Search

• A Machine Learning method is developed by integrating remotely sensed feedback of vegetation to antecedent rainfall to forecast seasonal soil moisture. • The method showed (>50%) and (>80%) variability in root-zone(0–1 m) and deep (1–6 m) soil can be explained 3 months ahead of time. • Integrating remotely sensed vegetation feedback to antecedent rainfall has improved soil moisture forecast by up to 33.3% • This model could be a steppingstone to developing and deploying a computationally light operational seasonal soil moisture forecast model. Seasonal forecast of soil moisture at large spatial scale over forested landscape has numerous implications in forest hydrology and bushfire risk planning. Remotely sensed plant response to rainfall, input meteorological forcing, and site-specific landscape attributes were integrated into a data-driven Gradient Boosting Machine Learning (ML) model to forecast summer season (December to February) soil moisture equivalent to those from Australian Water Resource Assessment-Landscape (AWRA-L) at root-zone (0–1 m) and deep (1–6 m) layers of the soil. Multispectral and thermal infrared bands from MODIS (Band 1–12, LST (day, night)) and meteorological forcing for 2000 – 2018 during and prior to winter (before August) and site-specific landscape attributes were used to generate the explanatory input variables to the model. Spatial and temporal forecasting skills of the model were evaluated using two types of cross-validation: two-fold cross-validation by space over the entire period (2000–2018) and two-fold cross-validation by time for all grid cells (2160 cells). In the first method, the model showed a high skill to forecast deep soil moisture (R2 = 0.81, NSE = 0.79, RMSE = 58.01 mm) and root-zone soil moisture (R2 = 0.65, NSE = 0.63, RMSE = 24.34 mm). In comparison, the second evaluation resulted in more accurate forecast deep-soil moisture (R2 = 0.88, NSE = 0.86, RMSE = 47.39 mm) but in lower performance for root-zone soil (R2 = 0.55, NSE = 0.47, RMSE = 29.2 mm). These outcomes have highlighted that the (>80%) variability in summers deep (1–6 m) soil moisture and (>50%) variability in root-zone (0–1 m) soil moisture over forested landscape can be explained at the end of winters. Furthermore, integrating remotely sensed observations has improved the results by showing an enhancement in the RMSE by 22.8–33.3% in the deep (1–6 m) soil moisture and 8.8–14.3% in root-zone (0–1 m) soil moisture. Overall, the integrated system using remotely sensed plant response, climate forcing and machine learning exhibit great potential to forecast summer soil moisture three months ahead in the Mediterranean climates. [ABSTRACT FROM AUTHOR]

Rice is one of the major crops cultivated predominantly in flooded paddies, thus a large amount of water is consumed during its growing season. Accurate paddy rice maps are therefore important inputs for improved estimates of actual evapotranspiration in the agricultural landscape. The main objective of this study was to obtain flooded paddy rice maps using multi-temporal images of Moderate Resolution Imaging Spectroradiometer (MODIS) in the Krishna River Basin, India. First, ground-based spectral samples collected by a field spectroradiometer, CROPSCAN, were used to demonstrate unique contrasts between the Normalized Difference Vegetation Index (NDVI) and the Land Surface Water Index (LSWI) observed during the transplanting season of rice. The contrast between Enhanced Vegetation Index (EVI) and Land Surface Water Index (LSWI) from MODIS time series data was then used to generate classification decision rules to map flooded rice paddies, for the transplanting seasons of Kharif and Rabi rice crops in the Krishna River Basin. Consistent with ground spectral observations, the relationship of the MODIS EVI vs. LSWI of paddy rice fields showed distinct features from other crops during the transplanting seasons. The MODIS-derived maps were validated against extensive reference data collected from multiple land use field surveys. The accuracy of the paddy rice maps, when determined using field plot data, was approximately 78%. The MODIS-derived rice crop areas were also compared with the areas reported by Department of Agriculture (DOA), Government of India (Government Statistics). The estimated root mean square difference (RMSD) of rice area estimated using MODIS and those reported by the Department of Agriculture over 10 districts varied between 3.4% and 6.6% during 10 years of our study period. Some of the major factors responsible for this difference include high noise of the MODIS images during the prolonged monsoon seasons (typically June-October) and the coarse spatial resolution (500 m) of the MODIS images compared to the small crop fields in the basin. However, this study demonstrates, based on multi-year analysis, that MODIS images can still provide robust and consistent flooded paddy rice extent and areas over a highly heterogeneous large river basin. [ABSTRACT FROM AUTHOR]

