Microwave Satellite Data Research Articles

Microwave Brightness Temperature Anomalies Associated With the 2015 Mw 7.8 Gorkha and Mw 7.3 Dolakha Earthquakes in Nepal

Two catastrophic earthquakes (EQs) with magnitudes of Mw 7.8 and 7.3 occurred in Gorkha and Dolakha, Nepal on April 25 and May 12, 2015, respectively. By employing the spatio-temporally weighted two-step method (STW-TSM), significant positive and negative microwave brightness temperature (MBT) anomalies preceding the two EQs were uncovered with satellite microwave data from FY-3B. Two strip-shaped positive MBT anomalies appeared in the Higher Himalayan 1 day before the Gorkha shock and the Dolakha shock, respectively, which were parallel to the thrust belt and whose intensity distributions exhibited good topographic consistency with the elevation profiles of the central Himalayas. Referring to the P-hole theory and seismogenic mechanism of the Nepal EQs, seismically activated mobile positive charges are presumed to have caused the surface positive MBT anomalies by reducing the superficial dielectric constant and increasing the local temperature of the Himalayan cliffs. A significant negative MBT anomaly occurred plausibly as early as March 2015 in the southern Gangetic Plain and developed into a stripe parallel to the thrust belt on the days of the Gorkha and Dolakha events. The negative MBT anomalies are suggested to result from soil moisture increases due to local synchronous precipitation. This work presents the first case study to use data from the FY-3B microwave radiation imager (MWRI) to investigate seismic anomalies, and discriminate both positive and negative MBT anomalies preceding an EQ for the first time.

Synergistic utilization of optical and microwave satellite data for coastal bathymetry estimation

The present study adopts a multi-sensor approach that utilizes optical remote sensing and Synthetic Aperture Radar (SAR) data complimentarily for coastal bathymetry estimation. The depths between 0–15 m are derived from Landsat 8 OLI imageries, by using the machine learning approach of Support Vector Regression (SVR); and high accuracies with RMSEs ranging between 0.17 m−0.95 m are obtained for the chosen study regions. Further, ALOS PALSAR images are used to obtain depths beyond 15 m, by applying the Wave Based Approach (WBA) which is based on the wave transformation and linear dispersion principles. The estimated depths are compared with the bathymetry obtained from admiralty charts or GEBCO and RMSEs between 0.83–1.34m are observed. Finally, the outputs of the SVR and WBA are combined to derive depths between 0–50 m at 75 m spatial resolution. The resultant depth maps agree with those obtained from GEBCO, with an additional advantage of higher accuracy in the near-shore region.

The 2020 Larsen C Ice Shelf surface melt is a 40-year record high

Abstract. Along with record-breaking summer air temperatures at an Antarctic Peninsula meteorological station in February 2020, the Larsen C ice shelf experienced an exceptionally long and extensive 2019/2020 melt season. We use a 40-year time series of passive and scatterometer satellite microwave data, which are sensitive to the presence of liquid water in the snow pack, to reveal that the extent and duration of melt observed on the ice shelf in the austral summer of 2019/2020 was the greatest on record. We find that unusual perturbations to Southern Hemisphere modes of atmospheric flow, including a persistently positive Indian Ocean Dipole in the spring and a very rare Southern Hemisphere sudden stratospheric warming in September 2019, preceded the exceptionally warm Antarctic Peninsula summer. It is likely that teleconnections between the tropics and southern high latitudes were able to bring sufficient heat via the atmosphere and ocean to the Antarctic Peninsula to drive the extreme Larsen C Ice Shelf melt. The record-breaking melt of 2019/2020 brought to an end the trend of decreasing melt that had begun in 1999/2000, will reinitiate earlier thinning of the ice shelf by depletion of the firn air content, and probably affected a much greater region than Larsen C Ice Shelf.

