Assessment of the improvements of a land process model through assimilation of LAI maps derived by Cosmo Sky-Med images

1Rinaldi M., 2Satalino G., 2Mattia F., 1Ruggieri S., 1 Consiglio per la Ricerca e la Sperimentazione in Agricoltura (CRA), Centro di Ricerca per la Cerealicoltura (CER), Foggia, Italy 2 Consiglio Nazionale delle Ricerche (CNR) – Istituto di Studi sui Sistemi Intelligenti per l’Automazione (ISSIA), Bari, Italy

Introduction AQUATER software, as a prototype of a Decision System Support (DSS), has been developed in order to schedule irrigation at district level in a Mediterranean area. The DSS integrates the information deriving from soil and climatic database with a crop simulation model. It simulates the most representative field crops in Southern Italy. Simulation output variables should be used as useful information to estimate the crop water requirement and irrigation needs at regional level. The DSS requires a detailed input data set, as pedological, meteorological, crops related data. To improve the simulation of different crops in a large area, a model input variable derived from remote sensing technology is needed. The main crop yield driving variable, allowing a derivation from remote images, is the Leaf Area Index (LAI), expressed as leaf area (m2) on ground area (m2) ratio. The objective of this paper was to assess the improvements of a land process model as a result of the integration of the COSMO-SkyMed-derived LAI maps.

Material and Methods The assessment of crop growth model implemented in the DSS AQUATER has been carried out with a spatial simulation of approximately 130 km2 in the district of Capitanata plain, using weather daily data. The comparison with measured values (ground truth) was carried out with 4 fields of durum wheat and sugar beet and 7 fields of processing tomato. Different combinations of LAI maps assimilations were assessed in order to obtain a final yield simulation closer to the measured values.

Results The validation study demonstrated that X-band COSMO-SkyMed data can be successfully used for the LAI retrieval, and, in particular, the HV (or VH) polarization holds the highest potential for such an application. For durum wheat the benefit of LAI assimilation, in terms of yield forecast, is marginal (if any), essentially because the DSS performed extremely well even without any additional LAI-information. This is likely due to the fact that CRA has accumulated an outstanding experience in modelling and simulating the growth and development of cereal crops characteristic of the Capitanata plain, however, different results may be obtained over areas not yet deeply investigated (e.g. Nord Africa); For sugar beet and tomato, the assimilation of, either measured or COSMO-SkyMed-derived, LAI maps indicated a significant improvement in terms of yield forecast, especially when the COSMO-SkyMed-derived LAI maps covered the most critical growth periods. In particular, for autumnal sugar beet the best period of assimilation resulted at the beginning of spring growth (April), when the LAI is about 2.5-3.5 m2 m-2, and the optimal number of assimilation values is one or two. Whereas, for tomato, due to its very short growing period (i.e. 3 months), the assimilation of highest number of LAI-maps provides the best result.

Conclusions The assimilation of LAI derived from COSMO-SkyMed, produced a good improvement for sugar beet and processing tomatoes yield, especially if assimilation occurred during specific growth periods. With adequate LAI assimilation values derived by COSMO SkyMed, the DSS can be run at district levels with a good accuracy in simulation several crops in hundreds of fields, obtaining yield and water indicators useful for yield forecasting and a better distribution of irrigation water.