1. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy PREFER WP 3.2 Information support to Recovery/Reconstruction Task 7 Damage Severity Map PREFER WP 3.2 Information support to Recovery/Reconstruction Task 7 Damage Severity Map

2. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy Remote sensing data for burn severity assessment The WP goal is to provide reliable information on fire effects over Mediterranean areas and do that in a way that is comparable from region to region and over time. Remote sensing data allows to evaluate the damages caused by fires even in remote or inaccessible zones. LANDSATE 8 FALSE COLOR IMAGES (burned areas in red)

3. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy Landsat and Sentinel satellites data are well suited for damage assessment. The 30-meter spatial resolution is effective, and the spectral signals allow to detect burned areas. Landsat and Sentinel provide continuous and repetitive coverage for most land areas of the world. This enables comparison of post-fire to pre-fire conditions. Further, we plan to find a method for extracting burn severity through: hyperspectral and SAR images if they will be available, but the primary research activity is focused on multispectral methods. Suited Remote Sensing Data

4. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy Burn Severity Scale/CBI No DamageLowMediumHigh 00.511.52.02.53.0 In Precedent Studies field-based indices have been introduced (CBI, geoCBI) based on a visual assessment of the quantity of fuel consumed. These indices were correlated to spectral indices based on multi-spectral images. (dNBR or the RdNBR). Radiative Tansfer Model and burn severity simulation State of art

5. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy A comprehensive approach to monitor burn severity on the landscape is composed of four interrelated elements: the definition of severity the algorithm for burn severity extraction the field measures to calibrate and/or validate remote sensing results; the implementation of a support chain which deliver product to users. Each element influences the others, and PREFER attempts to integrate these in a unified system. Objective

6. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy There is still some discrepancy in the way researchers and managers use the term “burn severity.” We define BURN SEVERITY as the degree of environmental change caused by fire, or how much fire has affected the ecological community. Fire Intensity and Burn Severity

7. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy Initial Assessments post-event image required as soon after fire as possible. In this case we register many fire effects, but likely miss the bulk of green-up from plants that survived fire. Extended Assessments post-event image acquired one growing season after include survivorship of plants that burned, and may be most relevant to the actual ecological severity of the burn. Acquisition Time importance

8. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy Burn severity at pixel level Pixel concept Vertical variability Horizontal variability

9. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy How fire changes a real site in nature is very complex, many response variables can be measured in order to define the change. Moreover, the site may be structurally composed of many strata. Terefore, the severity detected at this level (30 m) is an aggregate of many variables over many components of the site. This concept of severity is what we attempt to capture by high- resolution images. Conclusion

10. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy MULTISPECTRAL ALGORITHM

11. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy PRE-FIRE ----- Mean ----- Mean+/- std ----- Max/Min POST-FIRE ----- Mean ----- Mean+/- std ----- Max/Min Burn severity and Multi-Spectral High Resolution Data GOLFO ARANCI AREA OLI

12. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy 1 1 NBR

13. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy DNBR

14. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy 1 Pre-Fire CIR Post-Fire CIR DNBR -0.2 1 DNBR SOUTH OF SARDINIA OLI

15. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy Damage Level No Damage: the area is indistinguishable from pre-fire conditions. Low Damage: little change in cover and mortality of structurally vegetation Medium Damage: mixture of effects ranging in the pixel from low to high change High Damage: Vegetation has high to complete mortality. Burn Severity Map KompsatRapidEye PREFER BURNED AREA GOOGLE EARTH

16. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy Improvement to Burn severity Map DNBR tend to saturate. Our idea to improve DNBR consists in computing several indices, each one capable to assess different characteristics of the vegetation and possibly capable to evaluate the effect on it of fire.

17. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy Golfo Aranci

18. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy San Basilio

19. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy HYPERSPECTRAL ALGORITHM

20. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy The possibility provided by hyperspectral images to compute several indices In this study we want to base the severity of the damage based on physical measurements that can be measured in field and at the same time that can be estimated by hyperspectral satellite imagery. Spectral signatures were collected in field on two areas test representative of the typical Mediterranean vegetation. Methodology

21. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy Lunghezza d’onda Riflettanza Field data analysis

22. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy Field data analysis The spectra collected during the campaign have joined to a picture, this allows to visualize the examined vegetation damage level. Inverse radiative transfer model to extract biophysical characteristics from field spectral measures. The analysis results showed that: Cab chlorophyll content in µg.cm-2, Car carotenoid content µg.cm-2, Cbrown brown pigment content (%), Cw Equivalent Water Ticknes (cm), Cm Leaf Mass Area( LMA) in (g.cm-2 and Leaf Area Index ( LAI) leaf area index; are the best representative biophysical parameters for damage severity levels. LAI Cab Cw Car Cbrown Cm LAI Cab Cw Car Cbrown Cm

23. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy TRAINING SET Field Data Pre- processing Modtran Linear Mixing Radiative Transfer Model Sensor Transfer Function Biophysical parameter Biophysical parameter Simulate image Burned Area Simulation

24. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy Future Developments Multispectral - Improve the algorithm for damage assessment - Develop an automatic method for downloading (if possible) and pre-processing Landsat 8 images and calculate Burn Severity. - Definition of the strategy for the extended assessment of damages. Hyperspectral - Complete the algorithm for image simulation - Develop a method for burn severity assessment

25. PREFER 1 st Annual Review Meeting, 5-6 Dec 2013, Milano-Italy THANK YOU