Evaluation of a New Merged Sea-Ice Concentration Dataset at 1 km Resolution from Thermal Infrared and Passive Microwave Satellite Data in the Arctic

Sea-ice concentration (SIC) data with fine spatial resolution and spatially continuous coverage are needed, for example, for estimating heat fluxes. Passive microwave measurements of the Advanced Scanning Microwave Radiometer 2 (AMSR2) offer spatial continuity, but are limited to spatial resolutions of 5 km and coarser. Thermal infrared data of the Moderate Resolution Imaging Spectroradiometer (MODIS) provide a spatial resolution of 1 km, but are limited to cloud-free scenes. We exploit the benefits of both and present a merged SIC dataset with 1 km spatial resolution and spatially continuous coverage for the Arctic. MODIS and AMSR2 SIC are retrieved separately and then merged by tuning the MODIS SIC to preserve the mean AMSR2 SIC. We first evaluate the variability of the dynamically retrieved MODIS ice tie-point. Varying the starting position of the area used for the tie-point retrieval changes the MODIS SIC by on average 1.9%, which we mitigate by considering different starting positions and using the average as ice tie-point. Furthermore, the SIC datasets are evaluated against a reference dataset derived from Sentinel-2A/B reflectances between February and May 2019. We find that the merged SIC are 1.9% smaller than the reference SIC if thin ice is considered as ice and 4.9% higher if thin ice is considered as water. There is only a slight bias (0.3%) between the MODIS and the merged SIC; however, the root mean square deviation of 5% indicates that the two datasets do yield different results. In an example of poor-quality MODIS SIC, we identify an unscreened cloud and high ice-surface temperature as reasons for the poor quality. Still, the merged SIC are of similar quality as the passive microwave SIC in this example. The benefit of merging MODIS and AMSR2 data is demonstrated by showing that the finer resolution of the merged SIC compared to the AMSR2 SIC allows an enhanced potential for the retrieval of leads. At the same time, the data are available regardless of clouds. Last, we provide uncertainty estimates. The MODIS and merged SIC uncertainty are between 5% and 10% from February to April and increase up to 25% (merged SIC) and 35% (MODIS SIC) in May. They are identified as conservative uncertainty estimates.

Validation of remotely sensed estimates of snow water equivalent using multiple reference datasets from the middle and high latitudes of China

A key variable describing the mass of seasonal snow cover is snow water equivalent (SWE), which plays an important role in hydrological applications, weather forecasting and land surface process simulations. In this paper, the accuracy of an SWE product, GlobSnow-2, which combines microwave satellite data and in situ measurements, is assessed using three reference evaluation datasets north of 35°N in China. The GlobSnow-2 estimates are also compared with stand-alone satellite products (AMSR2, Chang and FY-3D SWE). The overall unbiased root mean square error (RMSE) and bias of the GlobSnow-2 SWE product validated with three reference datasets are 17.4 mm and 11.2 mm, respectively, which outperforms the AMSR2 SWE (39.3 mm and 37.3 mm, respectively) and Chang SWE (57.5 mm and 46.2 mm, respectively) products. The FY-3D SWE product performs better than the GlobSnow-2 estimate for shallow snow (SWE < 50 mm) and tends to underestimate snow cover, particularly when SWE exceeds 80 mm. A retrieval sensitivity analysis against land cover types shows that the highest SWE uncertainties for GlobSnow-2 are exhibited in grassland (unbiased RMSE, 27.8 mm), and the most serious overestimation occurs in forested areas (bias, 23.6 mm). The GlobSnow-2 performances at various elevations show an increasing bias trend, ranging from 5 to 61 mm with increasing elevation. The GlobSnow-2 estimate analyses under different snow regimes show that the GlobSnow-2 SWE product performs best in taiga snow, with high uncertainties (unbiased RMSE, 28.3 mm) in prairie snow and serious overestimations (bias, 23.2 mm) for alpine snow. The results of this study demonstrate that the GlobSnow-2 assimilation approach tends to overestimate SWE in China. One of the major reasons that overestimations occur is that the GlobSnow-2 SWE retrieval scheme utilizes a fixed density of 240 kg/m3, which is larger than the average value derived from ground measurements for China (180 kg/m3), which undoubtedly contributes to the observed SWE overestimation. Another reason is that forest effects on satellite signals remain challenging for SWE estimations in the GlobSnow-2 assimilation system. The retrieval errors in prairie and alpine are also higher than others due to the snowpack stratigraphy and complex topography. The GlobSnow-2 SWE product performance is evaluated over China in this study, and the major factors that affect the assimilation scheme accuracy are determined. These results will provide a reference to improve the GlobSnow-2 SWE product in future work.

