultivation. Historic remote sensing data reveals the year of abandonment of a shifting cultivation site which is the year zero for the biomass accumulation model. The developed tool is based on the following steps:

Description of dynamics of secondary forest area and location through remote sensing

Modelling of the biomass and carbon density

Computation of landscape carbon by coupling spatio-temporal information of secondary forest and modelled carbon density

The project does not only focus on the monitoring of carbon accumulation of secondary forests but it also considers carbon throughout the whole forest area.

Time-series of remotely sensed data is used to detect the changes in the forest. First, a forest – non-forest map of the area is made from optical imagery, which is augmented with radar data if needed. The forest is divided into primary, secondary and disturbed forest. After the initial mapping, clear-cut monitoring of the forest area starts. Both optical and radar data are used.  A record of the time of the clear-cut is kept and it is used to calculate the carbon for the desired point of time. A new forest – non-forest map is also made for that desired point of time.

In the conference, initial results of experiment will be presented. Proof of concept study is done in Chiapas, Mexico.  Spot5/Take5 is used together with Landsat-8 to detect changes which have happened between the end of 2014 and summer 2015. Change detection result from optical imagery is used as reference to verify the changes detected from Sentinel-1 time series. Ground reference data provided by the Mexican users is used to model the biomass levels. This data is combined with the modelled carbon calculated to the cleared forest patches.

The service developed in Accucarbon combines leading expertise on the analysis of tropical forest using Earth observation optical and active microwave data with carbon accumulation models that have been developed independently from the remote sensing community. Such models that will be applied in the disturbed forests can to our understanding greatly reduce the uncertainty on estimation of forest biomass and it can be used to provide information on the emission or removal of forest carbon to support the MRV process of REDD+. The project has strongly committed users from Mexico. The main user is Chiapas government as a public authority with a mandate to develop the REDD+ and MRV. The Chiapas user is supported by the Instituto de Silvicultura e Industria de la Madera, Universidad Juárez del Estado de Durango. As the biomass accumulation model can be parameterized to all humid climates, the tool to be developed in the AccuCarbon project can be applied with some modifications outside of Mexico.

AccuCarbon is performed under the Copernicus framework and the main earth observation data used is from Sentinel-1 and Sentinel-2 missions. Spot5/Take5 data is used in the first phase to substitute Sentinel-2 data and Landsat is used to supplement the time series. ESA funds the Accucarbon project through data user element (Innovators III).

The latest United Nations 2015 Forest survey (FRA 2015) showed that deforestation in temporal and boreal forests are now decreasing. The same report  however notes that deforestation in the tropics is still on-going with its major effects on biodiversity, local livelihoosds and GHG emisions. Although the FRA gives us a very good global and national picture, it is not comoplete. Global and local forest dynamics are complex and more detailed information is needed for better understanding and planning.  Remote sensing is able to add important dynamics that are not described in the FRA as:  where forest is dissapearing (change is ususall spatially defined in narrow area, e.g. Brasil's Amazon frontier), what are the impacts (GHG and biodiversity),  what are the driver (fires, man-made), where change in forest has a much higher risk of having negative mipacts (  E.g. although deforestation in some  countries  is low however it is treathening important bioviersity or national parks), the function of trees outside the forest, the GHG emisions deforestaiton is cuasing in the tropics and the and the distinction between natural forest and plantaiton forests.

  In this presentation Wolrd Resources Institute- Global Forest Watch (GFW) initative shows the use of several global products derived from satellites like:

- Global 30 m pan tropical above ground biomass density (from WHRC derived from Landsat and ground points),

- Annual global  tree cover change  ( from UMD derived from landsat),

- FORMA (forest alerts 250 m derived from MODIS), and

- Planatation mapping in 6 major countries (very high resoluiton data from several satelites)

- Dryland mapping of bushes and ofrest in the horn ofg Africa (CollectEarth and google)

- GHG emisions pan tropical  (combined data source)

- locally sources GIS data on land cover and land use and landuse

FRA data augmented by data mentioend above derived  from satellites and GIS, GFW will paint a more complete picture of forest trends and changes. GFW wil further look nito the need to make data open , transparent and avialbel to stakeholders (e.g. via its GFW platform or other ways) . The presentation wil further outline the need for more and better land cover and land use maps throughout the topics  that can be produced by the new satellites constelations (sentinel, Commerical providrs).

fAPAR is a radiometric quantity describing the fraction of photosynthetically active radiation (PAR) absorbed by a plant canopy. It is an important component of carbon cycle and energy balance models and has been named as one of the 50 Global Climate Observing System (GCOS) essential climate variables (ECVs). Space agencies such as ESA produce satellite fAPAR products in order to address the need for spatially explicit global data to address environmental and climate change issues. Given the derived nature of satellite fAPAR products it is essential to independently verify the results they produce. In order to do this, validation sites (or networks of sites) are needed that directly correspond to the measurands. Further to this, in order to understand divergences between product and validation data, uncertainty information should be provided with all measurement results.