The temperature vegetation dryness index (TVDI) has been commonly implemented to estimate regional soil moisture in arid and semi-arid regions. However, the parameterization of the dry edge in the TVDI model is performed with a constraint to define the maximum water stress conditions. Mismatch of the spatial scale between visible and thermal bands retrieved from remotely sensed data and terrain variations also affect the effectiveness of the TVDI. Therefore, this study proposed a new drought index named the condition vegetation drought index (CVDI) to monitor the temporal and spatial variations of soil moisture status by substituting the land surface temperature (LST) with the modified perpendicular drought index (MPDI). In situ soil moisture observations at crop and pasture sites in Victoria were used to validate the effectiveness of the CVDI. The results indicate that the dry and wet edges in the parameterization scheme of the CVDI formed a better-defined trapezoid shape than that of the TVDI. Compared with the MPDI and TVDI for soil moisture monitoring at crop sites, the CVDI exhibited a performance superior to the MPDI and TVDI in most days where the coefficients of determination (R2) achieved can reach to 0.67 on DOY023, 137, 274 and 0.71 on DOY 322 and reproduced more accurate spatial and seasonal variation of soil moisture. Moreover, the CVDI showed higher correlation with the Australian Water Resource Assessment Landscape (AWRA-L) soil moisture product on temporal scales. The R2 can reach to 0.69 and the root mean square error (RMSE) is also much better than that of the MPDI and TVDI. Overall, it can be concluded that the CVDI appears to be a feasible method and can be successfully used in regional soil moisture monitoring. [ABSTRACT FROM AUTHOR]

Floodplains are a key ecological feature of arid and semi-arid regions and flooding is the main source of vegetation productivity. Characterizing the spatiotemporal flood dynamics, such as surface water and changes in subsequent vegetation growth vigour and biomass, is essential in better understanding of ecology and hydrology of these regions. Recent remotely sensed flooding studies of arid zone floodplains have concentrated on improving classification of surface water, particularly in mixed water-vegetation areas; less is known about the effect of surface water flooding pattern on emergent vegetation dynamics. We use an integrated framework for mapping both flood extents and the persistence of floodplain response changes of water, vegetation and moisture in Cooper Creek, Australia. We analysed pixel-based time series of multiple indices generated by daily MODIS data for 14 highly variable flow flood events between 2000 and 2012. Results indicate that for the extremely flat Cooper Creek floodplain, mapping inundation area by changes in vegetation growth vigour and biomass was significantly larger than surface water mapping area (16.2% for the 2004 flood event) and the difference in inundation mapping mainly occurs around the inundated edges. In addition, by studying surface water and subsequent vegetation response together, it is possible to generate new information, such as the lag time between flooding and peak vegetation growth vigour and biomass, and persistence time of surface water and green vegetation, which provide important insights to arid zone floodplain behaviour. The large extent and high frequency of MODIS images provide advantages in characterizing inundation dynamics for large-scale floodplains where instantaneous (daily) inundation extent is considerably smaller than total cumulative inundation extent, compared with sensors with higher spatial resolution but lower temporal resolution; however, the coarse resolution of MODIS (500 m) limits its performance for small flood events. Globally, this approach is suitable for other large, low-gradient floodplains in arid zones that show similar, long duration vegetation responses as observed in Cooper Creek. [ABSTRACT FROM AUTHOR]

[Display omitted] • In asynchronous climates, the forest growth response to rainfall is not well known. • Lagged relationship of forest growth to antecedent rainfall was explored. • Vegetation in the asynchronous climate have a long memory to the rainfall. • The memory of vegetation to rain could indicate soil/rooting depth across the landscape. The out-of-phase rainfall and temperature and deep root system make the sequential connection between past rainfall events, soil water storage, and forest growth response complicated and temporally extended in asynchronous climates with Mediterranean-type settings. Unfortunately, these location-specific deep-soil water stores are rarely measured due to logistic and financial constraints, especially in the forest. Therefore, at a large spatio-temporal scale, forest growth relationship to growth drivers is still unknown in these ecosystems, limiting our knowledge to understand the functioning of these forests and their links with hydrological processes. Although process-based water balance models can analyze vegetation growth response to the input climate forcing, they rely upon some significant assumptions regarding plants access to deep soil water storage. Thus, this study aims to understand how the out-of-phase rainfall events affect the current ecosystem growth response, represented by the observed Normalized Difference Vegetation Index (NDVI), across the landscape. We have used an empirical approach on long term observed data without any assumption on access to deep-soil water stores. We estimated time lags between forest growth and rainfall events using a lagged correlation analysis applied to monthly anomalies of NDVI and rainfall against their climatological averages over 2002–2018. The study found that the forests in asynchronous climates exhibit unexpectedly long (10–25 months) memory to rain, and this memory has a systematic pattern across the landscape, which we contend has highlighted three things: 1) the forest in the middle aridity (∼3–4.5) range are relatively more sensitive to changes in the short-term rainfall than the forest in lower (<3) and higher (>5) aridity regions, could be due to rapid depletion of relatively small soil water storage in between the storms, 2) the variable memory of forest to rain across the landscape can be an indicator of soil depth/rooting depth, and 3) the variable sized location specific antecedent rainfall windows can explain significant variability in forest growth status in asynchronous climates, thus these rainfall windows can be employed to forecast forest growth with a lead time (>4 months). [ABSTRACT FROM AUTHOR]