Ground-Truth Experiments for the Calibration and Validation of Satellite Microwave Radiometer Data

Ground-truth experiments are needed to calibrate and validate satellite microwave data and to improve the quality and utilization efficiency of satellite data in solving problems of hydrometeorological support. This paper considers the arrangement of specialized test observatories in the satellite data validation subsystem exemplified by the geophysical observatory in Lehtusi (a town in Leningrad oblast), describes the state and prospects of its equipping with modern tools of contact and remote measurements of meteorological parameters, and indicates directions of scientific research.

A Data-Intensive Approach to Address Food Sustainability: Integrating Optic and Microwave Satellite Imagery for Developing Long-Term Global Cropping Intensity and Sowing Month from 2001 to 2015

It is necessary to develop a sustainable food production system to ensure future food security around the globe. Cropping intensity and sowing month are two essential parameters for analyzing the food–water–climate tradeoff as food sustainability indicators. This study presents a global-scale analysis of cropping intensity and sowing month from 2000 to 2015, divided into three groups of years. The study methodology integrates the satellite-derived normalized vegetation index (NDVI) of 16-day composite Moderate Resolution Imaging Spectroradiometer (MODIS) and daily land-surface-water coverage (LSWC) data obtained from The Advanced Microwave Scanning Radiometer (AMSR-E/2) in 1-km aggregate pixel resolution. A fast Fourier transform was applied to normalize the MODIS NDVI time-series data. By using advanced methods with intensive optic and microwave time-series data, this study set out to anticipate potential dynamic changes in global cropland activity over 15 years representing the Millennium Development Goal period. These products are the first global datasets that provide information on crop activities in 15-year data derived from optic and microwave satellite data. The results show that in 2000–2005, the total global double-crop intensity was 7.1 million km2, which increased to 8.3 million km2 in 2006–2010, and then to approximately 8.6 million km2 in 2011–2015. In the same periods, global triple-crop agriculture showed a rapid positive growth from 0.73 to 1.12 and then 1.28 million km2, respectively. The results show that Asia dominated double- and triple-crop growth, while showcasing the expansion of single-cropping area in Africa. The finer spatial resolution, combined with a long-term global analysis, means that this methodology has the potential to be applied in several sustainability studies, from global- to local-level perspectives.

The Stationary Banding Complex and Secondary Eyewall Formation in Tropical Cyclones

Abstract89 GHz passive microwave satellite data are used to develop a five year climatology of the incidence and morphology of the stationary banding complex in tropical cyclones and quantify changes in convective morphology prior to secondary eyewall formation. The stationary banding complex is shown to be present in 39% of passive microwave overpasses. Morphology varies substantially between tropical cyclones, with crossing angles ranging from 0.19° to 61.78° and azimuthal extents from 0.29 to 4.02 radians. Variations in the incidence and geometry of the stationary banding complex are observed in different environmental conditions and geographic locations. For 84 secondary eyewall formation events included in the sample, a stationary banding complex is observed within 6 hr of the secondary eyewall developing in 79% of cases. Within 12 hr prior to secondary eyewall formation, the crossing angle is significantly lower than its sample median, while the azimuthal extent is higher than its sample median. These results demonstrate that secondary eyewall formation is (most) often preceded by the formation and axisymmetrization of a stationary banding complex.

