|Paper title||Deriving high-resolution soil moisture-based drought indices from remote sensing: towards resilient irrigation practices|
|Form of presentation||Poster|
Climate change has greatly altered the occurrence of extreme events such as droughts, floods and wildfires in the past years. Dire consequences of intense drought have been affecting dryland crop yield production. Some of the areas (like the Mediterranean and the Sahel) have been shown to be more prone to climate change and thus droughts and their consequences are expected to exacerbate in the future. Soil moisture (SM) data was shown to be key in the detection of early onset drought. Current drought observation warning systems, such as the European Drought Observatory, the Global Drought Observatory, or the U.S. Drought Monitor offer maps of a combined drought index, derived from different data sources (meteorological and satellite measurements and models). SM anomaly is acknowledged to be a good metric for drought and consequently all the global Drought Observatories include remote sensing (RS) SM but at a low spatial resolution. Consequently, regional drought events are frequently not captured or their intensity is not fully pictured.
In order to detect the onset of crop water stress and to trigger irrigations to mitigate the effects of potential droughts, in situ SM measurements are used by modern irrigators. Unfortunately, they are costly; combined with the fact that they are available only over small areas and that they might not be representative at the field scale, remote sensing is a cost-effective approach for mapping and monitoring extended areas.
This study focuses on a new pilot project which has been implemented over two areas located in the Tarragona province of Catalonia, Spain, whose main aim is to help resilient irrigation practices by offering advice based on drought indices. For this purpose, spatialized drought indices at high (1 km) resolution from remotely sensed SM are derived on a weekly basis. These indices are then used to provide irrigation recommendations to farmers, which have recently switched from dryland crops to vineyards.
Different indices, such as the Palmer Drought Severity Index, the Crop Moisture Index, the Standardized Precipitation Index or the soil moisture deficit index (SMDI) have been developed in literature in order to provide insight on agricultural drought monitoring and forecasting. Most of the existing well-known drought indices have been developed in conjunction with hydrological and meteorological models, i.e., use parameters such as rainfall, evapotranspiration, run-off and other indicators derived from models in order to give a comprehensive picture for decision-making. When used in conjunction with remote sensing-derived parameters, certain artefacts can appear in the drought indices, brought about by the high variability of remotely sensed data in comparison with model data. More specifically, the presence of outliers can have a high impact on the remote-sensing derived drought indices. This study has focused on analysing the presence and the impact of such outliers in the computation of SMDI. High resolution (1 km) SMOS (Soil Moisture and Ocean Salinity) and SMAP (Soil Moisture Active Passive) SM were first derived by using the DISPATCH (DISaggregation based on a Physical and Theoretical scale CHange) methodology. Furthermore, high-resolution root zone soil moisture (RZSM) products were then derived from the 1 km surface SM (SSM) by applying a recursive formulation of an exponential filter. Both SSM and RZSM were consequently used in order to derive SMDI representative of both the surface and root zone layer, on a weekly basis, for a period spanning 2010-2021, for the two areas of the above-mentioned pilot project. In the computation of SMDI for a certain week belonging to a certain month, the historical maximum, minimum and median of the month in question are used. The presence of outliers in the historical maximum and minimum have been identified, after a close inspection of the estimated SMDI using the original definition. The outliers are in line with the nature of the sensor used to measure SM remotely, which naturally is more noisy than the in situ sensors. Therefore, a new strategy has been developed, which uses percentiles in order to compute values corresponding to a “maximum” and a “minimum”, which are not affected by the outliers. Results have shown that by using percentiles instead of directly the maximum and minimum values, the artefacts present in the SMDI have been mitigated. Moreover, when comparing the “corrected” SMDI derived from SSM with the “corrected” SMDI derived from RZSM, the results show that the SMDI based on RZSM is more representative of the hydric stress level of the plants, given that the RZSM is better suited than the SSM to describe the moisture conditions at the deeper layers, which are the ones used by plants during growth and development.
The study provides an insight into obtaining robust, high-resolution derived drought indices based on remote-sensing derived SSM and RZSM estimates, for the improvement of resilient irrigation techniques. With the SSM-derived SMDI being currently used operationally and the RZSM-derived SMDI planned to be available soon, any improvement in the SMDI estimates will further improve irrigation advice.