This work describes the development of a realistic 3D virtual forest site for radiative transfer model simulations that will act as the interface between the satellite estimates of fAPAR (including various assumptions, etc.) and the reference data; a wireless PAR network which collects the measurement results from a subsection of the field site; and the calibration and characterisation of the sensors that make up the network. The approach uses an accurate 3D forest model derived from terrestrial LiDAR scans of the field site to bridge the gap between the satellite fAPAR products and the in situ reference measurements. fAPAR can be simulated for the model, under various conditions, using librat (a Monte Carlo Ray Tracing library), while sensor models can be used to determine the theoretical values that are recorded by the sensors. The site, in Wytham Woods, Oxford (UK), is a 6 ha area of a larger deciduous forest, which was scanned leaf-on in June 2015 and is due to be scanned leaf-off in December 2015. The virtual site is constructed using the quantitative model approach (from [1-2]) for stem and branching structure, while the leaves are added based on the derived light availability from the point cloud. The reflectance of site constituents (canopy leaves, bark, understory and soil) was sampled during the June field campaign using an ASD Field Spectrometer.

The PAR sensor network covers an area of approximately 150 m². The system is managed wirelessly using an aggregator unit that collects the individual data logger readings. The PAR network layout replicates a hexagonal pattern with sensors located at 30 m intervals to avoid autocorrelation. The network is set up to record four-flux fAPAR (incoming and reflected above and below the canopy; see [3] for description of various types) using sixty-two below canopy sensors and two above canopy sensors. Each point in the network is comprised of two Apogee sensors simultaneously measuring the soil reflected and canopy transmitted PAR component; the above canopy pair record the sky irradiance and canopy reflected PAR component. Calibration and characterisation of the PAR sensors that are used in the network is critical to understanding the sources of uncertainty in satellite fAPAR products. The sensors are all individually calibrated at the National Physical Laboratory (NPL) with procedures that are traceable to SI. The angular, spectral and linearity response of the sensors are also characterised in order to a) derive the calibration coefficient and b) understand the uncertainty contributors. The individual sensor uncertainty, derived from these detailed measurements, will be one of the key inputs into the network uncertainty budget and therefore the validation data. 

Ultimately, the aim is to inform the production of better fAPAR products for use in downstream applications. In order to achieve this, the effect of model assumptions that are employed in fAPAR products must be investigated. Using the PAR network as an anchor point, and the model as translation tool, this can be achieved.

ACKNOWLEDGMENTS

We thank A. Barker, T. Jackson and D. Fox for their assistance with fieldwork.

REFERENCES

[1] Calders, K., Newnham, G., Burt, A., Murphy, S., Raumonen, P. et al. (2015). Nondestructive estimates of above-ground biomass using terrestrial laser scanning. Methods Ecol Evol, 6, 198–208.

[2] Raumonen, P., Kaasalainen, M., Åkerblom, M., Kaasalainen, S., Kaartinen, H. et al. (2013). Fast Automatic Precision Tree Models from Terrestrial Laser Scanner Data. Remote Sensing, 5, 491–520.

[3] Widlowski, J-L. (2010) On the bias of instantaneous FAPAR estimates in open-canopy forests. Agricultural and Forest Meteorology, 150, 1501-1522

The effect of climate change on forest ecosystems has substantial influence on their vitality. Due to the velocity of climate change forest ecosystems cannot adapt adequately and are affected increasingly. Through new operational optical spaceborne sensors the impact analysis of climate change on land surface will be increased and fostered. These sensors are already available (Sentinel-2A) or will be available in near future (EnMAP). Sentinel-2A was launched in April 2015 and offers high spatial resolution with 20m and a spectral resolution of 13 bands in visible/near infrared (VNIR) and shortwave infrared (SWIR) spectral region. The repetition rate of 5 days together with Sentinel-2B offers temporally high resolution operational data for continuous analyses. In 2018 EnMAP will be launched which is a hyperspectral operational mission for 30m spatial resolution with 250 spectral bands in VNIR and SWIR spectral region.        