Retrieving global surface soil moisture from GRACE satellite gravity data

Passive microwave radiometry from space through missions such as the Soil Moisture and Ocean Salinity (SMOS) and Soil Moisture Active Passive (SMAP) satellites is the most reliable means for mapping global surface soil moisture (SSM). Nonetheless, microwave SSM retrievals are uncertain over densely vegetated surfaces or areas with high radio frequency interference. This paper presents a new observationally driven approach to remote sensing of global SSM based on the terrestrial water storage anomaly (TWSA) data acquired from the Gravity Recovery and Climate Experiment (GRACE) satellite. This approach rests on a physically based, yet parsimonious, model based on the Richards’ equation and the assumption that the TWSA temporal rate of change (dS/dt) approximates the land surface net water flux (NWF) as the surface boundary condition. The GRACE-based SSM is found to be in a reasonable agreement with in-situ data and highly correlated with the SMAP and SMOS retrievals, especially over wet regions where the assumption of NWF ≈ dS/dt holds valid. The GRACE retrievals contain new SSM information relative to the microwave satellite data and provide a potential solution to improve the microwave data over densely vegetated surfaces or areas with high radio frequency interference.

Lagging recovery for tropical forests

Ecology Tropical forest aboveground carbon (AGC) stocks have yet to recover from the extremely hot and dry weather associated with the 2015–2016 El Nino event. Wigneron et al. used low-frequency microwave satellite data to monitor AGC changes from 2014 to 2017. By the end of 2017, AGC stocks had not reached 2014 predrought levels and continued to decline in some areas. The slow rate of recovery could be due, in part, to enhanced forest mortality and/or unaccounted deforestation and degradation, specifically in African humid forests. Such continuity in long-term records may improve understanding of climate change on ecosystems. Sci. Adv. 10.1126/sciadv.aay4603 (2020).

Overview of Researches on All-Sky Satellite Microwave Data Variational Assimilation

Overview of Researches on All-Sky Satellite Microwave Data Variational Assimilation

Mean European Carbon Sink Over 2010–2015 Estimated by Simultaneous Assimilation of Atmospheric CO2, Soil Moisture, and Vegetation Optical Depth

AbstractThe northern land biosphere is believed to be the main global sink of CO2, but the contribution of Europe is uncertain. While bottom‐up estimates and inverse atmospheric transport studies based on atmospheric CO2 observed in situ or from space by OCO‐2 point to a moderate rate of uptake, some other inversions based on remotely sensed atmospheric CO2 from GOSAT/SCIAMACHY and biomass estimates from passive microwave satellite data point to a large sink of around 1 Gt C/yr. We present results from combining both approaches in a data assimilation framework, inverting a biosphere model against in situ atmospheric CO2 and passive microwave measurements. When assimilating all observations, we estimate a European carbon sink of 0.303 ± 0.083 Gt C/yr for 2010–2015. The result agrees with other bottom‐up studies and atmospheric inversions using in situ CO2 or OCO‐2 observations pointing to potential data problems when using observations from GOSAT or SCIAMACHY to estimate the European CO2 sink.

Analysis of the Seasonal Dependence of the Brightness Temperature of the Glacier Sheet of Antarctica by Microwave Satellite Data

In this paper, an analysis of seasonal dependences of the brightness temperature of different regions of Antarctica according to radiometer data from MIRAS (SMOS satellite) and SSMIS (DMSP series satellites) is presented. Queen Maud Land (Norwegian: Dronning Maud Land) was chosen as the region of study, which includes the main zones of Antarctica: the dome, the zone of runoff winds, and the coastal zone. The interrelation of time dynamics of brightness temperature as a function of the change in climatic characteristics of regions is considered. The main factors influencing the change in brightness temperature in different regions of the Antarctic glacier sheet are revealed.

Comparison of Satellite Microwave and Visual Shipborne Data on Sea Ice Concentration

In this paper we compare the sea ice concentration data obtained by satellite microwave radiometry using the NASA Team, ASI, and VASIA2 algorithms and by special visual ship observations. A database of visual observations of the sea ice concentration obtained during 15 expeditions in the Arctic was used. The comparison for various concentration gradations (very open, open, close, very close, and compact ice) was carried out separately for summer and winter periods. On average, during the summer period, the NASA Team algorithm underestimates the sea ice concentration by 1 point, while the ASI and VASIA2 algorithms underestimate by 0.5 points when compared to ship observations. All three algorithms overestimate the total concentration in the zones of very open ice and underestimate it in the zones of close, very close, and compact ice. The maximal average errors are characteristic of open ice that is most often observed in ice edge zones. In winter, the average error for all algorithms is no more than 1 point when compared with ship observations. However, in very open and close, the average error in winter is significantly higher than that in summer. To assess the effect of melting processes on the concentration values obtained by satellite microwave radiometry, we used ship data on ice melt stage. The average error at the maximal area of ​​melt ponds on the ice surface reaches –2.9 points for the NASA Team algorithm, –2.8 points for the ASI algorithm, and –5.0 points for the VASIA2 algorithm. The results will be useful for determining the extent of ice-cover degradation in the Arctic Ocean observed in recent years.

Intercomparison of Snow Melt Onset Date Estimates From Optical and Microwave Satellite Instruments Over the Northern Hemisphere for the Period 1982–2015

AbstractRobust melt season timing and length estimates are important for hydrological and climatological applications; due to the large area and sparse in situ measurements, snow melt monitoring at the continental scale is only possible from satellites. We intercompared melt onset date (MOD) estimates obtained from optical and microwave satellite sensors over the Northern Hemisphere between 1982 and 2015 and subsequently analyzed the causes of the similarities and dissimilarities found. The optical satellite data are based on the mean surface albedo from the Satellite Application Facility for Climate Monitoring (CM SAF) CLouds, Albedo and RAdiation second release Surface ALbedo (CLARA‐A2 SAL) data set. The microwave satellite data are based on temporal variations in the differences of the brightness temperature from satellite passive microwave radiometers. The analysis shows that the microwave‐based method detects melt onset on average 10 days later than the albedo‐based method, which results from the different melt detection methods; the albedo‐based method observes the point when the spring snow metamorphism begins to have a detectable effect on snow albedo, whereas the microwave‐based method detects the appearance of meltwater in snowpack. The difference in MOD decreases in forests, because canopy protects snow from sunlight delaying snow metamorphism. Additionally, we analyzed the MOD estimates for trends across the Northern Hemisphere and separately for Eurasia and North America. A statistically significant negative trend toward earlier melt onset exists in all cases, which is consistent with previous studies.

Synergistic integration of optical and microwave satellite data for crop yield estimation

Developing accurate models of crop stress, phenology and productivity is of paramount importance, given the increasing need of food. Earth observation (EO) remote sensing data provides a unique source of information to monitor crops in a temporally resolved and spatially explicit way. In this study, we propose the combination of multisensor (optical and microwave) remote sensing data for crop yield estimation and forecasting using two novel approaches. We first propose the lag between Enhanced Vegetation Index (EVI) derived from MODIS and Vegetation Optical Depth (VOD) derived from SMAP as a new joint metric combining the information from the two satellite sensors in a unique feature or descriptor. Our second approach avoids summarizing statistics and uses machine learning to combine full time series of EVI and VOD. This study considers two statistical methods, a regularized linear regression and its nonlinear extension called kernel ridge regression to directly estimate the county-level surveyed total production, as well as individual yields of the major crops grown in the region: corn, soybean and wheat. The study area includes the US Corn Belt, and we use agricultural survey data from the National Agricultural Statistics Service (USDA-NASS) for year 2015 for quantitative assessment. Results show that (1) the proposed EVI-VOD lag metric correlates well with crop yield and outperforms common single-sensor metrics for crop yield estimation; (2) the statistical (machine learning) models working directly with the time series largely improve results compared to previously reported estimations; (3) the combined exploitation of information from the optical and microwave data leads to improved predictions over the use of single sensor approaches with coefficient of determination R≥20.76; (4) when models are used for within-season forecasting with limited time information, crop yield prediction is feasible up to four months before harvest (models reach a plateau in accuracy); and (5) the robustness of the approach is confirmed in a multi-year setting, reaching similar performances than when using single-year data. In conclusion, results confirm the value of using both EVI and VOD at the same time, and the advantage of using automatic machine learning models for crop yield/production estimation.