Vegetation pigment status such as Chlorophyll, Carotenoids and Xanthophyll content can be observed in the VNIR region. The SWIR spectral region, whereas, is important for detecting the water content of photosynthetic active biomass. This leads to the assumption that a high spatial and high spectral resolution of VNIR and SWIR region is of great importance for deriving these parameters. Adding the assumption that higher temporal resolution will support the detection of change over time leads to the hypothesis that high spatial, spectral and temporal resolution is vital for pigment and water derivation and their change regarding forest ecosystems.

This structured analysis examines the potential of high spatial resolution Sentinel-2 type data for detecting forest vegetation parameters, in particular Chlorophyll and canopy water content. Through selection of adequate vegetation indices the importance of relevant spectral features can be emphasized and evaluated. First a selection of appropriate bands will yield significant spectral regions for detecting the parameters. Out of these bands existing vegetation indices will be selected and applied onto the images. The evaluation of these indices will lead to the conclusion which spatial and spectral resolution is important for deriving Chlorophyll and canopy water content of forest vegetation. Additionally high spectral and spatial resolution EnMAP type data will be analysed for their importance for deriving Chlorophyll and canopy water content. Furthermore available multi-temporal data will be examined if high temporal resolution supplements additional value to the derivation of Chlorophyll and canopy water content and their change.  

Our study site is a temperate spruce forest in south eastern Germany where in year 2013 several trees were ring-barked for a controlled die-off. Additionally control trees were identified and observed same as the ring-barked trees. During this experiment hyperspectral HySpex VNIR and SWIR data with 1m spatial and 416 bands spectral resolution were acquired for twelve time steps during the vegetation periods of 2013 and 2014. Furthermore tree needle samples of all generations were gathered for several times corresponding to HySpex overflight dates. These samples were analysed regarding their spectral responses and chemically analysed for their Chlorophyll content. Out of the HySpex data Sentinel-2 and EnMAP data will be generated with their spatial and spectral characteristics for the structured analysis.

Evaporation from the land surface is needed to estimate regional water balance, which can be used for monitoring health of vegetated areas and ecosystem disturbances. The usage of the heat budget equation of land surface is one of methodological approaches for the evaporation rate mapping on the base of Earth Observing System (EOS) imagery:

LE=S*(1-A)-H-R-G,       (1)

where LE – the latent heat flux, W/m²; S – the incoming radiant energy (shortwave) falling on land surface at the angle, W/m²; A – albedo of land surface; H - the sensible heat flux, W/m²; R – the longwave radiation balance, W/m²; G – the heat flux in soil (ecosystem), caused by daily and annual thermal cycles, W/m².

The land surface energy budget approach can be divided into two sub-approaches. The latent heat flux may be calculated as a residue from the heat budget equation, if other terms of equation (1) are known. The errors of the evaporation rate calculation depends on the errors of the eddy coefficient estimating. The best estimation of the eddy coefficient can be obtained by using flux towers. Alternatively, evaporation rate may be mapped using thermal inertia technique. It can be obtained on diurnal variation in surface temperature measured by satellite. This approach has minimum of initial assumptions and the small dependence of the results from instantaneous changes in the weather conditions. It requires the complicate calculation of the coefficient of turbulent heat transfer in the atmospheric surface layer.

In this study we combine the thermal inertia approach with flux tower observations to develop a technique of land surface evapotranspiration mapping.

Study area consists of taiga forest in southern part of Finland and Leningrad district of Russia.

We used results of multiple Terra/Aqua (MODIS) satellite survey and Hyytiälä (forest) and Siikaneva (marsh) flux towers (Finland) regime observations, as well as, the thermal inertia approach. After analyses of heat and moister fluxes the new empirical equations for coefficients of turbulent heat and mass transfer at the active surface of forest and marsh ecosystems were presented with the shortwave solar radiation and the speed of wind, as arguments. The new model accuracy of latent heat flux is 3.6 W/m² (1%) for a forest and 12 W/m² for a marsh. Model precision is 38 W/m² for both ecosystems.

Finally, the map of daily averaged evapotranspiration was compiled with help of the thermal inertia approach for the territory of southern part of Finland and Leningrad district of Russia using the map of ecosystem classes. The evapotranspiration map reflects differences of evapotranspiration between ecosystem classes such as forest, marsh, agricultural and urban areas. Forested areas located near the border between Finland and Russia have different evapotranspiration, which could result from different forest management operations in these two countries.