Development of an Operational Algorithm for Automated Deforestation Mapping via the Bayesian Integration of Long-Term Optical and Microwave Satellite Data

The frequent fine-scale monitoring of deforestation using satellite sensors is important for the sustainable management of forests. Traditional optical satellite sensors suffer from cloud interruption, particularly in tropical regions, and recent active microwave sensors (i.e., synthetic aperture radar) demonstrate the difficulty in data interpretation owing to their inherent sensor noise and complicated backscatter features of forests. Although the sensor integration of optical and microwave sensors is of compelling research interest, particularly in the conduct of deforestation monitoring, this topic has not been widely studied. In this paper, we introduce an operational algorithm for automated deforestation mapping using long-term optical and L-band SAR data, including a simple time-series analysis of Landsat stacks and a multilayered neural network with Advanced Spaceborne Thermal Emission and Reflection Radiometer and Phased Array-type L-band Synthetic Aperture Radar-2, followed by sensor integration based on the Bayesian Updating of Land-Cover. We applied the algorithm over a deciduous tropical forest in Cambodia in 2003–2018 for validation, and the algorithm demonstrated better accuracy than existing approaches, which only depend on optical data or SAR data. Owing to the cloud penetration ability of SAR, observation gaps of optical data under cloudy conditions were filled, resulting in a prompter detection of deforestation even in the tropical rainy season. We also investigated the effect of posterior probability constraints in the Bayesian approach. The land-cover maps (forest/deforestation) created by the well-tuned Bayesian approach achieved 94.0% ± 4.5%, 80.0% ± 10.1%, and 96.4% ± 1.9% for the user’s accuracy, producer’s accuracy, and overall accuracy, respectively. In the future, small-scale commission errors in the resultant maps should be improved by using more sophisticated machine-learning approaches and considering the reforestation effects in the algorithm. The application of the algorithm to other landscapes with other sensor combinations is also desirable.

Estimating Root Zone Soil Moisture Across the Eastern United States with Passive Microwave Satellite Data and a Simple Hydrologic Model

Root zone soil moisture (RZSM) affects many natural processes and is an important component of environmental modeling, but it is expensive and challenging to monitor for relatively small spatial extents. Satellite datasets offer ample spatial coverage of near-surface soil moisture content at up to a daily time-step, but satellite-derived data products are currently too coarse in spatial resolution to use directly for many environmental applications, such as those for small catchments. This study investigated the use of passive microwave satellite soil moisture data products in a simple hydrologic model to provide root zone soil moisture estimates across a small catchment over a two year time period and the Eastern U.S. (EUS) at a 1 km resolution over a decadal time-scale. The physically based soil moisture analytical relationship (SMAR) was calibrated and tested with the Advanced Microwave Scanning Radiometer (AMSRE), Soil Moisture Ocean Salinity (SMOS), and Soil Moisture Active Passive (SMAP) data products. The SMAR spatial model relies on maps of soil physical properties and was first tested at the Shale Hills experimental catchment in central Pennsylvania. The model met a root mean square error (RMSE) benchmark of 0.06 cm3 cm−3 at 66% of the locations throughout the catchment. Then, the SMAR spatial model was calibrated at up to 68 sites (SCAN and AMERIFLUX network sites) that monitor soil moisture across the EUS region, and maps of SMAR parameters were generated for each satellite data product. The average RMSE for RZSM estimates from each satellite data product is &lt;0.06 cm3 cm−3. Lastly, the 1 km EUS regional RZSM maps were tested with data from the Shale Hills, which was set aside for validating the regional SMAR, and the RMSE between the RZSM predictions and the catchment average is 0.042 cm3 cm−3. This study offers a promising approach for generating long time-series of regional RZSM maps with the same spatial resolution of soil property maps.

Automatically Extracted Antarctic Coastline Using Remotely-Sensed Data: An Update

The temporal and spatial variability of the Antarctic coastline is a clear indicator of change in extent and mass balance of ice sheets and shelves. In this study, the Canny edge detector was utilized to automatically extract high-resolution information of the Antarctic coastline for 2005, 2010, and 2017, based on optical and microwave satellite data. In order to improve the accuracy of the extracted coastlines, we developed the Canny algorithm by automatically calculating the local low and high thresholds via the intensity histogram of each image to derive thresholds to distinguish ice sheet from water. A visual comparison between extracted coastlines and mosaics from remote sensing images shows good agreement. In addition, comparing manually extracted coastline, based on prior knowledge, the accuracy of planimetric position of automated extraction is better than two pixels of Landsat images (30 m resolution). Our study shows that the percentage of deviation (&lt;100 m) between automatically and manually extracted coastlines in nine areas around the Antarctica is 92.32%, and the mean deviation is 38.15 m. Our results reveal that the length of coastline around Antarctica increased from 35,114 km in 2005 to 35,281 km in 2010, and again to 35,672 km in 2017. Meanwhile, the total area of the Antarctica varied slightly from 1.3618 × 107 km2 (2005) to 1.3537 × 107 km2 (2010) and 1.3657 × 107 km2 (2017). We have found that the decline of the Antarctic area between 2005 and 2010 is related to the breakup of some individual ice shelves, mainly in the Antarctic Peninsula and off East Antarctica. We present a detailed analysis of the temporal and spatial change of coastline and area change for the six ice shelves that exhibited the largest change in the last decade. The largest area change (a loss of 4836 km2) occurred at the Wilkins Ice Shelf between 2005 and 2010.

SYNERGETIC USE OF OPTICAL, MICROWAVE AND THERMAL SATELLITE DATA FOR NON-PARAMETRIC ESTIMATION OF WHEAT GRAIN YIELD

Abstract. Crop yield maps are very crucial inputs for different practical applications like crop production estimation, pay-out of crop insurance, yield gap analysis etc. Satellite derived vegetation indices across different electromagnetic region has the ability to explain the variation in crop yield and can be used for prediction of yield before harvesting. This study utilised indices derived from multi-temporal Optical, Thermal and Radar data for developing model for Wheat (Triticum aestivum) grain yield using Machine learning approaches i.e., Random Forest Regression (RFR). Time series of Sentinel-2 derived Normalized difference vegetation index (NDVI), Normalized difference water Index (NDWI), Landsat-8 derived GPP using LST-EVI relationship (Temparature-Greeness model) and Sentinel-1 derived cross-polarization backscatter ratio (σVH/σVV) were used as predictor for wheat yield estimation. Actual grain yield measurements at ground were carried out at the end of the season over 178 locations. Seventy five percent of ground yield data were used for training of the model and rest twenty five percent data were used for its validation. All the datasets were grouped into ten fortnightly datasets ranging from November 2017 to March 2018. Through the random forest regression using time-series of NDVI alone, wheat grain yields were estimated with an RMSE of 9.8 Q ha−1. Subsequently by adding the multi-temporal NDWI, GPP and σVH/σVV led to the improvement of RMSE to 8.7, 7.6 and 7.4 Q ha−1 respectively. Variable importance based on the out of box error showed the significance of NDVI, NDWI and GPP during Dec-Jan and σVH/σVV during Feb for wheat grain estimation. It was concluded that the RFR algorithm together with the indices from optical, thermal and microwave satellite data can able to produced significantly accurate estimates of wheat grain yield.